1
00:00:00,840 --> 00:00:03,180
The following content is
provided under a Creative

2
00:00:03,180 --> 00:00:04,570
Commons license.

3
00:00:04,570 --> 00:00:06,780
Your support will help
MIT OpenCourseWare

4
00:00:06,780 --> 00:00:10,870
continue to offer high quality
educational resources for free.

5
00:00:10,870 --> 00:00:13,440
To make a donation or to
view additional materials

6
00:00:13,440 --> 00:00:17,400
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:17,400 --> 00:00:18,286
at ocw.mit.edu.

8
00:00:21,033 --> 00:00:22,783
GABRIEL SANCHEZ-MARTINEZ:
So today's topic

9
00:00:22,783 --> 00:00:24,160
is workforce planning.

10
00:00:24,160 --> 00:00:26,410
And we have Professor
Nigel Wilson,

11
00:00:26,410 --> 00:00:29,970
who is the father of
the Transit Lab here

12
00:00:29,970 --> 00:00:31,960
at MIT and of this course.

13
00:00:31,960 --> 00:00:36,490
He's sort of the master of
everything related to transit

14
00:00:36,490 --> 00:00:39,315
and the applications
of transit data,

15
00:00:39,315 --> 00:00:41,560
to operations
planning and study.

16
00:00:41,560 --> 00:00:47,760
You may remember that we
divided the course into themes.

17
00:00:47,760 --> 00:00:50,410
And so we already talked
about crew scheduling

18
00:00:50,410 --> 00:00:52,495
in the context of
operations planning.

19
00:00:52,495 --> 00:00:56,380
And now we're in the
strategy and policy theme.

20
00:00:56,380 --> 00:01:03,140
So this lecture is on looking,
again, at proofs and planning,

21
00:01:03,140 --> 00:01:05,950
but from a strategic
perspective.

22
00:01:05,950 --> 00:01:07,480
I think that's probably--

23
00:01:07,480 --> 00:01:08,040
NIGEL HM WILSON: That's great.

24
00:01:08,040 --> 00:01:08,900
GABRIEL SANCHEZ-MARTINEZ:
--enough of an introduction.

25
00:01:08,900 --> 00:01:09,670
Maybe you can provide--

26
00:01:09,670 --> 00:01:09,940
NIGEL HM WILSON: Yeah.

27
00:01:09,940 --> 00:01:11,350
GABRIEL SANCHEZ-MARTINEZ:
--a little more.

28
00:01:11,350 --> 00:01:12,850
NIGEL HM WILSON:
Thank you, Gabriel.

29
00:01:12,850 --> 00:01:14,950
It's good to see a fair
number of familiar faces

30
00:01:14,950 --> 00:01:17,110
and some new ones
I don't know yet.

31
00:01:17,110 --> 00:01:20,050
So I encourage you
all to jump in and ask

32
00:01:20,050 --> 00:01:22,780
questions if I'm not making
things clear or if you disagree

33
00:01:22,780 --> 00:01:24,130
or whatever.

34
00:01:24,130 --> 00:01:26,590
But I'm delighted to
be here with you today.

35
00:01:26,590 --> 00:01:29,080
So the work that I'm
going to be discussing

36
00:01:29,080 --> 00:01:32,800
is work that I
haven't done recently.

37
00:01:32,800 --> 00:01:35,620
It's about workforce planning.

38
00:01:35,620 --> 00:01:37,330
And it was done with the MBTA.

39
00:01:37,330 --> 00:01:40,990
Principally the results
I'll give are for the MBTA

40
00:01:40,990 --> 00:01:43,660
as it existed 20-odd years ago.

41
00:01:43,660 --> 00:01:47,710
But the issues are exactly
the same then as they are now.

42
00:01:47,710 --> 00:01:58,880
In fact, I gave a summary of
this presentation to Brian

43
00:01:58,880 --> 00:02:03,506
Shortsleeve and some of the
other people at the MBTA

44
00:02:03,506 --> 00:02:05,850
a year and a half ago, or so.

45
00:02:05,850 --> 00:02:07,289
And they were
thinking about maybe

46
00:02:07,289 --> 00:02:09,289
they really should
[AUDIO OUT] --thing something

47
00:02:09,289 --> 00:02:10,560
like this again.

48
00:02:10,560 --> 00:02:14,150
So what I want to discuss
the problem in general

49
00:02:14,150 --> 00:02:18,518
of the approach that
we've taken to this,

50
00:02:18,518 --> 00:02:20,810
and then look at three
different levels of the problem.

51
00:02:20,810 --> 00:02:22,640
One is the strategic
level, which

52
00:02:22,640 --> 00:02:25,340
is the long-term
view, which fits

53
00:02:25,340 --> 00:02:28,040
in with Gabriel's
statement about the theme

54
00:02:28,040 --> 00:02:30,620
of this part of the
course being on strategy.

55
00:02:30,620 --> 00:02:34,910
Then the tactical level,
which is shorter range.

56
00:02:34,910 --> 00:02:38,040
And then the operational level,
which is really short range.

57
00:02:38,040 --> 00:02:41,062
And these are all tightly
connected and interrelated,

58
00:02:41,062 --> 00:02:42,770
as I hope I'll be able
to explain to you.

59
00:02:46,310 --> 00:02:49,860
OK, so the challenge
that we have

60
00:02:49,860 --> 00:02:53,780
in this part of the
operations planning

61
00:02:53,780 --> 00:02:56,660
process, if you think
of it that way, is we

62
00:02:56,660 --> 00:02:58,100
want to provide good service.

63
00:02:58,100 --> 00:03:01,010
We've got hopefully a
decent operating plan

64
00:03:01,010 --> 00:03:02,900
that we've developed
and want to make sure

65
00:03:02,900 --> 00:03:05,910
that the service that's
delivered is reliable.

66
00:03:05,910 --> 00:03:10,010
That requires decent equipment,
decent infrastructure, and also

67
00:03:10,010 --> 00:03:12,230
a workforce that's
motivated and there

68
00:03:12,230 --> 00:03:17,060
when you need them, which
is pretty much every day.

69
00:03:17,060 --> 00:03:19,640
So the operations
plan you've all

70
00:03:19,640 --> 00:03:22,160
been through this at
this stage in the course.

71
00:03:22,160 --> 00:03:24,160
So you know what goes
into the operations plans.

72
00:03:24,160 --> 00:03:25,920
So what comes out of
the operations plan

73
00:03:25,920 --> 00:03:29,510
are a set of schedules,
timetables, runs, or duties

74
00:03:29,510 --> 00:03:32,610
for operators to perform.

75
00:03:32,610 --> 00:03:37,200
And that require a certain
amount of workforce or--

76
00:03:37,200 --> 00:03:40,230
excuse me for using the term
"manpower," but it's sort

77
00:03:40,230 --> 00:03:44,140
of the workforce that we need.

78
00:03:44,140 --> 00:03:46,590
So if you don't have the
right level of workforce

79
00:03:46,590 --> 00:03:49,560
available on a particular
day or at a particular time

80
00:03:49,560 --> 00:03:53,820
on that day, then you've
got a problem potentially.

81
00:03:53,820 --> 00:03:58,960
If you can buy work
on an overtime basis,

82
00:03:58,960 --> 00:04:01,870
you can do that and at
least not drop any trips.

83
00:04:01,870 --> 00:04:04,187
That's going to cost
you money though.

84
00:04:04,187 --> 00:04:06,270
And you may not be able
to get the overtime, which

85
00:04:06,270 --> 00:04:10,800
will, again, produce unreliable
service through dropped trips.

86
00:04:10,800 --> 00:04:15,690
It may also, if some
workers in the workforce

87
00:04:15,690 --> 00:04:19,200
feel that they can rely on
overtime, which gives them

88
00:04:19,200 --> 00:04:23,520
a 50% premium rate, then maybe
you can take more absence time.

89
00:04:23,520 --> 00:04:25,530
So there's a substitution
relationship here

90
00:04:25,530 --> 00:04:27,690
between absence and overtime.

91
00:04:27,690 --> 00:04:30,390
So if you rely more
heavily on overtime,

92
00:04:30,390 --> 00:04:32,010
perhaps your absence
rate will go up.

93
00:04:32,010 --> 00:04:33,910
That's not an outcome you want.

94
00:04:33,910 --> 00:04:36,480
So you've got to be
very careful about that.

95
00:04:36,480 --> 00:04:38,640
Certainly, if you
can't buy the overtime

96
00:04:38,640 --> 00:04:42,120
and you don't have enough
people on the property

97
00:04:42,120 --> 00:04:44,440
that particular day,
that particular time,

98
00:04:44,440 --> 00:04:47,010
then you can drop trips
because you don't have anyone

99
00:04:47,010 --> 00:04:48,870
to run the bus or
to run the train.

100
00:04:48,870 --> 00:04:50,460
And that's a
problem because that

101
00:04:50,460 --> 00:04:54,090
produces worse service
reliability, worse customer

102
00:04:54,090 --> 00:04:54,810
satisfaction.

103
00:04:54,810 --> 00:04:59,170
Politically, it's a very
negative [AUDIO OUT] as well.

104
00:04:59,170 --> 00:05:00,870
And so workforce
planning is actually

105
00:05:00,870 --> 00:05:04,260
a fairly critical piece of
getting from the operations

106
00:05:04,260 --> 00:05:08,085
plan to good service on
the street or on the rails.

107
00:05:11,470 --> 00:05:16,390
This is sort of a high
level view of what

108
00:05:16,390 --> 00:05:19,300
the problem is, as I see it.

109
00:05:19,300 --> 00:05:22,600
So on the left of
this chart, you

110
00:05:22,600 --> 00:05:24,850
have the operator availability.

111
00:05:24,850 --> 00:05:28,720
So that means the number
of drivers you have

112
00:05:28,720 --> 00:05:30,880
available on a particular day.

113
00:05:30,880 --> 00:05:34,600
That's affected by the
number of operators

114
00:05:34,600 --> 00:05:37,750
you had available
the previous day.

115
00:05:37,750 --> 00:05:40,360
If you hired people,
how many people

116
00:05:40,360 --> 00:05:44,420
are coming on to
increase the supply?

117
00:05:44,420 --> 00:05:47,200
What sort of attrition--
what sort of outflow

118
00:05:47,200 --> 00:05:49,870
do you have in your workforce
from your previous time, day,

119
00:05:49,870 --> 00:05:51,620
or time period?

120
00:05:51,620 --> 00:05:52,420
What's the absence?

121
00:05:52,420 --> 00:05:53,830
Because those guys
aren't going to be there.

122
00:05:53,830 --> 00:05:56,410
Those people aren't going to
be there to deliver service.

123
00:05:56,410 --> 00:05:58,470
And of course,
you have a roster,

124
00:05:58,470 --> 00:06:02,650
so everyone typically works five
days and takes two days off.

125
00:06:02,650 --> 00:06:04,910
And that rotates
among the workforce.

126
00:06:04,910 --> 00:06:10,130
So this is the supply side,
in terms of the operator.

127
00:06:10,130 --> 00:06:13,280
On the demand side, you
have the scheduled service,

128
00:06:13,280 --> 00:06:16,020
which is a result of
your operating plan--

129
00:06:16,020 --> 00:06:18,530
what you want to provide.

130
00:06:18,530 --> 00:06:22,820
You have unscheduled
service additions.

131
00:06:22,820 --> 00:06:25,820
And you potentially have
unscheduled service deletions.

132
00:06:25,820 --> 00:06:29,570
For instance, if there's
a school vacation,

133
00:06:29,570 --> 00:06:31,910
you will drop some
school trips--

134
00:06:31,910 --> 00:06:34,610
trips which are primarily
there on high frequency

135
00:06:34,610 --> 00:06:36,830
routes to provide
school service.

136
00:06:36,830 --> 00:06:39,910
And these don't fit naturally
into the timetable process

137
00:06:39,910 --> 00:06:42,410
because there's lot of overhead
in developing and timetable,

138
00:06:42,410 --> 00:06:45,600
as I'm sure Gabriel
has explained to you.

139
00:06:45,600 --> 00:06:49,190
So you'll just take those
trips off the table that day.

140
00:06:49,190 --> 00:06:52,190
And those operators are
then available to provide

141
00:06:52,190 --> 00:06:55,460
other service that you need and
to fill in for absent operators

142
00:06:55,460 --> 00:06:57,980
those days.

143
00:06:57,980 --> 00:06:59,810
So you've got the supply.

144
00:06:59,810 --> 00:07:01,100
You've got the demand.

145
00:07:01,100 --> 00:07:03,950
And so you're down here
looking at the availability

146
00:07:03,950 --> 00:07:06,445
versus the supply of drivers.

147
00:07:06,445 --> 00:07:08,080
I'll call them "drivers."

148
00:07:08,080 --> 00:07:13,040
The technical jargon in the
industry is "operators."

149
00:07:13,040 --> 00:07:16,160
In a perfect world, you'd
have an exact balance

150
00:07:16,160 --> 00:07:18,410
between availability
and demand, and you

151
00:07:18,410 --> 00:07:21,020
have no additional cost,
and you've got someone

152
00:07:21,020 --> 00:07:22,190
to cover every trip.

153
00:07:22,190 --> 00:07:24,090
That will never happen.

154
00:07:24,090 --> 00:07:24,890
Why?

155
00:07:24,890 --> 00:07:28,820
Because the operator
availability is stochastic.

156
00:07:28,820 --> 00:07:30,050
It's uncertain.

157
00:07:30,050 --> 00:07:32,840
You don't know what the absence
rate is going to be tomorrow.

158
00:07:32,840 --> 00:07:34,520
With the best will in the
world, with the best models

159
00:07:34,520 --> 00:07:36,187
in the world, you
don't know whether I'm

160
00:07:36,187 --> 00:07:38,660
going to be feeling well
enough to work tomorrow.

161
00:07:38,660 --> 00:07:40,310
So that's a problem.

162
00:07:40,310 --> 00:07:45,063
And for instance,
in the MBTA, you

163
00:07:45,063 --> 00:07:47,480
don't know if you going to
have a failure on the Red Line.

164
00:07:47,480 --> 00:07:49,820
You're going to have to
bus parts of the Red Line

165
00:07:49,820 --> 00:07:53,760
tomorrow at 8 o'clock
in the morning, OK.

166
00:07:53,760 --> 00:07:58,470
So all of those uncertainties
affect this balance

167
00:07:58,470 --> 00:08:00,960
between the
availability and demand.

168
00:08:00,960 --> 00:08:02,430
So you'll never be here.

169
00:08:02,430 --> 00:08:05,100
I mean, statistically,
it's virtually

170
00:08:05,100 --> 00:08:07,050
certain you'll never be there.

171
00:08:07,050 --> 00:08:10,020
You're either going to be over
here, where the availability is

172
00:08:10,020 --> 00:08:11,370
less than the demand.

173
00:08:11,370 --> 00:08:13,500
In that case, you have
additional time costs,

174
00:08:13,500 --> 00:08:16,260
because you're going to have
to try and buy operators

175
00:08:16,260 --> 00:08:17,400
to work that overtime.

176
00:08:20,580 --> 00:08:23,580
And perhaps more seriously, you
have the risk of dropping trips

177
00:08:23,580 --> 00:08:26,220
if you can't persuade
operators that they should

178
00:08:26,220 --> 00:08:27,480
work the overtime that day.

179
00:08:30,450 --> 00:08:32,625
Or you're over here,
where the availability

180
00:08:32,625 --> 00:08:33,750
is greater than the demand.

181
00:08:33,750 --> 00:08:36,150
In that case, you have
additional personnel costs

182
00:08:36,150 --> 00:08:39,900
who aren't going to be used as
productively as you might like.

183
00:08:39,900 --> 00:08:41,970
If they were going to
be used productively,

184
00:08:41,970 --> 00:08:44,520
you'd have built them
into the operating plan.

185
00:08:44,520 --> 00:08:48,600
So they're probably not going
to be used as productively

186
00:08:48,600 --> 00:08:50,190
as your average operator.

187
00:08:50,190 --> 00:08:52,880
So this is sort of one way
of looking at the framework.

188
00:08:56,260 --> 00:09:02,020
We can think of, again, three
levels of planning here.

189
00:09:02,020 --> 00:09:06,040
At the strategic level,
this is typically

190
00:09:06,040 --> 00:09:09,250
looking at a one-year or even
longer than one year horizon.

191
00:09:09,250 --> 00:09:11,920
So it's not strategic in
terms of infrastructure.

192
00:09:11,920 --> 00:09:15,120
It's strategic in
terms of operations.

193
00:09:15,120 --> 00:09:17,720
So that's how many
people will we

194
00:09:17,720 --> 00:09:23,270
need in the driver workforce
over the next 12 months.

195
00:09:23,270 --> 00:09:29,050
And how do we get
a hiring plan that

196
00:09:29,050 --> 00:09:33,700
will give us the right number
of people throughout the year?

197
00:09:33,700 --> 00:09:36,150
How do we allocate
our vacations?

198
00:09:36,150 --> 00:09:39,460
Do we have a certain
vacation liability?

199
00:09:39,460 --> 00:09:44,140
If I'm a 20-year employee of
the MBTA in the commons union,

200
00:09:44,140 --> 00:09:48,380
I probably have about five weeks
vacation time over the year.

201
00:09:48,380 --> 00:09:51,730
So I get my five weeks.

202
00:09:51,730 --> 00:09:53,590
But the MBTA
management decisions

203
00:09:53,590 --> 00:09:57,020
include when to provide
the vacation allocation.

204
00:09:57,020 --> 00:09:59,800
So you may have an overall
life vacation liability

205
00:09:59,800 --> 00:10:04,560
of 2,000 vacation weeks.

206
00:10:04,560 --> 00:10:06,750
So the question of
the strategic level,

207
00:10:06,750 --> 00:10:11,820
how do you allocate those 2,000
weeks of vacation liability

208
00:10:11,820 --> 00:10:13,687
over the course of the year?

