1 00:00:00,000 --> 00:00:02,430 The following content is provided under a Creative 2 00:00:02,430 --> 00:00:03,730 Commons license. 3 00:00:03,730 --> 00:00:06,060 Your support will help MIT OpenCourseWare 4 00:00:06,060 --> 00:00:10,090 continue to offer high-quality educational resources for free. 5 00:00:10,090 --> 00:00:12,690 To make a donation or to view additional materials 6 00:00:12,690 --> 00:00:16,560 from hundreds of MIT courses, visit MIT OpenCourseWare 7 00:00:16,560 --> 00:00:17,744 at ocw.mit.edu. 8 00:00:25,720 --> 00:00:29,320 HUGH MCMANUS: Welcome to the Variability Simulation. 9 00:00:29,320 --> 00:00:32,170 This should look kind of familiar. 10 00:00:32,170 --> 00:00:35,330 This is a much simpler simulation than before. 11 00:00:35,330 --> 00:00:37,960 It's not going to take all day, but it's 12 00:00:37,960 --> 00:00:41,650 going to illustrate one point very well, which 13 00:00:41,650 --> 00:00:44,500 is what's the impact of variability 14 00:00:44,500 --> 00:00:45,970 on process performance? 15 00:00:45,970 --> 00:00:48,070 That's our learning objective and we're 16 00:00:48,070 --> 00:00:52,430 going to look at it through the poker chip simulation. 17 00:00:52,430 --> 00:00:54,487 We're going to look at a computer simulation 18 00:00:54,487 --> 00:00:56,320 because one of the problems with variability 19 00:00:56,320 --> 00:00:58,570 is sometimes to understand a variable system, 20 00:00:58,570 --> 00:01:02,110 you need a lot of data, a lot of cases. 21 00:01:02,110 --> 00:01:08,500 And we're also going to explore some simple relationships that 22 00:01:08,500 --> 00:01:14,290 help us, including that cue-theory equation that Earll 23 00:01:14,290 --> 00:01:18,070 showed, to help us understand variable processes. 24 00:01:18,070 --> 00:01:23,170 This is a piece of a rather old data on an engineering 25 00:01:23,170 --> 00:01:25,060 process with high variability. 26 00:01:25,060 --> 00:01:27,970 The traditional process shown here 27 00:01:27,970 --> 00:01:30,910 a the high cycle time and high variability 28 00:01:30,910 --> 00:01:33,820 was actually the process that was value stream mapped 29 00:01:33,820 --> 00:01:36,850 in the module we showed, I believe 30 00:01:36,850 --> 00:01:40,840 it was two rays ago, that had that very messy value stream 31 00:01:40,840 --> 00:01:43,400 map with all the arrows going everywhere. 32 00:01:43,400 --> 00:01:45,370 It was a very unpredictable process. 33 00:01:45,370 --> 00:01:47,680 It ended up having very high variability. 34 00:01:47,680 --> 00:01:49,720 Now, the problem with that, given 35 00:01:49,720 --> 00:01:52,390 that that process was embedded in a larger process, which 36 00:01:52,390 --> 00:01:55,300 was to produce an airplane, it was just the drawing release 37 00:01:55,300 --> 00:01:56,110 process. 38 00:01:56,110 --> 00:01:58,270 The problem wasn't even so much the fact 39 00:01:58,270 --> 00:02:01,340 that it took a long time, although that wasn't great. 40 00:02:01,340 --> 00:02:04,720 The problem was that the variability was almost as high 41 00:02:04,720 --> 00:02:08,289 as the actual time, which means it could take almost no time, 42 00:02:08,289 --> 00:02:10,180 it could take double the amount of time. 43 00:02:10,180 --> 00:02:14,170 And how easy is it to do organized work 44 00:02:14,170 --> 00:02:19,210 in an environment where even the allegedly simple processes are 45 00:02:19,210 --> 00:02:21,220 so unpredictable? 46 00:02:21,220 --> 00:02:22,298 It's very bad. 47 00:02:22,298 --> 00:02:24,340 Now, they managed to actually do a lot about that 48 00:02:24,340 --> 00:02:28,090 by applying Lean, but you can see 49 00:02:28,090 --> 00:02:32,130 how disruptive that high variability is going to be. 50 00:02:32,130 --> 00:02:34,750 The accounts payable module also reinforced that, 51 00:02:34,750 --> 00:02:36,850 that the high variability was very 52 00:02:36,850 --> 00:02:38,560 destructive to that process. 53 00:02:38,560 --> 00:02:41,050 On top of that, you're going to have quality problems 54 00:02:41,050 --> 00:02:42,580 with a variable process. 55 00:02:42,580 --> 00:02:43,750 That's highly likely. 56 00:02:43,750 --> 00:02:46,300 They may be variable because of mistakes, 57 00:02:46,300 --> 00:02:49,330 or it may be that the variable process makes it easier 58 00:02:49,330 --> 00:02:51,340 to make mistakes. 