209
00:10:13,687 --> 00:10:15,270
And there are good
ways of doing that.

210
00:10:15,270 --> 00:10:17,950
And there are bad
ways of doing that.

211
00:10:17,950 --> 00:10:20,340
I mean, if you give
vacation liability

212
00:10:20,340 --> 00:10:22,440
when people don't want
to take vacations,

213
00:10:22,440 --> 00:10:24,740
what's likely to happen?

214
00:10:24,740 --> 00:10:26,682
Your absence rate's
going to increase.

215
00:10:26,682 --> 00:10:28,390
So you really want to
give vacations when

216
00:10:28,390 --> 00:10:30,280
people want to take vacations.

217
00:10:30,280 --> 00:10:32,743
But you don't want to
give them so much vacation

218
00:10:32,743 --> 00:10:34,660
when they want to take
vacation that it really

219
00:10:34,660 --> 00:10:36,250
costs you a lot more money.

220
00:10:36,250 --> 00:10:38,650
In that case, maybe
taking a higher absence

221
00:10:38,650 --> 00:10:40,610
will be a better way
of dealing with that.

222
00:10:40,610 --> 00:10:42,250
So there are really
important questions

223
00:10:42,250 --> 00:10:46,900
on how to meet the
vacation liability.

224
00:10:46,900 --> 00:10:48,460
And this is the timetable.

225
00:10:48,460 --> 00:10:50,760
We've decided what
we need in terms

226
00:10:50,760 --> 00:10:53,290
of finding decent service.

227
00:10:53,290 --> 00:10:54,670
And then, at the
strategic level,

228
00:10:54,670 --> 00:10:58,480
I'd also include
personnel policies.

229
00:10:58,480 --> 00:11:01,850
So as I think you all
know, in many agencies,

230
00:11:01,850 --> 00:11:04,090
including the MBTA,
you have a mix

231
00:11:04,090 --> 00:11:06,140
of part-timers and full-timers.

232
00:11:06,140 --> 00:11:08,323
In the MBTA, you have
Management Rights Act, which

233
00:11:08,323 --> 00:11:09,490
I'm sure you've heard about.

234
00:11:09,490 --> 00:11:12,330
Maybe Fred'll talk
about that next week.

235
00:11:12,330 --> 00:11:14,920
It's a very interesting topic.

236
00:11:14,920 --> 00:11:20,500
And Fred will make a meal
of it, I guarantee that.

237
00:11:20,500 --> 00:11:26,680
So the MBTA management was given
these rights by the legislature

238
00:11:26,680 --> 00:11:29,350
35 years ago, 36 years ago.

239
00:11:29,350 --> 00:11:35,610
And so they cannot negotiate
about not using part-timers.

240
00:11:35,610 --> 00:11:37,770
That's not part of the
contract negotiation.

241
00:11:37,770 --> 00:11:40,810
That's a right they've
been given legally.

242
00:11:40,810 --> 00:11:42,570
It's been challenged
in the courts.

243
00:11:42,570 --> 00:11:43,680
It's been upheld.

244
00:11:43,680 --> 00:11:45,870
And so they will
use part-timers.

245
00:11:45,870 --> 00:11:48,390
They've decided to use
part-timers in a way

246
00:11:48,390 --> 00:11:50,970
that I think is
very unproductive.

247
00:11:50,970 --> 00:11:56,710
They use part-timers as
apprentice full-timers.

248
00:11:56,710 --> 00:11:59,710
And so someone suffers
through perhaps five years

249
00:11:59,710 --> 00:12:02,440
as a part-timer at
the MBTA in order

250
00:12:02,440 --> 00:12:06,740
to reach nirvana of being
a full-timer at the MBTA.

251
00:12:06,740 --> 00:12:10,310
Well, that's one
way of doing it.

252
00:12:10,310 --> 00:12:13,630
But I'm not sure that's
the best way of doing it.

253
00:12:13,630 --> 00:12:15,970
Perhaps what you're
doing is, you're

254
00:12:15,970 --> 00:12:19,720
making these part-timers
suffer for five years,

255
00:12:19,720 --> 00:12:22,210
if that's how long it takes
to get promoted from part-time

256
00:12:22,210 --> 00:12:23,085
to full-timer status.

257
00:12:23,085 --> 00:12:25,090
So they're sort of unhappy
for that five years,

258
00:12:25,090 --> 00:12:28,590
because that's not
what they want.

259
00:12:28,590 --> 00:12:31,210
And it's pretty miserable
working conditions.

260
00:12:31,210 --> 00:12:32,850
They don't get full benefits.

261
00:12:32,850 --> 00:12:39,480
And typically, they work a
6-hour day spread over 13

262
00:12:39,480 --> 00:12:40,260
hours--

263
00:12:40,260 --> 00:12:42,480
so covering the AM
peak and the PM peak.

264
00:12:42,480 --> 00:12:45,610
And they're doing nothing
between the AM peak and PM

265
00:12:45,610 --> 00:12:46,110
peak.

266
00:12:46,110 --> 00:12:47,670
You hope they're not in bars.

267
00:12:47,670 --> 00:12:49,337
But they're certainly
not doing anything

268
00:12:49,337 --> 00:12:52,230
productive with the property.

269
00:12:52,230 --> 00:12:54,620
So that's the way
the MBTA decides

270
00:12:54,620 --> 00:12:56,520
to use the management
rights they

271
00:12:56,520 --> 00:12:58,320
were given by the legislature.

272
00:12:58,320 --> 00:13:00,160
I don't think that
was a good decision.

273
00:13:00,160 --> 00:13:03,810
If you have separate
AM part-timers,

274
00:13:03,810 --> 00:13:06,840
might you be able to
increase your workforce?

275
00:13:06,840 --> 00:13:08,100
I think so.

276
00:13:08,100 --> 00:13:11,640
You have people-- shop clerks
who don't really start work

277
00:13:11,640 --> 00:13:14,190
until after the
AM peak who might

278
00:13:14,190 --> 00:13:19,290
be very interested in
supplementing their shop clerk

279
00:13:19,290 --> 00:13:23,220
job by having an
AM-peak driving job.

280
00:13:23,220 --> 00:13:26,280
And similarly, in the
PM peak, you probably

281
00:13:26,280 --> 00:13:28,920
can increase your
labor pool with people

282
00:13:28,920 --> 00:13:32,940
who are not interested in a
full-time MBTA driver's job,

283
00:13:32,940 --> 00:13:36,930
but they would be interested
in just a sort of a guaranteed,

284
00:13:36,930 --> 00:13:39,670
3-hour piece of work every day.

285
00:13:39,670 --> 00:13:42,150
So there are lots of different
ways of implementing this.

286
00:13:42,150 --> 00:13:46,280
And I'm pretty sure the
MBTA didn't do it optimally.

287
00:13:46,280 --> 00:13:49,860
So that's what I mean
by personnel policies.

288
00:13:49,860 --> 00:13:52,760
So all of these are at
the strategic level.

289
00:13:52,760 --> 00:13:57,020
At the tactical level, what
we're making decisions on

290
00:13:57,020 --> 00:14:02,270
is the time horizon is
typically one timetable.

291
00:14:02,270 --> 00:14:05,653
How many timetables does
the MBTA have per year?

292
00:14:05,653 --> 00:14:06,278
AUDIENCE: Four.

293
00:14:06,278 --> 00:14:07,960
NIGEL HM WILSON: Yeah, four.

294
00:14:07,960 --> 00:14:09,940
So you're making
these decisions,

295
00:14:09,940 --> 00:14:12,430
running up to each timetable--

296
00:14:12,430 --> 00:14:14,800
the start of each
timetable period.

297
00:14:14,800 --> 00:14:18,730
And the main decisions
are the staff.

298
00:14:18,730 --> 00:14:22,240
And you've sort of got
your workforce fixed.

299
00:14:22,240 --> 00:14:23,980
And you're deciding
how to allocate them

300
00:14:23,980 --> 00:14:30,110
across depots or garages or rail
lines and by days of the week.

301
00:14:30,110 --> 00:14:32,920
Those are the tactical-level
decisions you're making.

302
00:14:32,920 --> 00:14:35,690
Those, it turns out, are
actually quite important.

303
00:14:35,690 --> 00:14:38,235
If you make poor
decisions, then you're

304
00:14:38,235 --> 00:14:40,360
going to be providing less
reliable service or more

305
00:14:40,360 --> 00:14:44,020
expensive service than
you otherwise would.

306
00:14:44,020 --> 00:14:47,140
And then at the final level,
your operational level,

307
00:14:47,140 --> 00:14:52,450
is here we are 4
o'clock on a Thursday.

308
00:14:52,450 --> 00:14:54,220
At 12 o'clock today--

309
00:14:54,220 --> 00:14:57,280
so four hours ago--
the MBTA was deciding

310
00:14:57,280 --> 00:15:01,450
when to ask each extra
board operator to show up

311
00:15:01,450 --> 00:15:05,020
for work tomorrow if they didn't
have a guaranteed assignment.

312
00:15:05,020 --> 00:15:06,310
OK?

313
00:15:06,310 --> 00:15:10,430
So there are unexpected absences
that will occur tomorrow.

314
00:15:10,430 --> 00:15:12,340
You don't know when
they're going to occur.

315
00:15:12,340 --> 00:15:13,715
You don't know
whether it's going

316
00:15:13,715 --> 00:15:16,840
to be an AM report or a PM
report or whatever it is.

317
00:15:16,840 --> 00:15:19,510
And you have to make
the decision about when,

318
00:15:19,510 --> 00:15:21,700
if you have five extra
board operators that

319
00:15:21,700 --> 00:15:23,740
don't have an assignment,
because a known,

320
00:15:23,740 --> 00:15:27,640
in advance absence, when
you want them to show up?

321
00:15:27,640 --> 00:15:31,000
And that the daily report times
for unassigned, extra staff

322
00:15:31,000 --> 00:15:35,060
is the operational-level
decision.

323
00:15:35,060 --> 00:15:37,160
Any questions about
this, in terms

324
00:15:37,160 --> 00:15:41,240
of strategic,
operational, tactical?

325
00:15:41,240 --> 00:15:46,740
And I'm not talking here about
policies that may change--

326
00:15:46,740 --> 00:15:48,630
absenteeism, for instance.

327
00:15:48,630 --> 00:15:50,910
Fred will talk about
some of those next week

328
00:15:50,910 --> 00:15:52,620
when he talks about
labor relations.

329
00:15:52,620 --> 00:15:57,510
But I'm talking about sort of a
narrower piece of this puzzle.

330
00:15:57,510 --> 00:15:58,900
Any questions on this?

331
00:15:58,900 --> 00:15:59,640
Yes, Kenji?

332
00:15:59,640 --> 00:16:05,083
AUDIENCE: [? Someone ?] become
the part-time operator--

333
00:16:05,083 --> 00:16:06,000
NIGEL HM WILSON: Yeah.

334
00:16:06,000 --> 00:16:09,428
AUDIENCE: So are they
required any license or--

335
00:16:09,428 --> 00:16:10,470
NIGEL HM WILSON: Oh, yes.

336
00:16:10,470 --> 00:16:13,920
A commercial driver's
license is required for them.

337
00:16:13,920 --> 00:16:18,360
But again, what I
was saying is, you

338
00:16:18,360 --> 00:16:25,410
have choice about both who to
recruit into those positions--

339
00:16:25,410 --> 00:16:29,970
do we want an AM part-timer
or a split-shift, full-day

340
00:16:29,970 --> 00:16:31,830
part-timer who's
going to be unhappy

341
00:16:31,830 --> 00:16:35,610
until they get their
full-time position at the MBTA

342
00:16:35,610 --> 00:16:36,940
that they really want.

343
00:16:36,940 --> 00:16:37,920
So yeah, sorry.

344
00:16:37,920 --> 00:16:41,110
The other point I wanted
to make is that how

345
00:16:41,110 --> 00:16:45,040
you select those people who
will get promoted to full-timer.

346
00:16:45,040 --> 00:16:47,610
So you've decided,
through attrition

347
00:16:47,610 --> 00:16:50,400
in the full-time ranks, you want
to promote some part-timers.

348
00:16:50,400 --> 00:16:52,570
How do you make that promotion?

349
00:16:52,570 --> 00:16:55,710
Do you promote based
strictly on seniority?

350
00:16:55,710 --> 00:16:58,740
In other words, the first
one to become a part-timer

351
00:16:58,740 --> 00:17:02,160
becomes the first one to be
promoted to a full-timer?

352
00:17:02,160 --> 00:17:04,200
Probably not,
because you'd really

353
00:17:04,200 --> 00:17:08,130
like people who are motivated,
who have a good absence record,

354
00:17:08,130 --> 00:17:09,480
a good accident record.

355
00:17:09,480 --> 00:17:13,140
So you'd like to use criteria in
that promotion, which are more

356
00:17:13,140 --> 00:17:15,609
than just strict seniority.

357
00:17:15,609 --> 00:17:18,400
And the MBTA, again,
has been, I would say,

358
00:17:18,400 --> 00:17:22,670
not very effective in using
that balance between seniority,

359
00:17:22,670 --> 00:17:24,760
which is an important issue--

360
00:17:24,760 --> 00:17:29,080
you want to avoid favoritism and
discrimination in the promotion

361
00:17:29,080 --> 00:17:29,890
process--

362
00:17:29,890 --> 00:17:33,370
with having some
elements of performance

363
00:17:33,370 --> 00:17:37,435
in the promotion
process, I would argue.

364
00:17:37,435 --> 00:17:38,560
Did I answer your question?

365
00:17:38,560 --> 00:17:39,647
AUDIENCE: Yeah.

366
00:17:39,647 --> 00:17:40,730
NIGEL HM WILSON: Go ahead.

367
00:17:40,730 --> 00:17:42,355
AUDIENCE: I was
interested in the topic

368
00:17:42,355 --> 00:17:47,340
because, yeah, we have a similar
culture for the station's

369
00:17:47,340 --> 00:17:47,960
time--

370
00:17:47,960 --> 00:17:48,470
NIGEL HM WILSON: Yes.

371
00:17:48,470 --> 00:17:49,345
AUDIENCE: --in Japan.

372
00:17:49,345 --> 00:17:55,922
But for the driver, they
require a very strict license--

373
00:17:55,922 --> 00:17:56,880
NIGEL HM WILSON: Right.

374
00:17:56,880 --> 00:17:58,713
AUDIENCE: --which takes
a long time to test.

375
00:17:58,713 --> 00:18:00,730
So we don't have any
part-time drivers.

376
00:18:00,730 --> 00:18:01,800
NIGEL HM WILSON: Right.

377
00:18:01,800 --> 00:18:02,580
Right, yes.

378
00:18:02,580 --> 00:18:06,360
And there are commercial
driver's licenses

379
00:18:06,360 --> 00:18:07,380
required for the MBTA.

380
00:18:07,380 --> 00:18:10,890
A bus driver or train driver has
to have a commercial driver's

381
00:18:10,890 --> 00:18:11,910
license.

382
00:18:11,910 --> 00:18:16,170
But it's not as a rigorous and
long period of qualification

383
00:18:16,170 --> 00:18:17,860
as it is, I'm sure in JR-East.

384
00:18:17,860 --> 00:18:20,430
That would be my guess, anyway.

385
00:18:20,430 --> 00:18:22,080
Any other comments or questions?

386
00:18:22,080 --> 00:18:24,460
OK.

387
00:18:24,460 --> 00:18:30,430
OK, so let's look at this
as three levels of planning.

388
00:18:30,430 --> 00:18:33,880
So strategic level-- we're
deciding the workforce size,

389
00:18:33,880 --> 00:18:36,280
vacation allocation
over the year,

390
00:18:36,280 --> 00:18:40,940
and hiring patterns over
the course of a year.

391
00:18:40,940 --> 00:18:43,840
So by period of the
year, the input into this

392
00:18:43,840 --> 00:18:46,570
includes absence hours.

393
00:18:46,570 --> 00:18:49,570
If we know we typically
have more absence hours

394
00:18:49,570 --> 00:18:52,660
in the winter because the
conditions are miserable--

395
00:18:52,660 --> 00:18:54,850
you don't drive
in snow and ice--

396
00:18:54,850 --> 00:18:58,330
and if you have observed
higher absence rates,

397
00:18:58,330 --> 00:19:00,820
that's important in
making the right decisions

398
00:19:00,820 --> 00:19:04,220
on the strategic level.

399
00:19:04,220 --> 00:19:05,390
Required extra work.

400
00:19:05,390 --> 00:19:11,210
If we know that we are going
to be bussing on the Braintree

401
00:19:11,210 --> 00:19:13,667
branch of the Red Line
because of reconstruction,

402
00:19:13,667 --> 00:19:16,000
then we'll build that into
our workforce planning thing.

403
00:19:16,000 --> 00:19:18,230
So that's an example
of required extra work.

404
00:19:18,230 --> 00:19:20,360
And attrition--
we do projections

405
00:19:20,360 --> 00:19:23,510
about the number of
full-timers and part-timers

406
00:19:23,510 --> 00:19:29,150
who will be retiring or leaving
the productive workforce

407
00:19:29,150 --> 00:19:32,360
over the next year.

408
00:19:32,360 --> 00:19:35,510
On the right-hand side, we
have input such as budget,

409
00:19:35,510 --> 00:19:39,260
the service plan, the operations
plan, vacation liability--

410
00:19:39,260 --> 00:19:42,740
how much vacation do we have
to provide in the next year--

411
00:19:42,740 --> 00:19:46,730
work rules and policies
that we want to meet.

412
00:19:46,730 --> 00:19:49,820
And what I'm showing
here is strategic level,

413
00:19:49,820 --> 00:19:54,200
tactical level, and
operational level.