59 00:02:51,340 --> 00:02:53,290 And then we're also going to talk, 60 00:02:53,290 --> 00:02:57,490 both in Six Sigma and today, about the process capability 61 00:02:57,490 --> 00:02:59,890 being affected by variability. 62 00:02:59,890 --> 00:03:03,510 So we're going to do, I said this already, a dice game. 63 00:03:03,510 --> 00:03:05,260 We're going to do a computer sim and we're 64 00:03:05,260 --> 00:03:07,480 going to look at some simple relationships. 65 00:03:07,480 --> 00:03:09,893 We're also going to sort of foreshadow 66 00:03:09,893 --> 00:03:12,310 some stuff we're going to do in the Quality and Six Sigma. 67 00:03:12,310 --> 00:03:14,040 Not on these bullets-- 68 00:03:14,040 --> 00:03:15,730 I'm talking fast right now because I'm 69 00:03:15,730 --> 00:03:18,550 going to try to carve out a little bit of time 70 00:03:18,550 --> 00:03:21,700 at the end to discuss some of the things 71 00:03:21,700 --> 00:03:24,340 you can do about variability because I 72 00:03:24,340 --> 00:03:26,590 think that's particularly important to our health care 73 00:03:26,590 --> 00:03:27,173 audience. 74 00:03:27,173 --> 00:03:29,590 To some extent, our engineers, but particularly our health 75 00:03:29,590 --> 00:03:30,730 care people. 76 00:03:30,730 --> 00:03:33,460 One of the big issues with applying 77 00:03:33,460 --> 00:03:37,450 any kind of process improvement in a very high variability 78 00:03:37,450 --> 00:03:39,400 environment is just getting your mind 79 00:03:39,400 --> 00:03:41,350 wrapped around what does an average mean, what 80 00:03:41,350 --> 00:03:46,450 does even a standard deviation mean in a process 81 00:03:46,450 --> 00:03:48,910 where the variability has weird distributions 82 00:03:48,910 --> 00:03:51,820 and odd special cases and outliers? 83 00:03:51,820 --> 00:03:53,860 We'll try to carve out a little bit of time 84 00:03:53,860 --> 00:03:56,140 to actually talk about some of the practical things 85 00:03:56,140 --> 00:03:59,050 you can do to get variability under control. 86 00:03:59,050 --> 00:04:00,910 Here's a perfect system. 87 00:04:00,910 --> 00:04:02,860 What you are sitting in front of is 88 00:04:02,860 --> 00:04:06,850 a perfect, balanced, single-piece flow 89 00:04:06,850 --> 00:04:08,800 system, or near single. 90 00:04:08,800 --> 00:04:10,330 It's a very small-batch system. 91 00:04:10,330 --> 00:04:14,710 It has tasks arranged in order in, actually, 92 00:04:14,710 --> 00:04:17,390 a U-shaped shaped cell. 93 00:04:17,390 --> 00:04:19,510 It's not a pull cell, it's more of a push cell, 94 00:04:19,510 --> 00:04:22,510 but still, it's a U-shaped cell. 95 00:04:22,510 --> 00:04:26,140 Every task has an inventory in front of it, a little buffer 96 00:04:26,140 --> 00:04:28,870 inventory to make sure that it has the things it 97 00:04:28,870 --> 00:04:30,790 needs to do to do its work. 98 00:04:30,790 --> 00:04:33,230 Every task is perfectly balanced, 99 00:04:33,230 --> 00:04:38,230 so there's no bottlenecks in the system. 100 00:04:38,230 --> 00:04:42,700 The only thing imperfect about the system is its variability. 101 00:04:42,700 --> 00:04:45,910 And this half of the room, your mats 102 00:04:45,910 --> 00:04:52,165 actually say mailroom and PFR check and, well, 103 00:04:52,165 --> 00:04:53,560 you get the picture. 104 00:04:53,560 --> 00:04:57,460 We're looking at the accounts payable process here. 105 00:04:57,460 --> 00:04:59,500 You guys on the health care side actually 106 00:04:59,500 --> 00:05:02,560 have familiar-looking mats from yesterday 107 00:05:02,560 --> 00:05:04,300 because you had a process that had 108 00:05:04,300 --> 00:05:07,750 variability issues yesterday. 109 00:05:07,750 --> 00:05:10,930 The point here, both sides of the room are identical. 110 00:05:10,930 --> 00:05:13,150 And the point we're actually trying to make here 111 00:05:13,150 --> 00:05:15,820 is an awful lot of processes look like this. 112 00:05:15,820 --> 00:05:19,070 Even if you get them perfect, if they have variability in them, 113 00:05:19,070 --> 00:05:22,360 they're going to behave in ways that are actually even 114 00:05:22,360 --> 00:05:24,890 worse than one might expect. 115 00:05:24,890 --> 00:05:28,300 So let's find out how this system behaves. 116 00:05:28,300 --> 00:05:29,530 This is the game. 117 00:05:29,530 --> 00:05:32,890 It's a very simple game. 118 00:05:32,890 --> 00:05:38,590 What we're going to ask you to do is every day-- 119 00:05:38,590 --> 00:05:41,590 we're going to call the shifts day-- 120 00:05:41,590 --> 00:05:43,600 we are going to process chips. 121 00:05:43,600 --> 00:05:48,190 We are going to process one six-sided die worth of chips, 122 00:05:48,190 --> 00:05:50,590 so anywhere from one to six chips. 123 00:05:50,590 --> 00:05:52,270 There's our variability. 124 00:05:52,270 --> 00:05:54,940 The way we are going to process them is we 125 00:05:54,940 --> 00:05:57,770 are going to pick up chips from our inbox. 126 00:05:57,770 --> 00:06:00,280 If you are the customer or patient person, 127 00:06:00,280 --> 00:06:02,020 you have a big pile of chips, so you have 128 00:06:02,020 --> 00:06:03,670 an infinite quantity of chips. 129 00:06:03,670 --> 00:06:05,060 Just pick up chips from there. 130 00:06:05,060 --> 00:06:06,880 Everybody else has a finite number 131 00:06:06,880 --> 00:06:09,730 of chips that they can pick up, the ones in their inbox. 132 00:06:09,730 --> 00:06:14,980 So you will be rolling a dice, picking up that number of chips 133 00:06:14,980 --> 00:06:15,760 from your inbox. 134 00:06:15,760 --> 00:06:18,225 If you don't have enough, OK, too bad. 135 00:06:18,225 --> 00:06:19,600 If you roll a 6 and you only have 136 00:06:19,600 --> 00:06:21,850 three, which is what everybody has now, 137 00:06:21,850 --> 00:06:24,400 then you just pick up the three. 138 00:06:24,400 --> 00:06:28,330 And then everybody at the same time, to avoid confusion, 139 00:06:28,330 --> 00:06:31,510 passes those chips, the number of chips 140 00:06:31,510 --> 00:06:36,310 that they can pass, to the next person in line. 141 00:06:36,310 --> 00:06:38,020 And then at the end of each day, we 142 00:06:38,020 --> 00:06:40,870 will write down how many chips we passed. 143 00:06:40,870 --> 00:06:46,450 The only tricky part about that is that both the archive here 144 00:06:46,450 --> 00:06:49,030 and the discharge there, a certain number 145 00:06:49,030 --> 00:06:51,610 of chips every day will go into that mat, 146 00:06:51,610 --> 00:06:55,570 and the customers or the patients 147 00:06:55,570 --> 00:06:57,310 have to record that number. 148 00:06:57,310 --> 00:07:00,340 That's kind of the overall throughput of the system. 149 00:07:00,340 --> 00:07:02,680 So that's a slight trickiness at the end. 150 00:07:02,680 --> 00:07:04,540 That's what we're going to do. 151 00:07:04,540 --> 00:07:07,900 So for example, here's day one. 152 00:07:07,900 --> 00:07:09,550 Everybody has three chips. 153 00:07:09,550 --> 00:07:12,430 If this person rolls a 3, they pick up all three chips, 154 00:07:12,430 --> 00:07:13,960 send them along. 155 00:07:13,960 --> 00:07:16,840 This person rolls a 2, they can only pick up two chips. 156 00:07:16,840 --> 00:07:19,990 This person rolls a 5, they can only pick up three chips 157 00:07:19,990 --> 00:07:21,610 because that's all they have. 158 00:07:21,610 --> 00:07:24,530 No waiting for this person to send the two along, 159 00:07:24,530 --> 00:07:25,577 so then you have five-- 160 00:07:25,577 --> 00:07:26,410 no, that's cheating. 161 00:07:30,190 --> 00:07:34,340 The chips only get passed one station per day. 162 00:07:34,340 --> 00:07:36,460 This person rolls a 1, et cetera. 163 00:07:36,460 --> 00:07:39,430 This person rolls a 6, they only had three, so they pass three. 164 00:07:39,430 --> 00:07:42,430 And then this is what it looks like at the end of that day. 165 00:07:42,430 --> 00:07:44,660 Three have been passed along, two out, 166 00:07:44,660 --> 00:07:46,390 so the inventory went up. 167 00:07:46,390 --> 00:07:50,140 Three came out, two went in, so that went down a little bit. 168 00:07:50,140 --> 00:07:54,310 Three went in, only one left, so that's got more. 169 00:07:54,310 --> 00:07:57,440 Let's see, one went in and three came out, 170 00:07:57,440 --> 00:07:59,080 so that person only has one. 171 00:07:59,080 --> 00:08:02,630 The overall throughput of the system was three. 172 00:08:02,630 --> 00:08:06,970 We're going to record how many did you do, how many did you 173 00:08:06,970 --> 00:08:07,820 pass along? 174 00:08:07,820 --> 00:08:10,510 So either your die roll or how many you had, 175 00:08:10,510 --> 00:08:11,980 whatever number is smaller. 176 00:08:11,980 --> 00:08:13,480 How many did you pass along? 177 00:08:13,480 --> 00:08:15,220 And then at the end of the round, 178 00:08:15,220 --> 00:08:16,510 what is your work in progress? 179 00:08:16,510 --> 00:08:19,300 So everyone starts with three, so that number is already 180 00:08:19,300 --> 00:08:20,170 filled in. 181 00:08:20,170 --> 00:08:23,240 At the end of each day, you're going to record your number. 