414
00:19:54,200 --> 00:19:57,980
And what you'll see is
these downward arrows

415
00:19:57,980 --> 00:20:01,340
from strategy to
tactics to operations,

416
00:20:01,340 --> 00:20:04,340
and the upward
pointing arrows back

417
00:20:04,340 --> 00:20:08,570
up the chain from operations
to tactical to strategic.

418
00:20:08,570 --> 00:20:12,230
So let's explain
each of these arrows.

419
00:20:12,230 --> 00:20:17,850
The downward arrows represent
constraints on your decisions.

420
00:20:17,850 --> 00:20:21,050
So if you've made some
strategic-level decisions

421
00:20:21,050 --> 00:20:23,390
on workforce size,
that limits what you

422
00:20:23,390 --> 00:20:25,220
can do at the tactical level.

423
00:20:25,220 --> 00:20:27,860
And that's this
downward arrow here.

424
00:20:27,860 --> 00:20:31,250
Similarly, you make your
tactical-level decisions

425
00:20:31,250 --> 00:20:33,030
at the beginning of
a timetable, that

426
00:20:33,030 --> 00:20:35,030
limits what you can do
at the operational level.

427
00:20:35,030 --> 00:20:38,180
These are these
downward-pointing arrows

428
00:20:38,180 --> 00:20:40,040
between these boxes.

429
00:20:40,040 --> 00:20:44,480
The upward pointing arrows
are the relationships

430
00:20:44,480 --> 00:20:47,540
between that you need
to know about how

431
00:20:47,540 --> 00:20:51,890
efficiently the operations
level is conducted

432
00:20:51,890 --> 00:20:56,030
to make good decisions at the
higher levels of the planning

433
00:20:56,030 --> 00:20:58,800
hierarchy.

434
00:20:58,800 --> 00:21:05,390
So for instance, if you know
that overtime is virtually

435
00:21:05,390 --> 00:21:07,130
impossible to get
from your workforce,

436
00:21:07,130 --> 00:21:09,000
for whatever reason--
they may say,

437
00:21:09,000 --> 00:21:12,170
I'm just not going to cooperate,
I'm not going to work,

438
00:21:12,170 --> 00:21:16,760
or that could be bad or
good reasons for that--

439
00:21:16,760 --> 00:21:18,710
then you need to
know that in order

440
00:21:18,710 --> 00:21:22,430
to make the right decisions
at the tactical level.

441
00:21:22,430 --> 00:21:24,500
And similarly, you need
to know that to make

442
00:21:24,500 --> 00:21:27,365
the right decisions at
the strategic level.

443
00:21:27,365 --> 00:21:30,990
So these are these
upward-pointing arrows here.

444
00:21:30,990 --> 00:21:35,220
And the difference in the
inputs are just more detail.

445
00:21:35,220 --> 00:21:39,500
So absence hours by
garage and by day of week

446
00:21:39,500 --> 00:21:44,850
is important input into
the tactical level.

447
00:21:44,850 --> 00:21:48,290
Absence hours by
garage and time of day

448
00:21:48,290 --> 00:21:54,420
is important input into
the operational level, OK.

449
00:21:54,420 --> 00:21:55,930
And you have work
rules and policies

450
00:21:55,930 --> 00:21:58,360
as critical inputs
on all these levels.

451
00:22:01,500 --> 00:22:07,650
So this is a slightly different
look at these relationships.

452
00:22:10,170 --> 00:22:15,110
And I want to lead you
through this a little bit.

453
00:22:15,110 --> 00:22:17,810
So this is our
strategic planning box

454
00:22:17,810 --> 00:22:20,300
where we're deciding
workforce size.

455
00:22:20,300 --> 00:22:21,140
What goes into this?

456
00:22:21,140 --> 00:22:23,030
The unit costs--
how much does it

457
00:22:23,030 --> 00:22:27,050
cost to provide a
platform hour, which

458
00:22:27,050 --> 00:22:30,650
is an hour of driving
a bus, by an overtime

459
00:22:30,650 --> 00:22:34,670
employee, a full-time employee,
and a part-time employee?

460
00:22:34,670 --> 00:22:38,300
Those are very important,
basic cost units

461
00:22:38,300 --> 00:22:41,520
that we need to have.

462
00:22:41,520 --> 00:22:43,630
What's the budget, and
what are the work rules?

463
00:22:43,630 --> 00:22:46,260
The work rules are important
because they tell us

464
00:22:46,260 --> 00:22:51,713
how easy it will be
to do certain things.

465
00:22:51,713 --> 00:22:53,630
And I'll give you an
example of that later on.

466
00:22:53,630 --> 00:22:55,693
And of course, you have
the scheduled hours,

467
00:22:55,693 --> 00:22:57,110
which comes out
of your operations

468
00:22:57,110 --> 00:23:01,300
plan-- the timetable
requirements.

469
00:23:01,300 --> 00:23:04,980
So given that you've made a
decision on the workforce size

470
00:23:04,980 --> 00:23:10,780
at the strategic level,
on a particular day,

471
00:23:10,780 --> 00:23:14,110
you may have an
amount of open work.

472
00:23:14,110 --> 00:23:15,860
What do I mean by open work?

473
00:23:15,860 --> 00:23:17,770
It's work that we
want to deliver,

474
00:23:17,770 --> 00:23:20,860
but there's no one
obvious to provide it.

475
00:23:20,860 --> 00:23:21,910
There's no one.

476
00:23:21,910 --> 00:23:29,260
So we'll try and buy time
on an overtime basis.

477
00:23:29,260 --> 00:23:32,000
So we're down here.

478
00:23:32,000 --> 00:23:35,842
And if we're lucky,
the overtime--

479
00:23:35,842 --> 00:23:37,300
there'll be someone
available who's

480
00:23:37,300 --> 00:23:41,500
willing to work the
overtime for increased pay.

481
00:23:41,500 --> 00:23:44,920
The negative side of that
is it increases the strain

482
00:23:44,920 --> 00:23:49,360
on that employee who's working
more hours than he or she was

483
00:23:49,360 --> 00:23:50,620
expecting to.

484
00:23:50,620 --> 00:23:54,900
That may produce this
negative feedback on absence.

485
00:23:54,900 --> 00:23:57,970
And I described it
previously as the trade-off

486
00:23:57,970 --> 00:24:01,690
between how much do I have to
work to get to a certain target

487
00:24:01,690 --> 00:24:02,410
income level?

488
00:24:02,410 --> 00:24:03,740
Well, that's one way
of looking at it.

489
00:24:03,740 --> 00:24:06,365
But the other way of looking at
it, which is equally realistic,

490
00:24:06,365 --> 00:24:10,060
is if you're working harder,
that puts more stress

491
00:24:10,060 --> 00:24:11,020
and strain on the body.

492
00:24:11,020 --> 00:24:13,330
You're more likely to be absent.

493
00:24:13,330 --> 00:24:16,120
Because there are two ways of
looking at it, which both raise

494
00:24:16,120 --> 00:24:20,560
the concern about relying
heavily on overtime and what

495
00:24:20,560 --> 00:24:23,000
that will do to absence.

496
00:24:23,000 --> 00:24:26,323
Again, over here, if you
can't get the overtime,

497
00:24:26,323 --> 00:24:27,740
you're going to
be dropping trips.

498
00:24:27,740 --> 00:24:29,960
So reliability suffers.

499
00:24:29,960 --> 00:24:33,447
Negative effect on service
reliability, service quality.

500
00:24:33,447 --> 00:24:35,030
Presumably, that
will affect ridership

501
00:24:35,030 --> 00:24:36,650
if it goes on long enough.

502
00:24:36,650 --> 00:24:38,450
And that will affect
your operating plan,

503
00:24:38,450 --> 00:24:40,460
because your ridership's down.

504
00:24:40,460 --> 00:24:43,350
And so there's that
feedback, as well.

505
00:24:43,350 --> 00:24:47,520
So this is sort of another way
of looking at the framework.

506
00:24:47,520 --> 00:24:54,920
Now let's look at the strategic
level in terms of the planning

507
00:24:54,920 --> 00:24:57,200
process.

508
00:24:57,200 --> 00:25:05,210
What I've shown here is
a month, as was said--

509
00:25:05,210 --> 00:25:07,940
four seasons in the MBTA.

510
00:25:07,940 --> 00:25:09,020
Some agencies have more.

511
00:25:09,020 --> 00:25:10,730
Some have less.

512
00:25:10,730 --> 00:25:15,060
Winter, spring,
summer, and fall.

513
00:25:15,060 --> 00:25:20,280
And let's have a hypothetical
workforce hiring scenario,

514
00:25:20,280 --> 00:25:23,430
where we hire up at the
beginning of each timetable.

515
00:25:23,430 --> 00:25:25,290
We sort of run
through the timetable.

516
00:25:25,290 --> 00:25:27,300
And then we hire up
for the beginning

517
00:25:27,300 --> 00:25:29,080
of the next timetable.

518
00:25:29,080 --> 00:25:30,790
That's sort of the hypothetical.

519
00:25:30,790 --> 00:25:33,910
That's not unusual.

520
00:25:33,910 --> 00:25:35,900
But it's not a very
effective way of doing it.

521
00:25:35,900 --> 00:25:39,260
And what I've tried to explain
here is why that is the case.

522
00:25:39,260 --> 00:25:48,430
So the bottom bar in this
graph indicates the timetable

523
00:25:48,430 --> 00:25:50,020
requirements.

524
00:25:50,020 --> 00:25:53,290
So typically, the timetable
requirements for the operating

525
00:25:53,290 --> 00:26:00,540
plan are pretty similar across
winter, spring, and fall,

526
00:26:00,540 --> 00:26:02,640
and they're lower in the summer.

527
00:26:02,640 --> 00:26:04,640
Why is that?

528
00:26:04,640 --> 00:26:06,900
No schools, and
people on vacation,

529
00:26:06,900 --> 00:26:08,340
so your demand is
generally down,

530
00:26:08,340 --> 00:26:13,800
unless it's a very strong summer
tourist place, like Boston.

531
00:26:13,800 --> 00:26:17,760
So that generally means that you
can reduce your operating plan

532
00:26:17,760 --> 00:26:20,190
requirements in the summer.

533
00:26:20,190 --> 00:26:21,450
That is a wonderful.

534
00:26:21,450 --> 00:26:22,500
Why?

535
00:26:22,500 --> 00:26:28,130
Because many drivers would like
to take vacation in the summer,

536
00:26:28,130 --> 00:26:29,980
OK.

537
00:26:29,980 --> 00:26:34,830
I mean, unless you're a serious
outdoor skier, ice skater,

538
00:26:34,830 --> 00:26:36,720
whatever have you in
the winter, you probably

539
00:26:36,720 --> 00:26:39,660
don't want to take
vacation in the winter

540
00:26:39,660 --> 00:26:43,020
because your children
are still in school.

541
00:26:43,020 --> 00:26:44,880
Parents want to go
on family vacations

542
00:26:44,880 --> 00:26:46,440
when the weather is decent.

543
00:26:46,440 --> 00:26:52,410
So this cross-hatched area
here is the vacation allocation

544
00:26:52,410 --> 00:26:54,600
you've made to
fulfill your vacation

545
00:26:54,600 --> 00:26:56,830
liability over the year.

546
00:26:56,830 --> 00:26:58,830
And so what you
see here is we're

547
00:26:58,830 --> 00:27:01,860
providing much more
vacation in the summer

548
00:27:01,860 --> 00:27:04,950
than we are in the
winter, or even

549
00:27:04,950 --> 00:27:08,160
in the spring and the autumn.

550
00:27:08,160 --> 00:27:08,660
Why?

551
00:27:08,660 --> 00:27:11,900
That's because our
demand is down,

552
00:27:11,900 --> 00:27:14,150
but the vacation demand is up.

553
00:27:14,150 --> 00:27:19,600
So that's a way of
minimizing absenteeism.

554
00:27:19,600 --> 00:27:22,660
Makes our workforce
happy, it's a good thing.

555
00:27:22,660 --> 00:27:29,180
So this gives us this
profile here about the number

556
00:27:29,180 --> 00:27:34,580
of people we need to cover the
work and cover the vacations.

557
00:27:34,580 --> 00:27:38,220
So the only thing we haven't
covered is the absence.

558
00:27:38,220 --> 00:27:39,975
And we have to cover
the absence as well,

559
00:27:39,975 --> 00:27:41,350
because those
people aren't there

560
00:27:41,350 --> 00:27:44,930
to work that particular day.

561
00:27:44,930 --> 00:27:47,620
So typically, if you did this--

562
00:27:47,620 --> 00:27:50,140
you hire up at the beginning
of each timetable model

563
00:27:50,140 --> 00:27:51,040
that I talked about--

564
00:27:54,220 --> 00:27:57,850
what you would have is
something like this, in terms

565
00:27:57,850 --> 00:28:03,940
of the extra board coverage--

566
00:28:03,940 --> 00:28:07,330
sort of a constant
level throughout this,

567
00:28:07,330 --> 00:28:09,610
except at the end.

568
00:28:09,610 --> 00:28:13,570
And the end of the
timetable is higher

569
00:28:13,570 --> 00:28:18,700
because people have
made changes and routes.

570
00:28:18,700 --> 00:28:19,660
There's retraining.

571
00:28:19,660 --> 00:28:21,340
There's training
of new operators.

572
00:28:21,340 --> 00:28:22,960
So actually, your
required work is

573
00:28:22,960 --> 00:28:25,870
greatest at the end of the
timetable in the last three

574
00:28:25,870 --> 00:28:28,820
or four weeks.

575
00:28:28,820 --> 00:28:33,550
So that's what this jump here
is, in terms of required work.

576
00:28:33,550 --> 00:28:35,185
However, look at the supply.

577
00:28:38,350 --> 00:28:40,210
Typically what
happens, is you have

578
00:28:40,210 --> 00:28:43,710
attrition over the
course of the timetable.

579
00:28:43,710 --> 00:28:46,470
And you have
increasing absenteeism

580
00:28:46,470 --> 00:28:47,940
over the course
of the timetable.

581
00:28:47,940 --> 00:28:48,840
Why?

582
00:28:48,840 --> 00:28:51,500
Because at the beginning
of the timetable,

583
00:28:51,500 --> 00:28:53,420
people have just
picked that work.

584
00:28:53,420 --> 00:28:56,450
They sort of want to do it.

585
00:28:56,450 --> 00:28:57,830
The grass is always greener.

586
00:28:57,830 --> 00:29:00,590
They're sort of enthusiastic at
the beginning of the timetable.

587
00:29:00,590 --> 00:29:02,090
They get out there,
and they realize

588
00:29:02,090 --> 00:29:04,230
it wasn't all it's
cracked up to be.

589
00:29:04,230 --> 00:29:06,410
And so when they're
not feeling up to it,

590
00:29:06,410 --> 00:29:11,370
they're more likely to be absent
perhaps, as time wears along.

591
00:29:11,370 --> 00:29:13,700
And then, towards
the end, they've

592
00:29:13,700 --> 00:29:16,950
made their new selection of work
for the following timetable.

593
00:29:16,950 --> 00:29:18,960
OK?

594
00:29:18,960 --> 00:29:24,690
So what this suggests is you
have a surplus of operators

595
00:29:24,690 --> 00:29:27,120
available at the
beginning of the timetable

596
00:29:27,120 --> 00:29:30,420
and a significant
deficit of operators

597
00:29:30,420 --> 00:29:33,240
at the end of a timetable.

598
00:29:33,240 --> 00:29:36,550
So you sort of have one day
in the middle of a timetable,

599
00:29:36,550 --> 00:29:38,470
which is perfect.

600
00:29:38,470 --> 00:29:41,020
But otherwise, you're off.

601
00:29:41,020 --> 00:29:44,080
You haven't got this good
balance between the operators

602
00:29:44,080 --> 00:29:45,990
that you have available
and what you need.

603
00:29:53,320 --> 00:29:58,850
So clearly doing workforce
hiring every month,

604
00:29:58,850 --> 00:30:01,450
for instance, is one way
of dealing with this.

605
00:30:01,450 --> 00:30:07,480
So you can reduce this hump
for training new operators

606
00:30:07,480 --> 00:30:13,210
and spread it over the whole
period because, in round terms,

607
00:30:13,210 --> 00:30:16,400
training for an operator
takes about a month

608
00:30:16,400 --> 00:30:18,040
of on-the-road training.

609
00:30:18,040 --> 00:30:22,315
And so you spread that over
the month, over the timetable.

610
00:30:26,040 --> 00:30:30,110
So going to monthly hiring
makes a lot of sense.

611
00:30:30,110 --> 00:30:32,630
And again, you
avoid these problems

612
00:30:32,630 --> 00:30:34,360
having unassigned cover time.

613
00:30:34,360 --> 00:30:35,860
What that means is
you've got people

614
00:30:35,860 --> 00:30:39,640
who are being paid not to be
very productive in the garage.

615
00:30:39,640 --> 00:30:41,890
You haven't got very, very
good things for them to do.

616
00:30:41,890 --> 00:30:43,473
Because otherwise,
you'd have built it

617
00:30:43,473 --> 00:30:44,590
into the operating plan.

618
00:30:44,590 --> 00:30:47,890
And large amounts of
overtime and unreliability

619
00:30:47,890 --> 00:30:49,790
at the end of the timetable.

620
00:30:49,790 --> 00:30:51,980
And I don't know what
your experience is,

621
00:30:51,980 --> 00:30:55,480
but in my experience,
that's exactly what you see.

622
00:30:55,480 --> 00:30:59,500
At the end of a timetable,
service goes down the tubes,

623
00:30:59,500 --> 00:31:04,120
because you get this imbalance
between higher absence

624
00:31:04,120 --> 00:31:06,190
and higher requirements
for running

625
00:31:06,190 --> 00:31:07,540
into the next timetable.

626
00:31:07,540 --> 00:31:11,490
So you want to try
and avoid that.

627
00:31:11,490 --> 00:31:11,990
OK.