182 00:08:23,240 --> 00:08:25,070 Yeah, so here's an example. 183 00:08:25,070 --> 00:08:29,420 Our person there passed one job along and some work built up, 184 00:08:29,420 --> 00:08:35,770 so at the end of that day, they had five, et cetera. 185 00:08:35,770 --> 00:08:37,780 The customer or patient worksheet 186 00:08:37,780 --> 00:08:40,150 is a little bit more complicated because they're 187 00:08:40,150 --> 00:08:42,870 sending the work out. 188 00:08:42,870 --> 00:08:45,370 That's basically always going to be the die roll because you 189 00:08:45,370 --> 00:08:48,930 have lots and lots of chips. 190 00:08:48,930 --> 00:08:51,930 You're going to record the number of jobs that actually 191 00:08:51,930 --> 00:08:54,660 came through the system, and here, the U-shaped cell 192 00:08:54,660 --> 00:08:55,930 is going to help us. 193 00:08:55,930 --> 00:08:57,630 So the number of jobs that are added-- 194 00:08:57,630 --> 00:09:02,450 not in, but added-- to the finished pile at each round 195 00:09:02,450 --> 00:09:04,763 is recorded here. 196 00:09:04,763 --> 00:09:06,180 And then you're also going to keep 197 00:09:06,180 --> 00:09:08,880 track of the total WIP on the whole table, the total work 198 00:09:08,880 --> 00:09:11,460 in progress in the whole table, which starts at 12. 199 00:09:11,460 --> 00:09:13,240 Now, you could just count every time, 200 00:09:13,240 --> 00:09:16,050 but that would be really slow, so instead, we're 201 00:09:16,050 --> 00:09:18,630 going to use a conservation law. 202 00:09:18,630 --> 00:09:21,870 Earll's an old aerodynamicist, so they love conservation laws, 203 00:09:21,870 --> 00:09:24,480 right? 204 00:09:24,480 --> 00:09:28,560 A former, a retired aerodynamicist. 205 00:09:28,560 --> 00:09:29,970 A former aerodynamicist. 206 00:09:29,970 --> 00:09:31,200 They love conservation laws. 207 00:09:31,200 --> 00:09:33,180 This is the conservation of chips laws. 208 00:09:33,180 --> 00:09:37,290 If you know that you sent out six and three got finished, 209 00:09:37,290 --> 00:09:39,540 then three must have gone into the system. 210 00:09:39,540 --> 00:09:42,510 And they didn't fall on the floor or get eaten or anything, 211 00:09:42,510 --> 00:09:45,083 so this number must have gone up by three. 212 00:09:45,083 --> 00:09:46,500 So there's a little equation there 213 00:09:46,500 --> 00:09:53,040 that will help you figure out the new work in progress. 214 00:09:53,040 --> 00:09:56,037 So everybody think they can do that much math? 215 00:09:56,037 --> 00:09:58,120 And I think that's actually written on your sheet, 216 00:09:58,120 --> 00:10:01,360 so you should be able to do that. 217 00:10:01,360 --> 00:10:02,890 OK, and there's an example. 218 00:10:02,890 --> 00:10:06,040 If three go in and three come out, it stays the same. 219 00:10:06,040 --> 00:10:10,540 If two go in and one comes out, then that extra chip 220 00:10:10,540 --> 00:10:13,450 must be sitting on the table, so it goes up by one. 221 00:10:13,450 --> 00:10:16,510 And we can keep doing that ad infinitum. 222 00:10:16,510 --> 00:10:18,267 OK, what do we think should happen? 223 00:10:18,267 --> 00:10:20,850 AUDIENCE: Maybe we should just ask if there are any questions. 224 00:10:20,850 --> 00:10:23,030 HUGH MCMANUS: Any questions about the process? 225 00:10:23,030 --> 00:10:26,950 AUDIENCE: So I [INAUDIBLE] and I roll the die, 226 00:10:26,950 --> 00:10:30,040 and I get like, 5. 227 00:10:30,040 --> 00:10:32,610 So I pass on three because I only have three chips. 228 00:10:32,610 --> 00:10:35,110 HUGH MCMANUS: Because you only have the three, that's right. 229 00:10:35,110 --> 00:10:38,160 AUDIENCE: So under patients completed, 230 00:10:38,160 --> 00:10:39,130 I'll have three there? 231 00:10:39,130 --> 00:10:40,547 HUGH MCMANUS: You put three, yeah. 232 00:10:40,547 --> 00:10:43,143 AUDIENCE: Then what will now be the work in progress? 233 00:10:43,143 --> 00:10:44,560 HUGH MCMANUS: The work in progress 234 00:10:44,560 --> 00:10:48,340 is whatever the person next to you gives to you. 235 00:10:48,340 --> 00:10:51,340 Because everybody's going to pass them at the same time. 236 00:10:51,340 --> 00:10:53,048 We're going to be very formal about that. 237 00:10:53,048 --> 00:10:54,882 We're going to have a standard process where 238 00:10:54,882 --> 00:10:57,770 we're going to ask everybody to pass them all at the same time. 239 00:10:57,770 --> 00:10:59,770 So then yeah, you'll just have to look and see-- 240 00:10:59,770 --> 00:11:00,820 AUDIENCE: It's actually what you have 241 00:11:00,820 --> 00:11:02,155 left by the end of the round. 