628
00:31:11,990 --> 00:31:15,510
So let's look at this from
a total cost perspective.

629
00:31:15,510 --> 00:31:17,210
This is a very simple sketch.

630
00:31:17,210 --> 00:31:19,610
I'll show you something more
realistic in a few minutes.

631
00:31:19,610 --> 00:31:22,880
But this is the
conceptual part of this.

632
00:31:22,880 --> 00:31:25,200
So on the x-axis--

633
00:31:25,200 --> 00:31:27,140
and I apologize,
this is a typo--

634
00:31:27,140 --> 00:31:28,880
that's number of extraboards.

635
00:31:28,880 --> 00:31:31,430
The extraboard is
the cover list--

636
00:31:31,430 --> 00:31:34,460
the people who are there
to cover absences or do

637
00:31:34,460 --> 00:31:37,580
extra work that's required
not built into the timetable.

638
00:31:37,580 --> 00:31:40,030
These are extraboards
or cover lists--

639
00:31:40,030 --> 00:31:41,240
whatever.

640
00:31:41,240 --> 00:31:43,310
And that should be the
number of drivers--

641
00:31:43,310 --> 00:31:46,130
"extraboard operators," not
"extraboard operations."

642
00:31:48,810 --> 00:31:52,090
And on the y-axis,
you have the cost.

643
00:31:52,090 --> 00:31:56,340
So if you have a small number
of extraboard operators,

644
00:31:56,340 --> 00:32:01,330
your fringe benefit
costs are low.

645
00:32:03,850 --> 00:32:05,930
Your fixed fringe benefit
cost will increase

646
00:32:05,930 --> 00:32:07,580
as you increase the number.

647
00:32:07,580 --> 00:32:11,110
But you'll be relying
very heavily on overtime.

648
00:32:11,110 --> 00:32:14,890
So you'll be having high
direct payment costs because

649
00:32:14,890 --> 00:32:16,390
of the reliance on overtime.

650
00:32:16,390 --> 00:32:20,440
And that gives you an
overall high cost solution

651
00:32:20,440 --> 00:32:22,840
beyond some point where
you get a good balance

652
00:32:22,840 --> 00:32:26,860
between the costs of overtime
and the cost of fixed fringe

653
00:32:26,860 --> 00:32:27,970
benefits.

654
00:32:27,970 --> 00:32:31,960
Then you'll get a sweet spot
you'll want to try and operate.

655
00:32:31,960 --> 00:32:34,655
And you want to avoid
this, where you've

656
00:32:34,655 --> 00:32:36,030
got so many
operators that you've

657
00:32:36,030 --> 00:32:38,030
got a lot of people who
don't have a lot of work

658
00:32:38,030 --> 00:32:39,460
to do on many days.

659
00:32:43,450 --> 00:32:44,950
Yes, question?

660
00:32:44,950 --> 00:32:45,760
Comment?

661
00:32:45,760 --> 00:32:47,510
GABRIEL SANCHEZ-MARTINEZ:
It's a question.

662
00:32:47,510 --> 00:32:49,060
NIGEL HM WILSON: OK.

663
00:32:49,060 --> 00:32:50,727
GABRIEL SANCHEZ-MARTINEZ:
I didn't quite

664
00:32:50,727 --> 00:32:53,130
understand why monthly
hiring would reduce the time.

665
00:32:53,130 --> 00:32:53,460
NIGEL HM WILSON: OK.

666
00:32:53,460 --> 00:32:54,490
GABRIEL SANCHEZ-MARTINEZ:
Can you explain that?

667
00:32:54,490 --> 00:32:54,940
NIGEL HM WILSON: Yeah, yeah.

668
00:32:54,940 --> 00:32:55,440
Sure.

669
00:32:58,970 --> 00:33:06,590
OK, so-- sorry, yeah.

670
00:33:06,590 --> 00:33:08,780
Thanks, Gabriel,
for reminding me.

671
00:33:08,780 --> 00:33:11,780
It will take care of this hump
at the end of the timetable.

672
00:33:11,780 --> 00:33:14,030
If there really is
an extra requirement

673
00:33:14,030 --> 00:33:15,800
for work at the
end of a timetable,

674
00:33:15,800 --> 00:33:19,140
you can hire up for
that a month before.

675
00:33:19,140 --> 00:33:24,560
So if you think of this
as a sawtooth pattern, OK,

676
00:33:24,560 --> 00:33:26,540
which is what it is--
you've got attrition,

677
00:33:26,540 --> 00:33:28,980
and then you've
got a bulk hiring--

678
00:33:28,980 --> 00:33:33,330
if you reduce the
periodicity of the saws--

679
00:33:33,330 --> 00:33:36,820
the gaps-- you'll
reduce the integrated

680
00:33:36,820 --> 00:33:39,530
imbalance between the
requirements and what you need.

681
00:33:39,530 --> 00:33:44,080
So that's, in essence,
what's going on here.

682
00:33:44,080 --> 00:33:44,740
Thanks.

683
00:33:44,740 --> 00:33:45,930
Good question.

684
00:33:45,930 --> 00:33:51,190
OK so this is sort of the
classic schema for analyzing

685
00:33:51,190 --> 00:33:55,170
this problem, for modeling it.

686
00:33:55,170 --> 00:34:01,950
So here on the x-axis, we
have number of scheduled runs.

687
00:34:01,950 --> 00:34:06,720
This is for an organization
that sort of looks a lot a lot

688
00:34:06,720 --> 00:34:07,345
like the MBTA.

689
00:34:11,030 --> 00:34:22,260
And this histogram here is
the cumulative distribution

690
00:34:22,260 --> 00:34:27,590
for number of absence days
in the previous period.

691
00:34:27,590 --> 00:34:28,550
OK.

692
00:34:28,550 --> 00:34:35,150
So now we're running into the
summer timetable for the MBTA.

693
00:34:35,150 --> 00:34:38,150
So when the MBTA was doing
the analysis on this,

694
00:34:38,150 --> 00:34:40,250
they were saying, well,
what level of absence

695
00:34:40,250 --> 00:34:41,780
can we expect this
coming summer?

696
00:34:41,780 --> 00:34:44,312
Well, we'll start by
looking at last summer.

697
00:34:44,312 --> 00:34:45,770
That's probably a
fairly good proxy

698
00:34:45,770 --> 00:34:47,489
for what we can
expect this summer.

699
00:34:47,489 --> 00:34:50,150
And so the way they
would look at that

700
00:34:50,150 --> 00:34:54,370
is look at the number
of absent operators

701
00:34:54,370 --> 00:35:01,190
by day of week over the
full range of experiences

702
00:35:01,190 --> 00:35:02,480
from last summer.

703
00:35:02,480 --> 00:35:06,800
So on your best day last
summer for the MBTA--

704
00:35:06,800 --> 00:35:09,050
these are hypothetical
numbers, but I want to give you

705
00:35:09,050 --> 00:35:10,670
the flavor of this--

706
00:35:10,670 --> 00:35:12,770
you had 16 absences.

707
00:35:12,770 --> 00:35:16,250
16 drivers didn't show up for
work, for whatever reason.

708
00:35:16,250 --> 00:35:21,980
On your worst day,
you had 60 operators

709
00:35:21,980 --> 00:35:24,660
who didn't show up
for work that day.

710
00:35:24,660 --> 00:35:28,030
So it's a tremendous range.

711
00:35:28,030 --> 00:35:28,530
Question?

712
00:35:28,530 --> 00:35:28,855
AUDIENCE: Sorry.

713
00:35:28,855 --> 00:35:29,180
AUDIENCE: Yeah.

714
00:35:29,180 --> 00:35:30,695
AUDIENCE: Can you explain how
the chart works a little bit

715
00:35:30,695 --> 00:35:30,800
more?

716
00:35:30,800 --> 00:35:31,717
NIGEL HM WILSON: Yeah.

717
00:35:31,717 --> 00:35:32,970
OK, so yeah.

718
00:35:32,970 --> 00:35:36,660
I'm trying to do a lot
in one chance here.

719
00:35:36,660 --> 00:35:39,075
So this, if we
look at that, it's

720
00:35:39,075 --> 00:35:44,250
the cumulative distribution
of absence days.

721
00:35:44,250 --> 00:35:50,880
So this is the percent of
days which have absences less

722
00:35:50,880 --> 00:35:52,223
than this number.

723
00:35:52,223 --> 00:35:53,640
So it's a cumulative
distribution,

724
00:35:53,640 --> 00:35:55,900
cumulative density function.

725
00:35:55,900 --> 00:35:59,790
So what we're saying
here is that every day

726
00:35:59,790 --> 00:36:03,760
had no more than 60
operators absent.

727
00:36:03,760 --> 00:36:07,860
Every day had at least
16 operators absent.

728
00:36:07,860 --> 00:36:10,200
And this is the
historical record

729
00:36:10,200 --> 00:36:11,980
of absences over that period.

730
00:36:11,980 --> 00:36:15,180
Now, the reason I've
drawn this line here

731
00:36:15,180 --> 00:36:19,500
is to show you what the
extraboard sizing trade off is.

732
00:36:19,500 --> 00:36:25,530
So we're looking at, say, 31
extraboard operators here.

733
00:36:25,530 --> 00:36:28,380
If we have 31 extra board
operators and we assume

734
00:36:28,380 --> 00:36:30,390
there's no absence in the
extraboard operators--

735
00:36:30,390 --> 00:36:33,080
of course, there will be--

736
00:36:33,080 --> 00:36:38,190
then we will be able to cover
all of the absences in area B

737
00:36:38,190 --> 00:36:40,920
without paying any overtime.

738
00:36:40,920 --> 00:36:45,000
But all of the absences
on days in area A,

739
00:36:45,000 --> 00:36:47,790
we'll have to pay
overtime for, and we'll

740
00:36:47,790 --> 00:36:54,380
have to deal with reliability
to drop trips on those days.

741
00:36:54,380 --> 00:36:56,110
So the first point
I want to make

742
00:36:56,110 --> 00:37:00,340
is this is a very simple
model, but it's very useful.

743
00:37:00,340 --> 00:37:02,110
It provides a lot
of insight into how

744
00:37:02,110 --> 00:37:04,950
you make these decisions.

745
00:37:04,950 --> 00:37:08,030
The first point to
make is, my god,

746
00:37:08,030 --> 00:37:10,520
that's a lot of variability--

747
00:37:10,520 --> 00:37:14,470
from 16 to 60 operators
absent on different days

748
00:37:14,470 --> 00:37:16,010
of the timetable.

749
00:37:16,010 --> 00:37:20,260
So probably some of
that is systematic.

750
00:37:20,260 --> 00:37:24,050
So some of them, probably you
have maybe a higher absence

751
00:37:24,050 --> 00:37:28,140
on Mondays and Fridays than you
have on Tuesdays, Wednesdays,

752
00:37:28,140 --> 00:37:30,530
and Thursdays.

753
00:37:30,530 --> 00:37:33,640
So you probably want
to do this analysis

754
00:37:33,640 --> 00:37:37,220
at the day-of-week level, OK.

755
00:37:37,220 --> 00:37:39,710
Maybe you can group Tuesdays,
Wednesdays, Thursdays.

756
00:37:39,710 --> 00:37:42,960
Maybe you can group
Mondays and Fridays.

757
00:37:42,960 --> 00:37:44,750
But the principles are the same.

758
00:37:44,750 --> 00:37:49,910
So there are different
ways of classifying days

759
00:37:49,910 --> 00:37:51,240
to reduce the variance.

760
00:37:51,240 --> 00:37:53,448
But you're basically going
to have this sort of trade

761
00:37:53,448 --> 00:37:57,560
off between your [AUDIO OUT]
--terministic number

762
00:37:57,560 --> 00:37:59,660
from this cumulative
density function,

763
00:37:59,660 --> 00:38:01,003
cumulative distribution.

764
00:38:01,003 --> 00:38:02,420
Is everyone
comfortable with this?

765
00:38:02,420 --> 00:38:03,940
Because this is important.

766
00:38:03,940 --> 00:38:04,455
Yeah.

767
00:38:04,455 --> 00:38:06,070
AUDIENCE: I'm just
curious about--

768
00:38:06,070 --> 00:38:08,570
so this is a broad spin on
public transportation here.

769
00:38:08,570 --> 00:38:11,960
But I'm thinking of like
delivery companies might

770
00:38:11,960 --> 00:38:13,060
be very similar.

771
00:38:13,060 --> 00:38:14,130
NIGEL HM WILSON: Yes.

772
00:38:14,130 --> 00:38:14,630
Yeah.

773
00:38:14,630 --> 00:38:17,420
AUDIENCE: [INAUDIBLE] consulted
by them for this type of work?

774
00:38:17,420 --> 00:38:17,920
[INAUDIBLE]

775
00:38:17,920 --> 00:38:19,420
NIGEL HM WILSON: I
don't personally,

776
00:38:19,420 --> 00:38:21,395
but the problems
are very similar.

777
00:38:24,790 --> 00:38:27,770
And the airlines have had
very similar problems.

778
00:38:27,770 --> 00:38:29,690
I was on a trip,
not too long ago,

779
00:38:29,690 --> 00:38:32,690
where I was delayed at
Logan Airport for an hour

780
00:38:32,690 --> 00:38:36,320
because they had an
unexpected absence.

781
00:38:36,320 --> 00:38:39,770
Their reserved crew
member happened

782
00:38:39,770 --> 00:38:41,450
to live in Nashua,
New Hampshire, which

783
00:38:41,450 --> 00:38:47,120
a lot of airline pilots
live in because it doesn't

784
00:38:47,120 --> 00:38:49,250
have any taxes in that state.

785
00:38:49,250 --> 00:38:51,597
And they had to drive
down, in the peak hour,

786
00:38:51,597 --> 00:38:52,430
to get to the plane.

787
00:38:52,430 --> 00:38:55,020
I was delayed for an hour.

788
00:38:55,020 --> 00:38:58,590
So every industry has
the same sort of problem

789
00:38:58,590 --> 00:39:03,803
when it's a transportation
industry, where

790
00:39:03,803 --> 00:39:04,970
you don't have a shelf life.

791
00:39:04,970 --> 00:39:06,220
It's not a production process.

792
00:39:06,220 --> 00:39:06,920
It's a service.

793
00:39:06,920 --> 00:39:10,620
And you have to deliver it at
that time, in that location.

794
00:39:10,620 --> 00:39:13,370
So there are similar problems.

795
00:39:13,370 --> 00:39:16,580
Again, it's not totally
dissimilar from staffing

796
00:39:16,580 --> 00:39:19,282
in hospitals for nurses, for
doctors, everything like that.

797
00:39:19,282 --> 00:39:20,990
So there's a whole
class of problems here

798
00:39:20,990 --> 00:39:22,350
which are quite similar.

799
00:39:22,350 --> 00:39:26,945
And I'm just presenting the
particular variant of this

800
00:39:26,945 --> 00:39:29,470
for the transit industry.

801
00:39:29,470 --> 00:39:32,610
Other questions or
comments on this?

802
00:39:32,610 --> 00:39:33,110
Yeah.

803
00:39:33,110 --> 00:39:34,735
AUDIENCE: Is this
axis mislabeled here?

804
00:39:34,735 --> 00:39:38,672
Should it [INAUDIBLE]
drivers, not scheduled runs?

805
00:39:38,672 --> 00:39:39,630
NIGEL HM WILSON: Sorry.

806
00:39:39,630 --> 00:39:40,330
[CHUCKLES]

807
00:39:41,510 --> 00:39:42,010
No.

808
00:39:42,010 --> 00:39:46,764
No, no-- it could be mislabeled.

809
00:39:49,715 --> 00:39:52,050
What I'm trying to
say is, these are

810
00:39:52,050 --> 00:39:53,750
the number of
scheduled operators

811
00:39:53,750 --> 00:39:55,370
who are absent on that day.

812
00:39:55,370 --> 00:39:57,980
That's what it really means, OK.

813
00:39:57,980 --> 00:40:00,363
AUDIENCE: Runs for which
the scheduled operator--

814
00:40:00,363 --> 00:40:01,280
NIGEL HM WILSON: Yeah.

815
00:40:01,280 --> 00:40:06,050
So the analysis is the sort
of full day for a full-timer.

816
00:40:06,050 --> 00:40:07,340
That's sort of the unit.

817
00:40:07,340 --> 00:40:10,640
And you can label the
x-axis in different ways.

818
00:40:10,640 --> 00:40:12,990
And this probably isn't
the best possible label.

819
00:40:12,990 --> 00:40:15,600
But that's what
I'm getting at, OK.

820
00:40:15,600 --> 00:40:16,970
Other questions or comments?

821
00:40:20,150 --> 00:40:26,540
OK, so let's quickly look at
the strategic-level model here.

822
00:40:26,540 --> 00:40:29,340
Decision variables--
are what we're

823
00:40:29,340 --> 00:40:34,130
trying to decide is the
workforce size for each period.

824
00:40:34,130 --> 00:40:36,650
And typically, as I
said, maybe the period

825
00:40:36,650 --> 00:40:39,948
of a month, four weeks.

826
00:40:39,948 --> 00:40:45,410
And vacation allocation for
each period and the vacation

827
00:40:45,410 --> 00:40:49,800
allocation will typically be
over the course of a timetable.

828
00:40:49,800 --> 00:40:52,260
So what's the average
number of vacation weeks

829
00:40:52,260 --> 00:40:56,850
we want to provide in the
summer at a particular garage

830
00:40:56,850 --> 00:40:58,890
or across the whole system?

831
00:40:58,890 --> 00:41:03,420
We'll provide the same
number of vacation weeks

832
00:41:03,420 --> 00:41:06,400
for all the bus
system in the MBTA.

833
00:41:06,400 --> 00:41:09,270
And there's someone who's
covering the vacation--

834
00:41:09,270 --> 00:41:11,760
just is covering the
different vacations

835
00:41:11,760 --> 00:41:13,680
every week of the timetable.

836
00:41:13,680 --> 00:41:16,110
So we have to have
the same number

837
00:41:16,110 --> 00:41:18,000
throughout the timetable.

838
00:41:18,000 --> 00:41:19,950
OK?

839
00:41:19,950 --> 00:41:23,970
And the optimal hiring level--
so you get the right number

840
00:41:23,970 --> 00:41:25,590
throughout the year.

841
00:41:25,590 --> 00:41:30,670
That's sort of what we're
trying to achieve here.

842
00:41:30,670 --> 00:41:34,150
And the objective is to
minimize the workforce cost,

843
00:41:34,150 --> 00:41:37,230
which includes the cost
of the schedule runs,

844
00:41:37,230 --> 00:41:39,480
the costs of the extraboard--
the people who are there

845
00:41:39,480 --> 00:41:42,030
to cover the absences
and the extra work--

846
00:41:42,030 --> 00:41:43,375
and the cost of overtime.

847
00:41:47,180 --> 00:41:50,240
And you'll notice there
that I didn't include

848
00:41:50,240 --> 00:41:52,490
costs of reliability
or unreliability,

849
00:41:52,490 --> 00:41:56,900
because I'm including
that in the constraint.

850
00:41:56,900 --> 00:41:58,960
I could have included in
the objective function.

851
00:41:58,960 --> 00:42:00,938
But I'm including it
in the constraint,

852
00:42:00,938 --> 00:42:02,480
just because it's
a little bit easier

853
00:42:02,480 --> 00:42:04,160
to conceptualize it that way.

854
00:42:04,160 --> 00:42:06,710
And we were doing this
work for the MBTA, who

855
00:42:06,710 --> 00:42:09,740
kind of like to think of
overtime as a constraint.

856
00:42:09,740 --> 00:42:12,010
Because overtime, as
you're probably aware,

857
00:42:12,010 --> 00:42:16,660
gets a lot of attention in
the medial, in the press.

858
00:42:16,660 --> 00:42:18,890
You know, "10% of
our budget is being

859
00:42:18,890 --> 00:42:21,020
spent on overtime for drivers."

860
00:42:21,020 --> 00:42:23,670
Those are headline
sorts of issues.

861
00:42:23,670 --> 00:42:25,820
So the MBTA and transit
agency generally

862
00:42:25,820 --> 00:42:27,950
are sensitive to the
level of overtime.

863
00:42:27,950 --> 00:42:30,590
That's why I structured
this as a constraint,

864
00:42:30,590 --> 00:42:34,940
rather than as a
multi-objective function--

865
00:42:34,940 --> 00:42:38,200
[INAUDIBLE] multiple objectives.

866
00:42:38,200 --> 00:42:40,330
Constraints we need to
meet-- vacation liability,

867
00:42:40,330 --> 00:42:45,120
overtime service reliability,
part-time operator constraint--

868
00:42:45,120 --> 00:42:46,750
sorry.

869
00:42:46,750 --> 00:42:49,120
And any other policy
constraints that we want to

870
00:42:49,120 --> 00:42:52,780
can be included in this
optimization problem.

871
00:42:52,780 --> 00:42:58,720
And the reliability model that
we have is basically this.

872
00:42:58,720 --> 00:43:01,590
So what this is saying
is we have open work.

873
00:43:01,590 --> 00:43:06,230
What open work means is we
have work we want to do.

874
00:43:06,230 --> 00:43:09,310
We don't have anyone who's
going to be willing to do it

875
00:43:09,310 --> 00:43:11,050
without paying overtime.

876
00:43:11,050 --> 00:43:13,100
That's what I mean by open work.

877
00:43:13,100 --> 00:43:15,010
It's work that wasn't
a regularly scheduled

878
00:43:15,010 --> 00:43:17,620
employee available to cover.

879
00:43:17,620 --> 00:43:19,720
So overtime is available.

880
00:43:19,720 --> 00:43:25,990
If an operator is available
and willing to work time,

881
00:43:25,990 --> 00:43:27,550
you come down here.

882
00:43:27,550 --> 00:43:28,930
And overtime is worked.

883
00:43:28,930 --> 00:43:30,610
That may produce more absence.

884
00:43:30,610 --> 00:43:34,120
May have those negative
feedback effects.

885
00:43:34,120 --> 00:43:38,200
But at least you've delivered
the service that day.

886
00:43:38,200 --> 00:43:41,300
If there's no operator
available or there's

887
00:43:41,300 --> 00:43:44,033
no operator willing to work the
overtime, you drop the trips.

888
00:43:44,033 --> 00:43:46,158
And you've got a reliability
problem on your hands.

889
00:43:51,043 --> 00:43:52,960
Don't worry about the
individual points there.

890
00:43:52,960 --> 00:44:02,360
But this is showing for the
bus network on the MBTA,

891
00:44:02,360 --> 00:44:05,202
on the x-axis, the open work.

892
00:44:05,202 --> 00:44:09,760
This is for a particular
timetable, system-wide

893
00:44:09,760 --> 00:44:13,510
for the bus, and the
number of missed trip hours

894
00:44:13,510 --> 00:44:15,200
that were incurred that day.

895
00:44:15,200 --> 00:44:19,930
So each point here
represents a day of operation

896
00:44:19,930 --> 00:44:22,340
on the MBTA bus system.

897
00:44:22,340 --> 00:44:27,710
And so this particular day,
you had close to 300 hours

898
00:44:27,710 --> 00:44:29,630
of open work.

899
00:44:29,630 --> 00:44:34,825
And you dropped close
to 150 hours of service.

900
00:44:37,550 --> 00:44:41,260
And if you do an OLS
regression on this,

901
00:44:41,260 --> 00:44:45,370
this was the best fit we got.

902
00:44:45,370 --> 00:44:48,850
So that means for every
100 hours of open work

903
00:44:48,850 --> 00:44:54,810
at the time we did this
analysis that you wanted to get,

904
00:44:54,810 --> 00:44:57,390
you were only able to get 72.

905
00:44:57,390 --> 00:45:01,370
And you wound up dropping
28 hours of service.

906
00:45:01,370 --> 00:45:03,140
That is pretty tough.

907
00:45:03,140 --> 00:45:06,570
That's actually
really quite bad.

908
00:45:06,570 --> 00:45:08,790
By comparison purposes,
I've done other work

909
00:45:08,790 --> 00:45:12,100
in Montreal,
Ottawa, and Toronto,

910
00:45:12,100 --> 00:45:16,230
which are pretty good, more or
less, east coast properties.

911
00:45:16,230 --> 00:45:20,050
In Canada, they're
never more than 10%.

912
00:45:20,050 --> 00:45:24,840
So there's something not
being done very efficiently

913
00:45:24,840 --> 00:45:29,380
in the way the MBTA
covers their overtime.

914
00:45:29,380 --> 00:45:30,960
And I think part
of this is probably

915
00:45:30,960 --> 00:45:36,390
over-reliance on seniority,
again, and probably not

916
00:45:36,390 --> 00:45:38,130
doing enough advance planning.

917
00:45:38,130 --> 00:45:47,370
And what do I mean by relying
too heavily on seniority?

918
00:45:47,370 --> 00:45:50,820
How many phone calls do you make
to the most senior operator who

919
00:45:50,820 --> 00:45:53,550
might be available before
you say, he or she isn't

920
00:45:53,550 --> 00:45:54,960
going to do this
work, and you go

921
00:45:54,960 --> 00:45:57,560
to the next person on the list?

922
00:45:57,560 --> 00:45:59,590
And how much advance
planning do you

923
00:45:59,590 --> 00:46:03,850
do to make sure that you've
got a set of operators who

924
00:46:03,850 --> 00:46:07,980
are really going to be available
if you have a need for them

925
00:46:07,980 --> 00:46:10,490
at 9 o'clock tomorrow morning?

926
00:46:10,490 --> 00:46:13,300
Those are the things that you
could do to try and reduce

927
00:46:13,300 --> 00:46:17,860
this 28% missed trips to 5%
or 10% missed trips, which

928
00:46:17,860 --> 00:46:21,692
would be a much better situation
for everyone concerned.

929
00:46:24,530 --> 00:46:26,970
So this is the MBTA
case study again.

930
00:46:26,970 --> 00:46:30,830
Bear in mind that this
was basically mid 1990s.

931
00:46:30,830 --> 00:46:31,810
So it's 20 years ago.

932
00:46:34,540 --> 00:46:36,850
But also, believe
me, nothing much

933
00:46:36,850 --> 00:46:38,980
has changed in this
regard at the MBTA.

934
00:46:38,980 --> 00:46:41,380
I'm not saying nothing's
changed in the MBTA

935
00:46:41,380 --> 00:46:43,480
in any regard over
the last 20 years.

936
00:46:43,480 --> 00:46:47,850
But certainly in this regard,
nothing much has changed.

937
00:46:47,850 --> 00:46:50,170
And I think there are still
some of the same problems

938
00:46:50,170 --> 00:46:53,350
that I'll talk about here
still in existence, which

939
00:46:53,350 --> 00:46:55,720
is sad for me to admit
because, obviously, I

940
00:46:55,720 --> 00:46:57,200
was trying to help them--

941
00:46:57,200 --> 00:47:00,200
didn't succeed.

942
00:47:00,200 --> 00:47:02,950
So at that point, the
part-time workforce

943
00:47:02,950 --> 00:47:05,980
was sized to about 40% of
the full-time workforce.

944
00:47:05,980 --> 00:47:09,930
That's a little bit
lower now, I believe.

945
00:47:09,930 --> 00:47:14,950
There's a large variability
in required work hours.

946
00:47:14,950 --> 00:47:16,550
So what do I mean by that?

947
00:47:16,550 --> 00:47:19,350
The mean daily
absence and extra work

948
00:47:19,350 --> 00:47:22,200
required-- this is apart
from the timetable.

949
00:47:22,200 --> 00:47:24,540
So this is shuttling
service when

950
00:47:24,540 --> 00:47:29,730
you get a line down
for signaling problem

951
00:47:29,730 --> 00:47:35,530
or disabled train or whatever it
is, or fatality on the tracks.

952
00:47:35,530 --> 00:47:37,650
The mean daily
absence and extra work

953
00:47:37,650 --> 00:47:40,020
for the whole MBTA
bus system was

954
00:47:40,020 --> 00:47:44,340
1,250 hours on a daily basis.

955
00:47:44,340 --> 00:47:47,220
And the standard deviation
of that was 290 hours.

956
00:47:47,220 --> 00:47:50,370
AUDIENCE: What percent
of that is total hours?

957
00:47:50,370 --> 00:47:51,750
NIGEL HM WILSON: It's coming up.

958
00:47:55,398 --> 00:47:56,940
I'm getting into it
in a few minutes.

959
00:47:56,940 --> 00:48:01,517
But round figure's 10%.

960
00:48:01,517 --> 00:48:02,100
AUDIENCE: Wow.

961
00:48:02,100 --> 00:48:03,380
NIGEL HM WILSON:
It's pretty high.

962
00:48:03,380 --> 00:48:04,450
That's not just absence.

963
00:48:04,450 --> 00:48:07,090
It doesn't include extra work.

964
00:48:07,090 --> 00:48:07,803
OK?

965
00:48:07,803 --> 00:48:10,220
AUDIENCE: But it does
include sick leave, as well?

966
00:48:10,220 --> 00:48:10,670
NIGEL HM WILSON: Yes.

967
00:48:10,670 --> 00:48:11,070
AUDIENCE: OK.

968
00:48:11,070 --> 00:48:11,490
NIGEL HM WILSON: Yes.

969
00:48:11,490 --> 00:48:12,100
AUDIENCE: Any kind of absence.

970
00:48:12,100 --> 00:48:12,975
NIGEL HM WILSON: Yes.

971
00:48:12,975 --> 00:48:14,420
Any sort of absence, exactly.

972
00:48:14,420 --> 00:48:17,125
So it's not in any
sense absenteeism.

973
00:48:17,125 --> 00:48:18,500
It's the amount
of extra work you

974
00:48:18,500 --> 00:48:22,625
have to do beyond what's
the scheduled work,

975
00:48:22,625 --> 00:48:25,800
the scheduled operating plan.

976
00:48:25,800 --> 00:48:29,750
But anyway, it's quite large,
and it's very variable.

977
00:48:29,750 --> 00:48:32,170
So it is highly stochastic.

978
00:48:32,170 --> 00:48:35,310
Now, these figures
were from 1996.

979
00:48:35,310 --> 00:48:38,740
And so the numbers will
have changed significantly

980
00:48:38,740 --> 00:48:40,190
over the last 20 years.

981
00:48:40,190 --> 00:48:42,770
But the relative
ratio of the numbers

982
00:48:42,770 --> 00:48:44,650
in the different
columns and cells

983
00:48:44,650 --> 00:48:48,380
has not changed significantly
in the last 20 years.

984
00:48:48,380 --> 00:48:49,780
So what this is saying--

985
00:48:49,780 --> 00:48:56,627
what we're showing here is the
driver cost per platform hour.

986
00:49:00,370 --> 00:49:04,360
And so this includes
the wage rate

987
00:49:04,360 --> 00:49:09,080
and benefits-- fixed and
variable fringe benefits.

988
00:49:09,080 --> 00:49:18,100
So an hour of overtime worked
by a driver at that time

989
00:49:18,100 --> 00:49:19,930
was around $30 an hour.

990
00:49:23,710 --> 00:49:27,310
So sorry, the first
row is the wage rate.

991
00:49:27,310 --> 00:49:30,040
The second row is the full cost.

992
00:49:30,040 --> 00:49:33,550
So the difference here is the
benefits that are included,

993
00:49:33,550 --> 00:49:38,000
and the productivity changes
between overtime, part-time,

994
00:49:38,000 --> 00:49:39,010
and full-timers.

995
00:49:39,010 --> 00:49:43,060
So for overtime, you get
a very small benefit load

996
00:49:43,060 --> 00:49:47,570
because most of the benefits are
covered through their base job,

997
00:49:47,570 --> 00:49:48,070
OK.

998
00:49:48,070 --> 00:49:50,980
So the marginal cost
of overtime work

999
00:49:50,980 --> 00:49:53,530
in terms of the benefit
load is quite small,

1000
00:49:53,530 --> 00:49:56,470
in the order of less than 10%--

1001
00:49:56,470 --> 00:49:58,590
or just about 10% of this.

1002
00:49:58,590 --> 00:50:01,570
Whereas for part-timers
and full-timers,

1003
00:50:01,570 --> 00:50:06,200
full-timers get somewhat
greater benefits.

1004
00:50:06,200 --> 00:50:08,650
So it's a significant increment.

1005
00:50:08,650 --> 00:50:11,170
So if you just scan
across here, this

1006
00:50:11,170 --> 00:50:13,990
gives you the straight,
direct comparison

1007
00:50:13,990 --> 00:50:17,590
between the cost of an hour of
work produced on an overtime

1008
00:50:17,590 --> 00:50:21,400
basis by a full-timer
or by a part-timer.

1009
00:50:21,400 --> 00:50:25,660
And in round terms,
you say, well,

1010
00:50:25,660 --> 00:50:29,290
if it's 40% part-timer,
60% full-timer, as the mix

1011
00:50:29,290 --> 00:50:34,630
of workforce, versus $32.72
for an overtime, it's a wash.

1012
00:50:34,630 --> 00:50:37,070
It's basically the same.

1013
00:50:37,070 --> 00:50:37,770
OK.

1014
00:50:37,770 --> 00:50:41,950
So strictly in terms of cost,
it doesn't really matter.

1015
00:50:41,950 --> 00:50:44,540
It matters a lot if you
have far too many people.

1016
00:50:44,540 --> 00:50:48,150
But it doesn't matter too
much in terms of the cost

1017
00:50:48,150 --> 00:50:50,390
if you have about
the right number

1018
00:50:50,390 --> 00:50:53,160
or slightly less
than that number.

1019
00:50:53,160 --> 00:50:56,520
If you have a lot
less than that number,

1020
00:50:56,520 --> 00:50:59,220
the short-run financial
costs won't change,

1021
00:50:59,220 --> 00:51:01,230
but you'll have
reliability problems

1022
00:51:01,230 --> 00:51:04,570
because you won't be
able to get the overtime.

1023
00:51:04,570 --> 00:51:07,500
However, what we really
should be looking at

1024
00:51:07,500 --> 00:51:12,570
is the marginal cost of the
last person on the cover list.

1025
00:51:12,570 --> 00:51:15,540
That's where we're making
the marginal decision about

1026
00:51:15,540 --> 00:51:17,610
whether we want to increase
the cover list by 1

1027
00:51:17,610 --> 00:51:20,730
or decrease it by 1 or
hold it where it is.

1028
00:51:20,730 --> 00:51:24,830
So if the last person
on the cover list

1029
00:51:24,830 --> 00:51:29,930
is being used 50% of
the time, then you're

1030
00:51:29,930 --> 00:51:34,720
making a trade off between
$32.72 for an overtime hour

1031
00:51:34,720 --> 00:51:41,170
versus $69.56 for the marginal
person on the cover list.

1032
00:51:43,850 --> 00:51:50,110
OK, so what a critical issue
is, in terms of any of you guys

1033
00:51:50,110 --> 00:51:51,850
doing analysis for
transit agency,

1034
00:51:51,850 --> 00:51:55,780
be it JR-East or the MBTA or
whatever it is, in this area,

1035
00:51:55,780 --> 00:51:59,320
you really want to look at what
the distribution of the costs

1036
00:51:59,320 --> 00:52:02,410
are of the cover list.

1037
00:52:02,410 --> 00:52:06,330
And a good proxy
for that is who's

1038
00:52:06,330 --> 00:52:09,600
picking the work
on the cover list.