242 00:11:02,155 --> 00:11:03,280 HUGH MCMANUS: That's right. 243 00:11:03,280 --> 00:11:04,330 That's correct. 244 00:11:04,330 --> 00:11:07,420 AUDIENCE: If you rolled a 3 and had four chips here, 245 00:11:07,420 --> 00:11:10,730 you would move three, and you'd take one left plus whatever 246 00:11:10,730 --> 00:11:11,230 new stuff-- 247 00:11:11,230 --> 00:11:12,490 HUGH MCMANUS: Plus whatever payment. 248 00:11:12,490 --> 00:11:14,080 So it's whatever's sitting on that mat 249 00:11:14,080 --> 00:11:17,590 at the end of the round is your work in progress. 250 00:11:17,590 --> 00:11:22,360 So the average of a six-sided die is 3 and 1/2, 251 00:11:22,360 --> 00:11:26,210 so what do we think this system should be able to do over-- 252 00:11:26,210 --> 00:11:27,790 we're going to do 20 rounds. 253 00:11:27,790 --> 00:11:30,123 What do we think it should be able to do over 20 rounds? 254 00:11:32,210 --> 00:11:33,640 Is it going to process 3 and 1/2? 255 00:11:33,640 --> 00:11:35,240 Are these die rolls going to even out? 256 00:11:38,440 --> 00:11:40,010 Maybe. 257 00:11:40,010 --> 00:11:42,688 Everyone's thinking. 258 00:11:42,688 --> 00:11:44,560 AUDIENCE: 60? 259 00:11:44,560 --> 00:11:46,880 HUGH MCMANUS: 60, so that would be 3ish. 260 00:11:46,880 --> 00:11:49,070 So it's not going to be perfect, but it's 261 00:11:49,070 --> 00:11:50,670 going to be close to evened out. 262 00:11:50,670 --> 00:11:52,420 AUDIENCE: So we don't process more than 3. 263 00:11:52,420 --> 00:11:54,355 [INAUDIBLE] 264 00:11:54,355 --> 00:11:55,980 HUGH MCMANUS: Yeah, in the first round, 265 00:11:55,980 --> 00:11:57,680 you're only going to get 3. 266 00:11:57,680 --> 00:11:59,840 More will come in as the customer 267 00:11:59,840 --> 00:12:02,360 person rolls their dice. 268 00:12:02,360 --> 00:12:04,664 The initiating person. 269 00:12:04,664 --> 00:12:07,075 [INTERPOSING VOICES] 270 00:12:07,075 --> 00:12:07,818 You think less? 271 00:12:07,818 --> 00:12:08,318 Less, OK. 272 00:12:08,318 --> 00:12:18,300 [INTERPOSING VOICES] 273 00:12:18,300 --> 00:12:18,800 Less than? 274 00:12:18,800 --> 00:12:19,805 AUDIENCE: 35. 275 00:12:19,805 --> 00:12:20,930 HUGH MCMANUS: Less than 35. 276 00:12:20,930 --> 00:12:22,850 OK, you're thinking a lot less, OK. 277 00:12:22,850 --> 00:12:23,600 Yeah. 278 00:12:23,600 --> 00:12:25,860 [INTERPOSING VOICES] 279 00:12:25,860 --> 00:12:26,360 OK. 280 00:12:26,360 --> 00:12:28,310 So yeah, people are struggling with this. 281 00:12:28,310 --> 00:12:30,370 Intuitively, this is hard. 282 00:12:30,370 --> 00:12:33,290 It's hard to figure out what this thing should do. 283 00:12:33,290 --> 00:12:35,720 I've already sort of biased the conversation 284 00:12:35,720 --> 00:12:39,410 by intimating that the system's going to behave badly. 285 00:12:39,410 --> 00:12:42,080 But one's intuition, unfortunately, 286 00:12:42,080 --> 00:12:46,040 about random systems is that they should even out. 287 00:12:46,040 --> 00:12:49,310 The way the human brain works is badly 288 00:12:49,310 --> 00:12:53,420 biased towards just figuring random things will even out. 289 00:12:53,420 --> 00:12:56,150 There's all sorts of interesting gambling fallacies 290 00:12:56,150 --> 00:12:58,955 that are almost hardwired into people's brains. 291 00:12:58,955 --> 00:13:02,660 The "law of averages"-- now, there is no law of averages. 292 00:13:02,660 --> 00:13:06,260 It tends that there is-- 293 00:13:06,260 --> 00:13:07,070 we'll see. 294 00:13:07,070 --> 00:13:10,400 I won't go into that any deeper than that because what we're 295 00:13:10,400 --> 00:13:14,100 going to do, given that we can't really figure it out a priori, 296 00:13:14,100 --> 00:13:15,650 and if we try to do statistics on it, 297 00:13:15,650 --> 00:13:17,660 it actually gets really complicated really fast, 298 00:13:17,660 --> 00:13:18,993 we're going to do it, all right? 299 00:13:18,993 --> 00:13:23,432 So let's go ahead and do our process. 300 00:13:23,432 --> 00:13:24,890 OK, what I'm going to ask everybody 301 00:13:24,890 --> 00:13:27,620 to do the first couple of rounds is follow my cue. 302 00:13:27,620 --> 00:13:30,202 I'm going to dictate the process. 