1039
00:52:09,600 --> 00:52:11,300
If it's the most
senior operators who

1040
00:52:11,300 --> 00:52:13,460
are picking the work
on the cover list,

1041
00:52:13,460 --> 00:52:15,230
that means that's
pretty easy work.

1042
00:52:18,020 --> 00:52:21,050
If the senior operators are
not picking the cover list

1043
00:52:21,050 --> 00:52:24,380
and it's being left to the
junior people on the seniority

1044
00:52:24,380 --> 00:52:29,010
list, that means the senior
operators want to avoid it

1045
00:52:29,010 --> 00:52:31,770
because it's difficult work.

1046
00:52:31,770 --> 00:52:35,130
So that's a sort of
very simple shortcut--

1047
00:52:35,130 --> 00:52:38,020
do we have an oversized cover
list or an undersized cover

1048
00:52:38,020 --> 00:52:38,520
list?

1049
00:52:38,520 --> 00:52:41,730
Look at who's picking that
work on the cover list.

1050
00:52:41,730 --> 00:52:44,980
Is it very senior
operators or not?

1051
00:52:44,980 --> 00:52:46,997
Any questions on this?

1052
00:52:46,997 --> 00:52:49,330
AUDIENCE: So cover lists--
are those people who come in,

1053
00:52:49,330 --> 00:52:50,670
and they're sort of waiting
to go on a [INAUDIBLE]??

1054
00:52:50,670 --> 00:52:51,100
NIGEL HM WILSON: Yeah.

1055
00:52:51,100 --> 00:52:51,196
Yeah.

1056
00:52:51,196 --> 00:52:52,240
AUDIENCE: They're not people
who are sitting at home,

1057
00:52:52,240 --> 00:52:52,780
waiting for a phone call.

1058
00:52:52,780 --> 00:52:53,947
NIGEL HM WILSON: No, no, no.

1059
00:52:53,947 --> 00:52:55,713
They're in the garage--

1060
00:52:55,713 --> 00:52:57,130
maybe playing cards
in the garage.

1061
00:52:57,130 --> 00:52:58,540
But they're in the garage,
and they're available.

1062
00:52:58,540 --> 00:52:59,037
AUDIENCE: OK.

1063
00:52:59,037 --> 00:52:59,870
NIGEL HM WILSON: OK.

1064
00:53:03,020 --> 00:53:06,020
OK, so I'm not going to
go through the model.

1065
00:53:06,020 --> 00:53:10,670
Again, I'm happy to give you
the papers and theses on this.

1066
00:53:10,670 --> 00:53:13,453
One of these theses was done
by Haris Koutsopoulos, who

1067
00:53:13,453 --> 00:53:14,620
many of you know quite well.

1068
00:53:14,620 --> 00:53:17,450
He's a senior faculty
member at Northeastern.

1069
00:53:17,450 --> 00:53:21,560
Another of the theses was
done by Yoram Shiftan,

1070
00:53:21,560 --> 00:53:26,137
who was entertaining me in
Israel just three days ago.

1071
00:53:26,137 --> 00:53:28,470
So there's some very good
work that's been done on this.

1072
00:53:28,470 --> 00:53:34,838
But the model that
came out of this--

1073
00:53:34,838 --> 00:53:36,380
I'm going to go show
you some results

1074
00:53:36,380 --> 00:53:40,840
and talk about what they mean.

1075
00:53:40,840 --> 00:53:45,913
What this is showing is,
under different scenarios,

1076
00:53:45,913 --> 00:53:47,330
the number of
full-time operators,

1077
00:53:47,330 --> 00:53:51,950
the number of part-time
operators that produce

1078
00:53:51,950 --> 00:53:57,230
the optimal solution,
the overtime as a percent

1079
00:53:57,230 --> 00:54:01,810
of the full cost--

1080
00:54:01,810 --> 00:54:06,070
the full cost, $96.4 million
in this case, per year.

1081
00:54:09,450 --> 00:54:12,932
Sorry, that's the regular
cost, regular time cost.

1082
00:54:12,932 --> 00:54:13,890
It's the overtime cost.

1083
00:54:13,890 --> 00:54:15,810
You add them up, and
you get the total cost--

1084
00:54:15,810 --> 00:54:17,370
$97.8 here.

1085
00:54:17,370 --> 00:54:21,230
And this is adding
different constraints.

1086
00:54:21,230 --> 00:54:25,720
So the base case, without any
constraints beyond vacation

1087
00:54:25,720 --> 00:54:29,720
or liability-- the things that
we've already talked about--

1088
00:54:29,720 --> 00:54:42,430
are that the cost of the
MBTA bus drivers should be--

1089
00:54:42,430 --> 00:54:48,710
these are 1996 dollars
again-- but $98 million.

1090
00:54:48,710 --> 00:54:52,220
And the reliability,
given 1.5% of the time

1091
00:54:52,220 --> 00:54:53,870
has to be worked on
overtime, and given

1092
00:54:53,870 --> 00:54:56,360
the way they manage their
overtime, which is not

1093
00:54:56,360 --> 00:54:59,540
very good, not
very well, they'll

1094
00:54:59,540 --> 00:55:03,320
be dropping about
0.4% of the trips--

1095
00:55:03,320 --> 00:55:06,770
4 trips out of every 1,000.

1096
00:55:06,770 --> 00:55:09,450
That was their existing
level of unreliability

1097
00:55:09,450 --> 00:55:12,920
in terms of dropping service.

1098
00:55:12,920 --> 00:55:14,720
So that's the base case numbers.

1099
00:55:14,720 --> 00:55:17,340
You could then add
different constraints.

1100
00:55:17,340 --> 00:55:20,570
If you add a constant
hiring constraint, which

1101
00:55:20,570 --> 00:55:25,840
means that we hire the same
number of operators every month

1102
00:55:25,840 --> 00:55:27,460
over the course
of the year and we

1103
00:55:27,460 --> 00:55:30,080
don't allow it to vary
from month to month,

1104
00:55:30,080 --> 00:55:31,640
what does that do?

1105
00:55:31,640 --> 00:55:34,480
It makes it easier for
the human resources folks.

1106
00:55:34,480 --> 00:55:36,540
It makes it easier for
the training people,

1107
00:55:36,540 --> 00:55:38,290
because they know what
their load is going

1108
00:55:38,290 --> 00:55:40,460
to be throughout the year.

1109
00:55:40,460 --> 00:55:43,860
And in fact, you can incorporate
that single constraint

1110
00:55:43,860 --> 00:55:48,790
with zero impact on the
budget or reliability.

1111
00:55:48,790 --> 00:55:50,670
So the solution is
exactly the same.

1112
00:55:50,670 --> 00:55:52,680
We've just juggled
vacation liability

1113
00:55:52,680 --> 00:55:54,760
between different periods.

1114
00:55:54,760 --> 00:55:57,410
OK?

1115
00:55:57,410 --> 00:56:00,980
If we wanted to have a
constant vacation as a policy

1116
00:56:00,980 --> 00:56:03,560
in the organization, we
add that as a constraint--

1117
00:56:03,560 --> 00:56:06,960
so the same month vacation
throughout the year--

1118
00:56:06,960 --> 00:56:11,570
then that would result in
about an almost $2 million

1119
00:56:11,570 --> 00:56:18,020
increase in the total cost
of the MBTA bus drivers.

1120
00:56:18,020 --> 00:56:20,240
We'd have less overtime.

1121
00:56:20,240 --> 00:56:22,850
And so our reliability
would improve

1122
00:56:22,850 --> 00:56:26,330
to dropping 2 trips
out of every 1,000,

1123
00:56:26,330 --> 00:56:28,700
rather than 4 trips
out of every 1,000.

1124
00:56:28,700 --> 00:56:31,460
And if we included in them
both the constant hiring

1125
00:56:31,460 --> 00:56:33,715
and vacation, which I
don't think you'd want to,

1126
00:56:33,715 --> 00:56:35,090
but this is just
to show what you

1127
00:56:35,090 --> 00:56:37,280
can do with this
sort of model, then

1128
00:56:37,280 --> 00:56:41,180
that would increase the budget
by an additional $1.5 million

1129
00:56:41,180 --> 00:56:42,580
a year.

1130
00:56:42,580 --> 00:56:46,670
And because you'd be
relying less on overtime,

1131
00:56:46,670 --> 00:56:49,140
you'd be having even
more reliable service.

1132
00:56:49,140 --> 00:56:53,780
So it gives the MBTA managers
a sense of the trade offs here,

1133
00:56:53,780 --> 00:56:57,440
but between the size of the
workforce optimally structured

1134
00:56:57,440 --> 00:57:01,184
over the course of the year
and reliability and overtime.

1135
00:57:04,170 --> 00:57:06,900
One different set
of results here.

1136
00:57:06,900 --> 00:57:09,420
This is showing,
OK, so what happens

1137
00:57:09,420 --> 00:57:13,780
if we change the level of
the overtime constraint?

1138
00:57:13,780 --> 00:57:21,450
So the base case here is
1.5% time as a constraint.

1139
00:57:21,450 --> 00:57:24,870
If we just eliminated the
overtime constraint completely

1140
00:57:24,870 --> 00:57:29,630
and said, just choose
the amount of overtime

1141
00:57:29,630 --> 00:57:32,990
that gives you the
minimal cost solution,

1142
00:57:32,990 --> 00:57:38,860
then your overtime costs would
be close to $12 million a year.

1143
00:57:38,860 --> 00:57:41,590
And that would enable you
to save about $1.3 million

1144
00:57:41,590 --> 00:57:44,050
per year on the
overall budget, because

1145
00:57:44,050 --> 00:57:49,610
of the relative costs of
overtime versus regular work.

1146
00:57:49,610 --> 00:57:54,110
However, your reliability would
go to hell in a hand basket.

1147
00:57:54,110 --> 00:57:56,405
And you'd be dropping 3
trips out of every 100--

1148
00:57:56,405 --> 00:57:58,550
3 hours of service
out of every 100.

1149
00:57:58,550 --> 00:58:00,230
And probably, this
is much better

1150
00:58:00,230 --> 00:58:02,540
than it would really
be if you did that.

1151
00:58:02,540 --> 00:58:05,630
Because this is not a
linear function, OK.

1152
00:58:05,630 --> 00:58:07,380
You're going to have
more and more trouble

1153
00:58:07,380 --> 00:58:09,005
covering those trips
and getting people

1154
00:58:09,005 --> 00:58:11,820
to agree to work overtime if
there's so much overtime being

1155
00:58:11,820 --> 00:58:13,990
asked of the workforce.

1156
00:58:13,990 --> 00:58:16,592
So this is a gross
underestimate.

1157
00:58:16,592 --> 00:58:21,370
But if you looked at 5%,
which is still pretty high,

1158
00:58:21,370 --> 00:58:26,700
then you'd save, in theory,
about $800,000 a year--

1159
00:58:26,700 --> 00:58:28,470
less than $1 million a year.

1160
00:58:28,470 --> 00:58:33,720
And your reliability would
decrease significantly

1161
00:58:33,720 --> 00:58:36,900
from dropping 4 trips
out of every 1,000

1162
00:58:36,900 --> 00:58:43,410
to dropping 120 trips
out of every 1,000--

1163
00:58:43,410 --> 00:58:45,980
no, 12-- from 4 to 12.

1164
00:58:49,730 --> 00:58:53,480
And then if you tighten
the overtime constraint

1165
00:58:53,480 --> 00:58:57,420
for whatever reason, then this
is what it would cost you.

1166
00:58:57,420 --> 00:59:00,870
It would cost you about half
a million dollars a year.

1167
00:59:00,870 --> 00:59:04,460
And again, you get some
improvement in reliability.

1168
00:59:04,460 --> 00:59:08,210
So these are the sort of
uses of the model that you

1169
00:59:08,210 --> 00:59:12,560
can use to inform management
decisions about how you

1170
00:59:12,560 --> 00:59:15,010
structure your hiring
process and training process

1171
00:59:15,010 --> 00:59:16,137
and extraboard size.

1172
00:59:19,150 --> 00:59:19,650
OK.

1173
00:59:19,650 --> 00:59:22,200
That's all I want to say
on the strategic level.

1174
00:59:22,200 --> 00:59:24,420
Any questions?

1175
00:59:24,420 --> 00:59:25,626
Yes.

1176
00:59:25,626 --> 00:59:28,494
AUDIENCE: Do you have a
sense of how difficult it is

1177
00:59:28,494 --> 00:59:34,150
or how costly it is to hire
additional, new drivers?

1178
00:59:34,150 --> 00:59:35,140
NIGEL HM WILSON: Yes.

1179
00:59:35,140 --> 00:59:36,400
Yes.

1180
00:59:36,400 --> 00:59:38,890
So there is a training period.

1181
00:59:38,890 --> 00:59:41,320
The MBTA, since
Fred was Secretary,

1182
00:59:41,320 --> 00:59:46,690
has had a lottery process for
selection of new employees.

1183
00:59:46,690 --> 00:59:48,910
And every agency,
including the MBTA,

1184
00:59:48,910 --> 00:59:50,650
has a pretty rigorous
testing procedure

1185
00:59:50,650 --> 00:59:53,920
you have to go through to get
to the point of being hired.

1186
00:59:53,920 --> 00:59:56,020
You then go through training,
the criminal records

1187
00:59:56,020 --> 00:59:59,530
check, all sorts of checks
before the training begins.

1188
00:59:59,530 --> 01:00:02,045
And then the training--

1189
01:00:02,045 --> 01:00:05,410
I'm trying to remember, I don't
have it on top of a head--

1190
01:00:05,410 --> 01:00:10,210
I think that the
total training is

1191
01:00:10,210 --> 01:00:12,610
in the order of six
weeks, which includes

1192
01:00:12,610 --> 01:00:14,410
getting your commercial
driver's licenses,

1193
01:00:14,410 --> 01:00:16,960
as we talked about earlier.

1194
01:00:16,960 --> 01:00:21,400
So you've got a certain amount
of advance planning in here.

1195
01:00:21,400 --> 01:00:24,890
You can't decide today, I
want to have more people

1196
01:00:24,890 --> 01:00:26,442
on the driver
workforce tomorrow.

1197
01:00:26,442 --> 01:00:27,400
It's not going to work.

1198
01:00:27,400 --> 01:00:30,280
So that's why you need to
have a strategic view of this.

1199
01:00:30,280 --> 01:00:31,115
Yes?

1200
01:00:31,115 --> 01:00:35,075
AUDIENCE: Does this model
capture the behavioral feedback

1201
01:00:35,075 --> 01:00:35,882
loop?

1202
01:00:35,882 --> 01:00:37,090
NIGEL HM WILSON: It does not.

1203
01:00:37,090 --> 01:00:38,440
So that was separate work.

1204
01:00:38,440 --> 01:00:41,040
And we did look at that.

1205
01:00:41,040 --> 01:00:43,560
But it's not in the
results I presented.

1206
01:00:43,560 --> 01:00:45,490
Though, as I said,
Yoram Shiftan looked

1207
01:00:45,490 --> 01:00:49,630
at feedback effect between
if you provide more overtime,

1208
01:00:49,630 --> 01:00:52,170
what does that do to absence?

1209
01:00:52,170 --> 01:00:55,200
And what he found
was, it's variable.

1210
01:00:55,200 --> 01:00:58,170
For some operators, there
is a feedback effect.

1211
01:00:58,170 --> 01:01:02,650
But it's not as strong
as you might expect, OK.

1212
01:01:02,650 --> 01:01:04,420
So the MBTA management,
for instance,

1213
01:01:04,420 --> 01:01:09,880
was pretty convinced this
was really a serious issue.

1214
01:01:09,880 --> 01:01:11,450
It wasn't, in our analysis.

1215
01:01:11,450 --> 01:01:14,110
But there is a feedback effect.

1216
01:01:14,110 --> 01:01:17,330
It's not reflected in here.

1217
01:01:17,330 --> 01:01:18,810
Yeah, Scott.

1218
01:01:18,810 --> 01:01:20,310
AUDIENCE: Earlier,
you said you look

1219
01:01:20,310 --> 01:01:24,938
at the seniority of the
extraboards that are taking

1220
01:01:24,938 --> 01:01:25,980
work from the cover list.

1221
01:01:25,980 --> 01:01:26,370
NIGEL HM WILSON: Yes.

1222
01:01:26,370 --> 01:01:27,950
AUDIENCE: And you said
it's approximately

1223
01:01:27,950 --> 01:01:29,410
more than the work
that's desirable, right?

1224
01:01:29,410 --> 01:01:29,810
NIGEL HM WILSON: Yes.

1225
01:01:29,810 --> 01:01:30,810
AUDIENCE: The more senior
person takes the work

1226
01:01:30,810 --> 01:01:31,685
from the [INAUDIBLE]?

1227
01:01:31,685 --> 01:01:32,810
NIGEL HM WILSON: Yes, yes.

1228
01:01:32,810 --> 01:01:35,310
AUDIENCE: And I didn't quite
get the implication there

1229
01:01:35,310 --> 01:01:38,680
for the amount of
work, [INAUDIBLE]..

1230
01:01:38,680 --> 01:01:40,400
NIGEL HM WILSON: OK.

1231
01:01:40,400 --> 01:01:40,900
So yeah.

1232
01:01:40,900 --> 01:01:44,020
So if the most
senior operators--

1233
01:01:44,020 --> 01:01:46,575
who can pick anything--
are picking the extraboard,

1234
01:01:46,575 --> 01:01:47,950
that means you
don't have to work

1235
01:01:47,950 --> 01:01:49,300
very hard in the extraboard.

1236
01:01:49,300 --> 01:01:51,880
The extraboard is
probably too large.

1237
01:01:51,880 --> 01:01:52,380
OK.

1238
01:01:52,380 --> 01:01:54,570
That's the simple proxy.

1239
01:01:54,570 --> 01:01:57,810
Similarly, if the
most senior operators

1240
01:01:57,810 --> 01:02:02,520
are picking certain duties
in a garage in terms

1241
01:02:02,520 --> 01:02:06,060
of the types of shifts
and the days they work,

1242
01:02:06,060 --> 01:02:10,180
that's a good proxy for what's
easy, what's not so easy.

1243
01:02:10,180 --> 01:02:12,860
That's not to say
senior operators aren't

1244
01:02:12,860 --> 01:02:13,650
highly motivated.

1245
01:02:13,650 --> 01:02:14,150
They are.

1246
01:02:14,150 --> 01:02:16,360
But they also have their
own self-interest at heart.

1247
01:02:16,360 --> 01:02:18,840
They don't necessarily want to
be on a dangerous, difficult

1248
01:02:18,840 --> 01:02:20,448
route to operate.

1249
01:02:20,448 --> 01:02:22,490
Now, the problem here,
which Fred will get into I

1250
01:02:22,490 --> 01:02:26,142
think next week, in terms
of the way the US transit

1251
01:02:26,142 --> 01:02:28,100
industry operates-- there
are lots of problems.

1252
01:02:28,100 --> 01:02:31,040
But one of the problems in
terms of the way the US transit

1253
01:02:31,040 --> 01:02:35,010
industry operates, compared
with European industry,

1254
01:02:35,010 --> 01:02:39,280
is they rely very
heavily on seniority.

1255
01:02:39,280 --> 01:02:41,890
So what does that mean?

1256
01:02:41,890 --> 01:02:45,580
It means the least
capable drivers, who

1257
01:02:45,580 --> 01:02:47,330
are the people who
are new in the job

1258
01:02:47,330 --> 01:02:49,500
and haven't got a
lot of experience,

1259
01:02:49,500 --> 01:02:52,630
have the most
difficult assignments.

1260
01:02:52,630 --> 01:02:56,550
That's almost
certainly not optimal.

1261
01:02:56,550 --> 01:03:01,430
So it means you have
very poor drivers,

1262
01:03:01,430 --> 01:03:03,110
in terms of not
their capability,

1263
01:03:03,110 --> 01:03:05,630
but where they are on
the learning curve,

1264
01:03:05,630 --> 01:03:08,860
driving very difficult routes.

1265
01:03:08,860 --> 01:03:12,550
And the European
approach to this

1266
01:03:12,550 --> 01:03:16,130
is typically to have
a rotating roster, OK.

1267
01:03:18,690 --> 01:03:23,720
And again, in the
US transit industry,

1268
01:03:23,720 --> 01:03:26,600
one of the main benefits
of more seniority

1269
01:03:26,600 --> 01:03:31,470
for a particular individual
is you get to pick your work,

1270
01:03:31,470 --> 01:03:35,010
and you get to pick
your vacation time.

1271
01:03:35,010 --> 01:03:36,897
So if you get to 10
years at the MBTA,

1272
01:03:36,897 --> 01:03:38,980
perhaps you never have to
work weekends unless you

1273
01:03:38,980 --> 01:03:42,180
particularly want it,
which you often won't.

1274
01:03:42,180 --> 01:03:48,150
Whereas in most
European systems,

1275
01:03:48,150 --> 01:03:50,670
you're assigned to a
roster that rotates.

1276
01:03:50,670 --> 01:03:53,610
So you will all go through
the same experience

1277
01:03:53,610 --> 01:03:57,100
over the course of two months.

1278
01:03:57,100 --> 01:03:59,023
I don't know how
JR-East does it, Kenji.

1279
01:03:59,023 --> 01:04:00,190
AUDIENCE: Similar to Europe.

1280
01:04:00,190 --> 01:04:01,648
NIGEL HM WILSON:
Similar to Europe.

1281
01:04:01,648 --> 01:04:04,405
That's really the
dominant model elsewhere

1282
01:04:04,405 --> 01:04:05,950
that I'm familiar with.

1283
01:04:05,950 --> 01:04:09,670
There are trade offs in this.

1284
01:04:09,670 --> 01:04:13,930
So one of the trade
offs is that if you

1285
01:04:13,930 --> 01:04:17,560
use that as your main
motivation for keeping people

1286
01:04:17,560 --> 01:04:18,715
on the workforce--

1287
01:04:18,715 --> 01:04:20,590
you want to keep your
good operators working,

1288
01:04:20,590 --> 01:04:21,923
you don't have to replace them--

1289
01:04:24,530 --> 01:04:29,860
you can either pay a
wage rate incentive

1290
01:04:29,860 --> 01:04:35,960
for longevity, which is more
common in other agencies

1291
01:04:35,960 --> 01:04:37,110
around the world.

1292
01:04:37,110 --> 01:04:39,070
But it's not very
common in the US.

1293
01:04:39,070 --> 01:04:45,680
And so you progress from the
entry level wage in the MBTA

1294
01:04:45,680 --> 01:04:49,690
to the top level
in about two years.

1295
01:04:49,690 --> 01:04:54,520
OK, and that's not unusual
in the US transit industry.

1296
01:04:54,520 --> 01:04:57,670
And so because that is
such a minimal incentive

1297
01:04:57,670 --> 01:05:01,270
to hang around for
20 years, you want

1298
01:05:01,270 --> 01:05:02,800
to provide other incentives.

1299
01:05:02,800 --> 01:05:07,090
And so seniority, picking your
work, picking your vacation,

1300
01:05:07,090 --> 01:05:09,460
is the way the US transit
industry has followed this.

1301
01:05:09,460 --> 01:05:11,348
Gabriel, you were going
to make a comment.

1302
01:05:11,348 --> 01:05:13,890
GABRIEL SANCHEZ-MARTINEZ: It's
a half-comment, half-question.

1303
01:05:13,890 --> 01:05:14,100
NIGEL HM WILSON: Yeah.

1304
01:05:14,100 --> 01:05:15,517
GABRIEL SANCHEZ-MARTINEZ:
But when

1305
01:05:15,517 --> 01:05:17,362
we discussed crew
scheduling, we briefly

1306
01:05:17,362 --> 01:05:19,603
touched on rostering
and differences

1307
01:05:19,603 --> 01:05:21,520
between North American
and European practices.

1308
01:05:21,520 --> 01:05:21,860
NIGEL HM WILSON: Yes.

1309
01:05:21,860 --> 01:05:23,860
GABRIEL SANCHEZ-MARTINEZ:
And well, part of that

1310
01:05:23,860 --> 01:05:29,040
was that in European practice,
rostering is done sometimes

1311
01:05:29,040 --> 01:05:31,930
with some help of
alarm or helping

1312
01:05:31,930 --> 01:05:35,283
them to make the different
rosters as even as possible.

1313
01:05:35,283 --> 01:05:36,241
NIGEL HM WILSON: Right.

1314
01:05:36,241 --> 01:05:37,658
GABRIEL SANCHEZ-MARTINEZ:
And that

1315
01:05:37,658 --> 01:05:40,295
is slightly different
from picking them.

1316
01:05:40,295 --> 01:05:41,170
NIGEL HM WILSON: Yes.

1317
01:05:41,170 --> 01:05:42,390
GABRIEL SANCHEZ-MARTINEZ:
Right, so in North American,

1318
01:05:42,390 --> 01:05:44,160
these rosters are
formed by people,

1319
01:05:44,160 --> 01:05:48,230
but then they assemble the
rosters from crew schedules.

1320
01:05:48,230 --> 01:05:49,293
Is that correct?

1321
01:05:49,293 --> 01:05:50,710
NIGEL HM WILSON:
Well, the roster,

1322
01:05:50,710 --> 01:05:56,110
as I'm describing the roster,
is the rotation, if any,

1323
01:05:56,110 --> 01:05:58,510
of the work that's
done among individuals

1324
01:05:58,510 --> 01:06:01,390
over the course of a period,
maybe a timetable or two

1325
01:06:01,390 --> 01:06:03,070
months.

1326
01:06:03,070 --> 01:06:05,110
If you have a
rotating roster, that

1327
01:06:05,110 --> 01:06:06,490
means everyone in
this group goes

1328
01:06:06,490 --> 01:06:09,490
through the same set of
driving responsibilities

1329
01:06:09,490 --> 01:06:12,100
over the course of a two-month
period or a timetable.

1330
01:06:14,820 --> 01:06:18,910
In the US transit industry case,
you don't have any rotation.

1331
01:06:18,910 --> 01:06:20,590
You don't even have a roster.

1332
01:06:20,590 --> 01:06:25,890
So you're there, you do
the same trip every day

1333
01:06:25,890 --> 01:06:27,330
if you show up for work.

1334
01:06:27,330 --> 01:06:31,440
If you don't show up for
work, you don't do the duty.

1335
01:06:31,440 --> 01:06:37,740
So it's a trade off.

1336
01:06:37,740 --> 01:06:40,860
Personally-- maybe it's
my European roots--

1337
01:06:40,860 --> 01:06:44,200
I kind of tend to prefer a
European approach to this

1338
01:06:44,200 --> 01:06:46,960
than the US approach to this.

1339
01:06:46,960 --> 01:06:47,460
OK.

1340
01:06:47,460 --> 01:06:51,280
Let me quickly jump
into the tactical level.

1341
01:06:51,280 --> 01:06:54,540
So here, we're going
into the summer timetable

1342
01:06:54,540 --> 01:06:57,890
the MBTA in about
a month's time.

1343
01:06:57,890 --> 01:06:59,700
And so we'll be deciding.

1344
01:06:59,700 --> 01:07:04,700
We know how many drivers
we have on the property.

1345
01:07:04,700 --> 01:07:07,850
What we're deciding
now is when to assign

1346
01:07:07,850 --> 01:07:11,060
the work on a daily basis
and how many operators

1347
01:07:11,060 --> 01:07:17,270
to provide to Lynn versus Quincy
versus Arborway versus Cabot--

1348
01:07:17,270 --> 01:07:20,210
the different depots
in the system.

1349
01:07:20,210 --> 01:07:23,380
And again, so we're
allocating the extra staff,

1350
01:07:23,380 --> 01:07:29,320
these floaters, by garage
and by day of week.

1351
01:07:29,320 --> 01:07:31,510
And so we want to
know the absence level

1352
01:07:31,510 --> 01:07:36,240
on a more detailed basis here.

1353
01:07:36,240 --> 01:07:39,840
We have basically the same
sort of issues and concerns,

1354
01:07:39,840 --> 01:07:43,770
just at a more detailed
level of analysis.

1355
01:07:43,770 --> 01:07:46,290
And we can either use a
heuristic or optimization

1356
01:07:46,290 --> 01:07:47,160
for this.

1357
01:07:47,160 --> 01:07:49,470
And heuristics actually give
you a very good solution

1358
01:07:49,470 --> 01:07:50,970
because it's much simpler.

1359
01:07:50,970 --> 01:07:54,490
This is real data from the
MBTA, again, from 20 years ago.

1360
01:07:54,490 --> 01:07:57,600
So I'm not saying it's
exactly the same today.

1361
01:07:57,600 --> 01:08:03,760
But I would bet it's not
that different from now

1362
01:08:03,760 --> 01:08:05,270
from what it was then.

1363
01:08:05,270 --> 01:08:09,470
So this is a single MBTA garage.

1364
01:08:09,470 --> 01:08:12,770
This is the Cabot garage, which
is one of the biggest MBTA bus

1365
01:08:12,770 --> 01:08:14,540
garages.

1366
01:08:14,540 --> 01:08:18,890
And for that garage, this
shows the mean number

1367
01:08:18,890 --> 01:08:24,200
of open hours of work on a
daily basis over a timetable.

1368
01:08:24,200 --> 01:08:28,250
So on Mondays, 259 hours
was the mean number

1369
01:08:28,250 --> 01:08:30,620
of hours you had to
cover because you

1370
01:08:30,620 --> 01:08:32,450
didn't have someone available.

1371
01:08:32,450 --> 01:08:34,979
Standard deviation
of that was 36.

1372
01:08:34,979 --> 01:08:38,240
So that's quite variable.

1373
01:08:38,240 --> 01:08:42,109
Tuesday and Wednesday is
lower in terms of absences--

1374
01:08:42,109 --> 01:08:44,210
also quite variable.

1375
01:08:44,210 --> 01:08:47,687
Thursday and particularly
Friday is much higher and also

1376
01:08:47,687 --> 01:08:48,479
even more variable.

1377
01:08:53,310 --> 01:08:57,000
So following our
analysis principles,

1378
01:08:57,000 --> 01:08:59,850
the cumulative density
function, then you

1379
01:08:59,850 --> 01:09:04,630
would expect to want to
allocate your workforce,

1380
01:09:04,630 --> 01:09:10,180
your spare workforce,
reflecting the level of absence.

1381
01:09:10,180 --> 01:09:10,680
But no.

1382
01:09:13,920 --> 01:09:19,160
The MBTA was allocating their
extraboard exactly equally

1383
01:09:19,160 --> 01:09:21,380
across all weekdays--

1384
01:09:21,380 --> 01:09:25,350
20 extraboard operators on
Monday, Tuesday, Wednesday,

1385
01:09:25,350 --> 01:09:26,800
Thursday, Friday, Saturday--

1386
01:09:26,800 --> 01:09:29,270
Wednesday, Thursday,
Friday, lower levels

1387
01:09:29,270 --> 01:09:33,290
on Saturday and Sunday, but the
same percentage, independent

1388
01:09:33,290 --> 01:09:36,319
of the observed level
of difference in absence

1389
01:09:36,319 --> 01:09:40,060
or extra work required
on those days.

1390
01:09:40,060 --> 01:09:45,550
The model showed that if
you optimally reallocated

1391
01:09:45,550 --> 01:09:50,090
the same level of resources,
basically the same level down

1392
01:09:50,090 --> 01:09:52,729
here in terms of
operator days, you

1393
01:09:52,729 --> 01:09:58,430
provide more cover on
Fridays, on Mondays,

1394
01:09:58,430 --> 01:10:03,570
and weekends, and less coverage
on Tuesdays, Wednesdays,

1395
01:10:03,570 --> 01:10:06,720
and Thursdays, reflecting
the absence patterns.

1396
01:10:06,720 --> 01:10:09,521
That's optimizing the daily
level for a particular garage.

1397
01:10:13,580 --> 01:10:17,900
And at that point, they
weren't using part-timers.

1398
01:10:17,900 --> 01:10:22,490
They were using part-timers,
but the part-timers

1399
01:10:22,490 --> 01:10:26,028
were not on the extraboard
and not being used in quite

1400
01:10:26,028 --> 01:10:27,320
the same way that they are now.

1401
01:10:30,110 --> 01:10:33,230
And what you get out of this
is the expected overtime

1402
01:10:33,230 --> 01:10:37,190
hours over the whole
timetable, recognizing

1403
01:10:37,190 --> 01:10:40,940
the variability on
a particular day

1404
01:10:40,940 --> 01:10:43,340
is the expected
level of overtime

1405
01:10:43,340 --> 01:10:53,190
hours over the whole week is
140 hours versus 99 hours.

1406
01:10:53,190 --> 01:10:58,760
So you're saving about 25% to
30% of your overtime costs,

1407
01:10:58,760 --> 01:11:01,910
just by reallocating your
existing workforce more

1408
01:11:01,910 --> 01:11:05,500
effectively across
days of the week.

1409
01:11:05,500 --> 01:11:07,130
This is not rocket science.

1410
01:11:07,130 --> 01:11:09,430
This is actually
pretty easy to do.

1411
01:11:09,430 --> 01:11:11,990
I mean, you can get these
absence figures quite easily

1412
01:11:11,990 --> 01:11:15,800
and implement them.

1413
01:11:15,800 --> 01:11:19,780
So the tactical level findings
are significant variation,

1414
01:11:19,780 --> 01:11:26,140
absence, and required extra work
by garage and by day of week.

1415
01:11:26,140 --> 01:11:31,060
So what would you
expect by garage,

1416
01:11:31,060 --> 01:11:32,920
in terms of absence rates?

1417
01:11:32,920 --> 01:11:35,080
You might expect--
and you'd be right--

1418
01:11:35,080 --> 01:11:37,750
that the most difficult garages
would have the highest absence

1419
01:11:37,750 --> 01:11:38,950
rates.

1420
01:11:38,950 --> 01:11:40,670
Those are tough conditions.

1421
01:11:40,670 --> 01:11:44,110
So if you look at Cabot,
those are very tough routes.

1422
01:11:44,110 --> 01:11:47,380
And so when you do a
system-wide pick that

1423
01:11:47,380 --> 01:11:52,266
allows the operators to select
the garage they work at,

1424
01:11:52,266 --> 01:11:56,610
so the most senior
operators will avoid Cabot.

1425
01:11:56,610 --> 01:11:58,570
So it's the most
junior operators

1426
01:11:58,570 --> 01:12:02,650
that wind up at Cabot on
the most difficult routes.

1427
01:12:02,650 --> 01:12:06,970
Not a great surprise your
absence rate is higher at Cabot

1428
01:12:06,970 --> 01:12:08,710
than it is at Lynn or Quincy--

1429
01:12:08,710 --> 01:12:10,360
nice, comfortable
suburban garages.

1430
01:12:14,340 --> 01:12:15,342
Yeah.

1431
01:12:15,342 --> 01:12:18,037
AUDIENCE: How do you
define a difficult route?

1432
01:12:18,037 --> 01:12:19,250
What constitutes [INAUDIBLE]?

1433
01:12:19,250 --> 01:12:20,250
NIGEL HM WILSON: Oh, OK.

1434
01:12:20,250 --> 01:12:24,750
So there can be any
number of things--

1435
01:12:24,750 --> 01:12:29,390
very heavy traffic, danger of
accidents, danger of assault,

1436
01:12:29,390 --> 01:12:34,650
crime, danger of
altercations among customers,

1437
01:12:34,650 --> 01:12:37,810
passengers on the bus,
altercations with the driver.

1438
01:12:37,810 --> 01:12:42,300
So a number of things like that.