303 00:13:30,202 --> 00:13:31,910 AUDIENCE: And we're going to try to check 304 00:13:31,910 --> 00:13:32,900 and make sure you're doing it. 305 00:13:32,900 --> 00:13:33,530 HUGH MCMANUS: Right. 306 00:13:33,530 --> 00:13:35,300 And our facilitators will check and make 307 00:13:35,300 --> 00:13:36,110 sure you're doing it right. 308 00:13:36,110 --> 00:13:37,400 The first thing we do is pick up the dice. 309 00:13:37,400 --> 00:13:38,990 Everybody holding their dice? 310 00:13:38,990 --> 00:13:40,970 OK, roll it. 311 00:13:40,970 --> 00:13:46,190 OK, pick up that many chips, or three if you only got three. 312 00:13:46,190 --> 00:13:52,070 And now everybody together pass, passing to the right, yeah. 313 00:13:52,070 --> 00:13:53,342 This is a properly done-- 314 00:13:53,342 --> 00:13:54,800 actually, more or less by accident, 315 00:13:54,800 --> 00:13:56,180 properly done U-shaped cell. 316 00:13:56,180 --> 00:13:58,010 We're passing to the right, unless you're 317 00:13:58,010 --> 00:14:01,180 left-handed, in which case it's not so proper. 318 00:14:01,180 --> 00:14:02,180 OK, everybody done that? 319 00:14:02,180 --> 00:14:06,590 Now write down what you did, how many did you pass? 320 00:14:06,590 --> 00:14:11,170 OK, the customer person has a slightly harder job. 321 00:14:11,170 --> 00:14:14,150 Everybody figuring that out? 322 00:14:14,150 --> 00:14:15,670 Yeah, OK. 323 00:14:15,670 --> 00:14:19,690 So the only communication necessary 324 00:14:19,690 --> 00:14:25,240 should be the number of chips finished, right? 325 00:14:25,240 --> 00:14:27,160 All right, day two. 326 00:14:27,160 --> 00:14:28,710 Everybody roll the dice. 327 00:14:32,750 --> 00:14:36,230 Pick up that number of dice or whatever you have, 328 00:14:36,230 --> 00:14:39,430 and pass it to the right. 329 00:14:39,430 --> 00:14:40,480 Write down your results. 330 00:14:44,350 --> 00:14:47,680 The amount that was finished needs to be communicated, 331 00:14:47,680 --> 00:14:49,720 no other talking please. 332 00:14:49,720 --> 00:14:51,615 And finally, day 10. 333 00:14:51,615 --> 00:14:53,290 Not finally, but day 10. 334 00:14:53,290 --> 00:14:58,980 Roll dice, pick up chips, pass them to the right, 335 00:14:58,980 --> 00:15:00,390 and write everything down. 336 00:15:02,910 --> 00:15:07,190 Now we're going to pause and ask how this process is doing. 337 00:15:07,190 --> 00:15:12,630 Who has only one chip? 338 00:15:12,630 --> 00:15:14,820 A couple of people have only one chip. 339 00:15:14,820 --> 00:15:19,730 Who has eight ships or more? 340 00:15:19,730 --> 00:15:20,900 Eight or more. 341 00:15:20,900 --> 00:15:21,680 10 or more? 342 00:15:24,733 --> 00:15:25,900 Looks like we have a winner. 343 00:15:25,900 --> 00:15:27,078 How many? 344 00:15:27,078 --> 00:15:27,620 AUDIENCE: 14. 345 00:15:27,620 --> 00:15:30,295 HUGH MCMANUS: 14 chips. 346 00:15:30,295 --> 00:15:31,390 AUDIENCE: Just yesterday. 347 00:15:33,325 --> 00:15:34,450 HUGH MCMANUS: That's right. 348 00:15:34,450 --> 00:15:37,240 Yeah, 6's are good today. 349 00:15:37,240 --> 00:15:40,430 So yeah, that's the one difference. 350 00:15:40,430 --> 00:15:42,490 So what should we do about this process? 351 00:15:42,490 --> 00:15:44,800 We've got at least two people who 352 00:15:44,800 --> 00:15:47,110 are keeping their in-bin clear. 353 00:15:47,110 --> 00:15:51,880 They're sitting back, definitely they have extra capacity. 354 00:15:51,880 --> 00:15:54,550 Yeah, we've got at least one person who just cannot keep up 355 00:15:54,550 --> 00:15:59,173 and probably has to go, right? 356 00:15:59,173 --> 00:16:03,010 [INTERPOSING VOICES] 357 00:16:03,010 --> 00:16:07,550 So does that mean there's a bottleneck there? 358 00:16:07,550 --> 00:16:10,170 Must be, right? 359 00:16:10,170 --> 00:16:12,760 No, of course not, she has the same dice as everybody else. 360 00:16:12,760 --> 00:16:15,170 This is a transparent-- 361 00:16:15,170 --> 00:16:18,090 this process is behaving badly and is actually throwing out 362 00:16:18,090 --> 00:16:20,850 bad clues about itself. 363 00:16:20,850 --> 00:16:22,380 That looks like a bottleneck. 364 00:16:22,380 --> 00:16:24,600 It isn't, it's just bad luck. 365 00:16:24,600 --> 00:16:26,760 Some people look like they're doing really well. 366 00:16:26,760 --> 00:16:30,060 No, they're just lucky. 367 00:16:30,060 --> 00:16:34,500 And let's see if the trends continue. 368 00:16:34,500 --> 00:16:37,960 Day 11, roll the dice, pick up the chips, 369 00:16:37,960 --> 00:16:40,630 pass them to the right, write things down. 370 00:16:45,370 --> 00:16:46,300 And 20. 371 00:16:50,120 --> 00:16:57,800 Pick them up, hand them along, write everything down. 372 00:16:57,800 --> 00:17:00,600 OK, so how did that go? 373 00:17:00,600 --> 00:17:04,170 Let's see, who has 14 or better? 374 00:17:04,170 --> 00:17:07,050 See if we broke our old record. 375 00:17:07,050 --> 00:17:07,829 Several people. 376 00:17:07,829 --> 00:17:09,317 What have we got? 377 00:17:09,317 --> 00:17:10,109 AUDIENCE: I got 15. 378 00:17:10,109 --> 00:17:13,050 HUGH MCMANUS: 15. 379 00:17:13,050 --> 00:17:14,547 What do we got here? 380 00:17:14,547 --> 00:17:15,089 AUDIENCE: 13. 381 00:17:15,089 --> 00:17:17,040 HUGH MCMANUS: 13, oh not quite. 382 00:17:17,040 --> 00:17:18,540 So 15, we have a new winner. 383 00:17:21,630 --> 00:17:23,575 How is our problem employee now? 384 00:17:23,575 --> 00:17:25,450 AUDIENCE: A little bit better, only at eight. 385 00:17:25,450 --> 00:17:26,450 HUGH MCMANUS: Eight, OK. 386 00:17:26,450 --> 00:17:29,100 Down to eight, so magically got better. 387 00:17:29,100 --> 00:17:31,290 How about our people that had one before? 388 00:17:31,290 --> 00:17:33,780 Anybody have one now? 389 00:17:33,780 --> 00:17:37,350 We still got some people with not so many. 390 00:17:37,350 --> 00:17:39,690 I think we have one repeat winner, do we not? 391 00:17:39,690 --> 00:17:40,770 OK. 392 00:17:40,770 --> 00:17:42,780 So there we go. 393 00:17:42,780 --> 00:17:45,330 The system is, in fact, behaving fairly badly. 394 00:17:45,330 --> 00:17:48,700 We even had an unintended lesson in failure of standard process, 395 00:17:48,700 --> 00:17:49,200 right? 396 00:17:49,200 --> 00:17:50,790 When we got off standard process, 397 00:17:50,790 --> 00:17:55,260 we also had some high variability in the performance 398 00:17:55,260 --> 00:17:57,605 of our process. 399 00:17:57,605 --> 00:17:58,980 There's some calculations you can 400 00:17:58,980 --> 00:18:00,330 do they're self-explanatory. 401 00:18:00,330 --> 00:18:02,970 We're going to give you a few minutes to do that. 402 00:18:02,970 --> 00:18:04,500 Only a few. 403 00:18:04,500 --> 00:18:07,960 If you don't quite get them finished, that's probably OK. 404 00:18:07,960 --> 00:18:10,080 The one thing we would really like to see 405 00:18:10,080 --> 00:18:14,290 is the customer just getting the average throughput. 406 00:18:14,290 --> 00:18:17,110 That first average for the customer 407 00:18:17,110 --> 00:18:19,620 is an important number because we 408 00:18:19,620 --> 00:18:24,857 want to see what our performance looks like. 409 00:18:24,857 --> 00:18:25,690 Oh that's a new one. 410 00:18:25,690 --> 00:18:26,540 OK, very good. 411 00:18:29,080 --> 00:18:32,710 So you got 2.7 average throughput on m1. 412 00:18:35,370 --> 00:18:37,750 What's your average throughput, 2.6? 413 00:18:37,750 --> 00:18:39,360 OK. 414 00:18:39,360 --> 00:18:41,340 Looks like people are finishing up the numbers. 415 00:18:41,340 --> 00:18:49,615 2.6, average 3.05. 416 00:18:49,615 --> 00:18:55,610 2.3 and got another average? 417 00:18:55,610 --> 00:18:57,950 Still working on it, OK. 418 00:18:57,950 --> 00:19:00,410 What about utilizations? 419 00:19:00,410 --> 00:19:05,140 What was our average utilization at our first medical table? 420 00:19:05,140 --> 00:19:06,020 AUDIENCE: 0.77 421 00:19:06,020 --> 00:19:10,629 HUGH MCMANUS: 0.77, OK. 422 00:19:10,629 --> 00:19:11,605 AUDIENCE: 0.74. 423 00:19:11,605 --> 00:19:12,480 HUGH MCMANUS: Pardon? 424 00:19:12,480 --> 00:19:13,200 AUDIENCE: 0.74. 425 00:19:13,200 --> 00:19:18,380 HUGH MCMANUS: 0.74, OK. 426 00:19:18,380 --> 00:19:19,860 AUDIENCE: 0.87. 427 00:19:19,860 --> 00:19:21,090 HUGH MCMANUS: 0.87, OK. 428 00:19:21,090 --> 00:19:22,896 We got a good one there. 429 00:19:22,896 --> 00:19:24,146 AUDIENCE: 0.66. 430 00:19:24,146 --> 00:19:32,320 HUGH MCMANUS: 0.66 2.8, and what's the utilization? 431 00:19:32,320 --> 00:19:33,200 AUDIENCE: 0.8. 432 00:19:33,200 --> 00:19:35,030 HUGH MCMANUS: 0.8, OK. 433 00:19:35,030 --> 00:19:36,770 So interesting thing here. 434 00:19:39,710 --> 00:19:42,570 And we can look at some of the other details. 435 00:19:42,570 --> 00:19:46,310 Another thing, I just asked you to look at your personal sheet 436 00:19:46,310 --> 00:19:47,270 there. 437 00:19:47,270 --> 00:19:50,210 Is there any strong trend in the inventories? 