1439
01:12:42,300 --> 01:12:48,480
And so I'm not saying all
the routes in the Lynn garage

1440
01:12:48,480 --> 01:12:49,950
are great routes.

1441
01:12:49,950 --> 01:12:54,420
But compared with operating
in the inner city, which

1442
01:12:54,420 --> 01:12:56,370
is where the Cabot
routes typically serve,

1443
01:12:56,370 --> 01:12:57,468
there is a difference.

1444
01:12:57,468 --> 01:12:59,760
And certainly in Quincy,
there will be a big difference

1445
01:12:59,760 --> 01:13:00,090
as well.

1446
01:13:00,090 --> 01:13:00,590
Yes?

1447
01:13:03,070 --> 01:13:06,510
AUDIENCE: So the workforce
would have differing levels

1448
01:13:06,510 --> 01:13:09,860
of information about which
route or which garage would

1449
01:13:09,860 --> 01:13:11,163
be more difficult [INAUDIBLE]?

1450
01:13:11,163 --> 01:13:12,580
NIGEL HM WILSON:
By word of mouth.

1451
01:13:12,580 --> 01:13:16,320
It's not a sort of a
conscious campaign.

1452
01:13:16,320 --> 01:13:20,580
The idea is basically,
you say, hey,

1453
01:13:20,580 --> 01:13:22,210
we've got a seniority system.

1454
01:13:22,210 --> 01:13:25,860
So you'll get to know other
operators of different depots,

1455
01:13:25,860 --> 01:13:26,910
word of mouth.

1456
01:13:26,910 --> 01:13:29,520
And you'll select, based
on what's available

1457
01:13:29,520 --> 01:13:32,520
when it gets to you of your
place in your seniority list,

1458
01:13:32,520 --> 01:13:34,170
where you want to work.

1459
01:13:34,170 --> 01:13:36,330
And often, it's based
on where you live.

1460
01:13:36,330 --> 01:13:41,185
You don't want to have to
commute from Lynn to Quincy

1461
01:13:41,185 --> 01:13:43,270
every day to do your work.

1462
01:13:43,270 --> 01:13:46,380
So it's a question of
proximity to where you live

1463
01:13:46,380 --> 01:13:52,380
and the ease, the comfort
level you'll have on the routes

1464
01:13:52,380 --> 01:13:55,440
and the conditions that
you'll be operating on.

1465
01:13:55,440 --> 01:13:56,360
OK?

1466
01:13:56,360 --> 01:13:59,300
And the policy of the
MBTA at this point

1467
01:13:59,300 --> 01:14:04,200
was to provide the same level
of cover at every garage.

1468
01:14:04,200 --> 01:14:04,700
Why?

1469
01:14:07,960 --> 01:14:09,850
Their argument that
they articulated to me--

1470
01:14:09,850 --> 01:14:14,095
and obviously, I discussed
this with them extensively--

1471
01:14:14,095 --> 01:14:17,110
they would not dispute the fact
that Cabot has a higher level

1472
01:14:17,110 --> 01:14:19,522
of absence than Quincy or Lynn.

1473
01:14:19,522 --> 01:14:20,230
They accept that.

1474
01:14:20,230 --> 01:14:21,920
That's not in dispute.

1475
01:14:21,920 --> 01:14:23,500
But their argument
would be, you're

1476
01:14:23,500 --> 01:14:29,450
rewarding the manager
of Cabot by giving

1477
01:14:29,450 --> 01:14:32,540
them more manpower if you
recognize the existing absence

1478
01:14:32,540 --> 01:14:33,350
level.

1479
01:14:33,350 --> 01:14:36,700
So they're concerned
about this feedback loop.

1480
01:14:36,700 --> 01:14:40,350
If you have more absence,
you get more manpower.

1481
01:14:40,350 --> 01:14:42,450
To me, that's a
completely backwards way

1482
01:14:42,450 --> 01:14:43,500
of thinking about this.

1483
01:14:43,500 --> 01:14:46,080
You need to have
incentives for the managers

1484
01:14:46,080 --> 01:14:48,570
to do a good job, hopefully.

1485
01:14:48,570 --> 01:14:51,570
But you shouldn't let
that be the motivation

1486
01:14:51,570 --> 01:14:53,370
for making bad
decisions about how

1487
01:14:53,370 --> 01:14:55,440
you use your resources,
which is basically

1488
01:14:55,440 --> 01:14:58,460
what's happening here.

1489
01:14:58,460 --> 01:15:01,520
And they were arguing the same
thing in terms of day of week,

1490
01:15:01,520 --> 01:15:02,120
as well.

1491
01:15:02,120 --> 01:15:05,930
So they were really
fighting that, that concept.

1492
01:15:05,930 --> 01:15:11,540
And I wasn't successful
in persuading them

1493
01:15:11,540 --> 01:15:15,170
that they should really
separate out the management

1494
01:15:15,170 --> 01:15:18,480
incentives from the efficiency
of the decisions they're

1495
01:15:18,480 --> 01:15:18,980
making.

1496
01:15:21,980 --> 01:15:24,660
OK.

1497
01:15:24,660 --> 01:15:25,160
All right.

1498
01:15:25,160 --> 01:15:26,780
The final level--

1499
01:15:26,780 --> 01:15:28,850
I've got a few minutes left--

1500
01:15:28,850 --> 01:15:30,840
is the operational level.

1501
01:15:30,840 --> 01:15:34,520
So again, we've gone from
the one-year planning horizon

1502
01:15:34,520 --> 01:15:39,680
to the timetable planning
horizon for three or four

1503
01:15:39,680 --> 01:15:42,990
months, now down
to the daily level.

1504
01:15:42,990 --> 01:15:44,120
OK.

1505
01:15:44,120 --> 01:15:46,520
So the objective here is
pretty much the same thing.

1506
01:15:46,520 --> 01:15:48,782
We're trying to minimize
the weighted sum of overtime

1507
01:15:48,782 --> 01:15:49,490
and missed trips.

1508
01:15:54,590 --> 01:15:55,263
Oh, I'm sorry.

1509
01:15:55,263 --> 01:15:56,930
I'm still at the
operational level here.

1510
01:16:01,430 --> 01:16:04,830
Yeah, no this is right, sorry.

1511
01:16:04,830 --> 01:16:11,430
So what we want to do here
is decide the report times

1512
01:16:11,430 --> 01:16:17,090
for any unassigned, extra
staff for the next day.

1513
01:16:17,090 --> 01:16:20,110
So as I said, at
12 o'clock today,

1514
01:16:20,110 --> 01:16:22,450
if there were five
operators who had not

1515
01:16:22,450 --> 01:16:28,420
been assigned work for tomorrow
at Cabot, then the MBTA

1516
01:16:28,420 --> 01:16:29,530
manager--

1517
01:16:29,530 --> 01:16:34,460
desk superintendent or whoever
you want to call him or her--

1518
01:16:34,460 --> 01:16:37,600
at Cabot will be
posting the report times

1519
01:16:37,600 --> 01:16:39,640
for those five people tomorrow.

1520
01:16:39,640 --> 01:16:42,070
You don't know what the
absence is going to be.

1521
01:16:42,070 --> 01:16:43,570
You didn't have any
known in advance

1522
01:16:43,570 --> 01:16:46,030
absence to give
them work to cover.

1523
01:16:46,030 --> 01:16:47,950
But you want them
to be available

1524
01:16:47,950 --> 01:16:49,762
so that when you do
get the absences, which

1525
01:16:49,762 --> 01:16:51,970
you don't know about, they
will be in a good position

1526
01:16:51,970 --> 01:16:52,926
to cover them.

1527
01:16:59,140 --> 01:17:02,238
And so this gets into the
most detailed level of this.

1528
01:17:02,238 --> 01:17:03,780
So it's not just a
question of having

1529
01:17:03,780 --> 01:17:08,630
the right number of drivers
available on a particular day.

1530
01:17:08,630 --> 01:17:10,380
It's a question of
having the right number

1531
01:17:10,380 --> 01:17:14,960
of available drivers at
every time of that day.

1532
01:17:14,960 --> 01:17:16,860
That's the problem
we're dealing with here.

1533
01:17:16,860 --> 01:17:19,850
So again, much more
uncertainty about how

1534
01:17:19,850 --> 01:17:27,070
the absence is at 9 o'clock
AM versus 5 o'clock PM.

1535
01:17:27,070 --> 01:17:30,445
And what this is showing
is this glorious concept.

1536
01:17:30,445 --> 01:17:33,220
It used to be called
"slop overtime."

1537
01:17:33,220 --> 01:17:35,320
The MBTA didn't like the
term "slop overtime,"

1538
01:17:35,320 --> 01:17:37,810
so it's now "excess overtime."

1539
01:17:37,810 --> 01:17:40,540
So you have regular
overtime, which is just,

1540
01:17:40,540 --> 01:17:44,690
you didn't have enough
personnel on the payroll.

1541
01:17:44,690 --> 01:17:46,440
And so you had to buy
overtime if you were

1542
01:17:46,440 --> 01:17:48,220
going to deliver the service.

1543
01:17:48,220 --> 01:17:52,850
Excess overtime is where, in
theory you have enough people

1544
01:17:52,850 --> 01:17:55,220
to cover the needs, but
they're not available

1545
01:17:55,220 --> 01:17:59,070
when you want them to be,
because you couldn't make

1546
01:17:59,070 --> 01:18:02,370
perfect decisions about when
they should show up to make

1547
01:18:02,370 --> 01:18:05,000
sure they're productive.

1548
01:18:05,000 --> 01:18:08,240
So here, I'm showing,
here, this is

1549
01:18:08,240 --> 01:18:12,890
time of day from 5:00 AM to
1:00 AM the next morning.

1550
01:18:15,430 --> 01:18:21,790
This shows the open run
profile, OK, like that.

1551
01:18:24,610 --> 01:18:29,320
Sorry, this is the open
run profile like this.

1552
01:18:29,320 --> 01:18:32,260
And this is the available cover
profile, based on your report

1553
01:18:32,260 --> 01:18:36,160
time decisions, like this.

1554
01:18:36,160 --> 01:18:38,710
And so what this shows is
that you have these periods

1555
01:18:38,710 --> 01:18:43,900
here and here and
here, where you

1556
01:18:43,900 --> 01:18:50,290
had more operators available
than you had work to cover.

1557
01:18:50,290 --> 01:18:53,940
And so that means that there
were other periods of the day

1558
01:18:53,940 --> 01:18:56,790
here and here and
here, where you

1559
01:18:56,790 --> 01:18:58,140
didn't have enough operators.

1560
01:18:58,140 --> 01:19:00,270
You had the right
number on the whole day,

1561
01:19:00,270 --> 01:19:04,020
you just didn't have them at
the right time in the day.

1562
01:19:04,020 --> 01:19:06,240
And it's not just you
made bad decisions,

1563
01:19:06,240 --> 01:19:09,090
it's that this is stochastic.

1564
01:19:09,090 --> 01:19:10,568
You can't make
perfect decisions.

1565
01:19:10,568 --> 01:19:12,110
So you've got to
expect there's going

1566
01:19:12,110 --> 01:19:19,410
to be some excess
overtime built in here.

1567
01:19:19,410 --> 01:19:22,470
And this is the
excess overtime curve.

1568
01:19:22,470 --> 01:19:26,427
And I will wrap this up in
four minutes, I promise.

1569
01:19:26,427 --> 01:19:27,760
I may not go through everything.

1570
01:19:27,760 --> 01:19:31,630
But I'll cover most
of what we want.

1571
01:19:31,630 --> 01:19:34,650
So this, on the x-axis,
is the available cover

1572
01:19:34,650 --> 01:19:36,540
minus the required cover.

1573
01:19:36,540 --> 01:19:41,070
So again, if it's
a perfect balance,

1574
01:19:41,070 --> 01:19:42,990
excess cover minus
required cover

1575
01:19:42,990 --> 01:19:47,340
is 0 because you
had a perfect fit.

1576
01:19:47,340 --> 01:19:49,620
But what that gives you
is the greatest number

1577
01:19:49,620 --> 01:19:51,120
of excess overtime hours.

1578
01:19:51,120 --> 01:19:54,710
Because unless you made
every decision perfectly,

1579
01:19:54,710 --> 01:19:57,200
you're going to have
excess overtime hours.

1580
01:19:57,200 --> 01:20:00,980
If you go way out here,
you had far more people

1581
01:20:00,980 --> 01:20:02,930
available than you needed.

1582
01:20:02,930 --> 01:20:04,680
So you shouldn't have
any access overtime,

1583
01:20:04,680 --> 01:20:07,040
even if you make
lousy decisions.

1584
01:20:07,040 --> 01:20:10,380
You put them out any
time, they'll be covered.

1585
01:20:10,380 --> 01:20:12,300
And you put them
out-- if you have

1586
01:20:12,300 --> 01:20:15,030
far fewer people available
than you require,

1587
01:20:15,030 --> 01:20:16,050
you'll be out here.

1588
01:20:16,050 --> 01:20:18,900
And you shouldn't have any
access overtime either,

1589
01:20:18,900 --> 01:20:22,310
because you don't
have to be very

1590
01:20:22,310 --> 01:20:25,310
good to make decisions which
will have these people be fully

1591
01:20:25,310 --> 01:20:27,305
employed.

1592
01:20:27,305 --> 01:20:30,100
OK.

1593
01:20:30,100 --> 01:20:33,970
So it's when you have
this exact balance,

1594
01:20:33,970 --> 01:20:36,220
you have to make every
decision perfectly

1595
01:20:36,220 --> 01:20:37,617
to avoid excess overtime.

1596
01:20:37,617 --> 01:20:39,034
And you're never
going to do that.

1597
01:20:44,010 --> 01:20:47,530
OK, so I've got
some results here,

1598
01:20:47,530 --> 01:20:51,670
which you can go
through on your own.

1599
01:20:51,670 --> 01:20:56,410
And I don't think I need
to cover this specifically.

1600
01:20:56,410 --> 01:20:59,760
But let me get to the
conclusion on this.

1601
01:21:02,520 --> 01:21:04,980
The analysis we did on
this level of the model

1602
01:21:04,980 --> 01:21:10,860
shows, because of the natural
peaking in work requirements,

1603
01:21:10,860 --> 01:21:13,380
that you have much more
service in the peak

1604
01:21:13,380 --> 01:21:16,320
than in the middle
of the day, the base,

1605
01:21:16,320 --> 01:21:19,770
it doesn't matter that
absence rates are actually

1606
01:21:19,770 --> 01:21:22,380
higher in the peak than they
are in the middle of the day.

1607
01:21:22,380 --> 01:21:24,570
Because your
decisions are driven

1608
01:21:24,570 --> 01:21:28,370
by the same underlying dynamic.

1609
01:21:28,370 --> 01:21:33,770
And so you just can make
good report time decisions,

1610
01:21:33,770 --> 01:21:42,090
based on the average absence
rate on that day at that depot.

1611
01:21:42,090 --> 01:21:44,760
So you don't have to go into
super detailed analysis, which

1612
01:21:44,760 --> 01:21:47,510
is good news.

1613
01:21:47,510 --> 01:21:51,050
And in fact, you can
wind up with having

1614
01:21:51,050 --> 01:21:56,420
a ranked list of report times
that the desk manager at Cabot

1615
01:21:56,420 --> 01:21:59,550
can just work off that
list for tomorrow,

1616
01:21:59,550 --> 01:22:01,580
based on the number
of unassigned cover

1617
01:22:01,580 --> 01:22:03,050
operators you have.

1618
01:22:03,050 --> 01:22:05,540
And it will always give you
a good solution, as opposed

1619
01:22:05,540 --> 01:22:10,860
to the desk manager making
decisions de novo, every day.

1620
01:22:10,860 --> 01:22:17,450
This is a very simple decision
support tool that can help.

1621
01:22:17,450 --> 01:22:19,010
So the operational
level findings--

1622
01:22:19,010 --> 01:22:22,010
significant improvements
are possible,

1623
01:22:22,010 --> 01:22:25,070
can reduce the overtime
and missed trips,

1624
01:22:25,070 --> 01:22:27,920
and a single set of report times
can be used across all weekdays

1625
01:22:27,920 --> 01:22:31,160
and seasons for each garage.

1626
01:22:31,160 --> 01:22:34,430
Separate ranked report times
for Saturdays and Sundays,

1627
01:22:34,430 --> 01:22:36,290
and constant absence
rates can be assumed

1628
01:22:36,290 --> 01:22:37,870
by hour of day and day of week.

1629
01:22:37,870 --> 01:22:40,330
Those are the main conclusions.

1630
01:22:40,330 --> 01:22:43,280
OK, sorry I raced
towards the end.

1631
01:22:43,280 --> 01:22:45,380
But I'm available.

1632
01:22:45,380 --> 01:22:48,888
I am retiring, as
Gabriel well knows,

1633
01:22:48,888 --> 01:22:49,930
at the end of the summer.

1634
01:22:49,930 --> 01:22:54,110
This is probably my last lecture
as a non-retired MIT faculty

1635
01:22:54,110 --> 01:22:54,980
member.

1636
01:22:54,980 --> 01:22:58,310
It's been fun doing it with you.

1637
01:22:58,310 --> 01:23:00,120
And I'm around.

1638
01:23:00,120 --> 01:23:02,960
And I will be around after
the end of the summer as well.

1639
01:23:02,960 --> 01:23:05,420
Just I won't be
being paid by MIT.

1640
01:23:05,420 --> 01:23:07,340
So if anyone wants to
discuss any of this

1641
01:23:07,340 --> 01:23:09,810
or go through any of
the background topics,

1642
01:23:09,810 --> 01:23:12,080
please feel free to come see me.

1643
01:23:12,080 --> 01:23:12,830
OK?

1644
01:23:12,830 --> 01:23:13,730
Thanks very much.

1645
01:23:13,730 --> 01:23:17,080
[APPLAUSE]