438 00:19:52,910 --> 00:19:56,350 Inventories were definitely going up from 3 maybe. 439 00:19:56,350 --> 00:19:59,320 Some of them were going down. 440 00:19:59,320 --> 00:20:03,520 What's the overall trend in inventory? 441 00:20:03,520 --> 00:20:05,644 Anyone see a trend? 442 00:20:05,644 --> 00:20:07,630 AUDIENCE: It went up and stabilized. 443 00:20:07,630 --> 00:20:09,310 HUGH MCMANUS: Went up and stabilized? 444 00:20:09,310 --> 00:20:10,270 AUDIENCE: Give or take. 445 00:20:10,270 --> 00:20:11,395 HUGH MCMANUS: Give or take. 446 00:20:11,395 --> 00:20:13,713 Went up and stabilized, give or take. 447 00:20:13,713 --> 00:20:15,130 AUDIENCE: Ours just kept going up. 448 00:20:15,130 --> 00:20:17,690 HUGH MCMANUS: Just kept going up. 449 00:20:17,690 --> 00:20:18,190 Went up? 450 00:20:18,190 --> 00:20:18,850 OK. 451 00:20:18,850 --> 00:20:21,310 That was even with three people with only one? 452 00:20:21,310 --> 00:20:23,120 AUDIENCE: The average overall went up. 453 00:20:23,120 --> 00:20:24,770 HUGH MCMANUS: The average overall still went up, 454 00:20:24,770 --> 00:20:25,562 that's interesting. 455 00:20:25,562 --> 00:20:27,116 That's interesting. 456 00:20:27,116 --> 00:20:34,090 And what was the personal productivity, the utilization 457 00:20:34,090 --> 00:20:37,810 factor, of, say, you, the one who never had any inventory? 458 00:20:37,810 --> 00:20:39,460 What was your personal productivity? 459 00:20:39,460 --> 00:20:40,360 AUDIENCE: 0.74. 460 00:20:40,360 --> 00:20:41,500 HUGH MCMANUS: 0.74. 461 00:20:41,500 --> 00:20:45,437 And what about the people that had lots of inventory? 462 00:20:48,250 --> 00:20:49,210 AUDIENCE: 0.68 463 00:20:49,210 --> 00:20:51,313 HUGH MCMANUS: 0.68. 464 00:20:51,313 --> 00:20:52,230 AUDIENCE: Utilization? 465 00:20:52,230 --> 00:20:53,022 HUGH MCMANUS: Yeah. 466 00:20:53,022 --> 00:20:53,730 AUDIENCE: 0.88. 467 00:20:53,730 --> 00:20:55,710 HUGH MCMANUS: 0.88, OK. 468 00:20:55,710 --> 00:20:57,580 So this person's working very efficiently 469 00:20:57,580 --> 00:20:59,250 but has lots of inventory. 470 00:20:59,250 --> 00:21:01,390 This person is working not very efficiently, 471 00:21:01,390 --> 00:21:03,230 but has lots of inventory. 472 00:21:03,230 --> 00:21:04,950 This person doesn't have very much-- 473 00:21:04,950 --> 00:21:07,020 this system is behaving very oddly. 474 00:21:07,020 --> 00:21:14,740 It's behaving in ways that are difficult to understand. 475 00:21:14,740 --> 00:21:17,315 So why? 476 00:21:17,315 --> 00:21:19,440 AUDIENCE: We're not always passing the exact number 477 00:21:19,440 --> 00:21:20,077 that we roll. 478 00:21:20,077 --> 00:21:21,660 HUGH MCMANUS: We're not always passing 479 00:21:21,660 --> 00:21:22,905 the exact number we roll. 480 00:21:22,905 --> 00:21:27,985 AUDIENCE: In a lower average process [INAUDIBLE].. 481 00:21:27,985 --> 00:21:30,360 HUGH MCMANUS: OK, so there's some unused capacity, right? 482 00:21:30,360 --> 00:21:33,430 You roll a 6, you only have one. 483 00:21:33,430 --> 00:21:35,000 But the inventory's going up. 484 00:21:35,000 --> 00:21:37,720 Is that problem going to go away? 485 00:21:37,720 --> 00:21:39,070 AUDIENCE: Eventually. 486 00:21:39,070 --> 00:21:41,650 HUGH MCMANUS: Eventually. 487 00:21:41,650 --> 00:21:42,340 Good answer. 488 00:21:46,270 --> 00:21:48,770 I mean, we got at the root of it. 489 00:21:48,770 --> 00:21:50,650 It was also sort of a system problem, right? 490 00:21:50,650 --> 00:21:53,780 Which is OK, it's behaving badly because of that, 491 00:21:53,780 --> 00:21:56,830 which makes it behave worse because your inventory is 492 00:21:56,830 --> 00:21:57,760 unpredictable. 493 00:21:57,760 --> 00:22:00,733 So even if there's lots of inventory, 494 00:22:00,733 --> 00:22:02,650 the chance that there would be a low inventory 495 00:22:02,650 --> 00:22:04,910 and you'd waste a 6 is still there. 496 00:22:04,910 --> 00:22:08,380 It doesn't go away even as the inventory comes up. 497 00:22:08,380 --> 00:22:12,250 We don't know what's going to happen 498 00:22:12,250 --> 00:22:15,238 and things depend on each other. 499 00:22:15,238 --> 00:22:16,780 The amount that you have in inventory 500 00:22:16,780 --> 00:22:20,350 depends on the person next to you who's rolling dice. 501 00:22:20,350 --> 00:22:22,713 Even if you're doing great, if they 502 00:22:22,713 --> 00:22:24,130 roll a whole string of 1's, you've 503 00:22:24,130 --> 00:22:26,560 got nothing to work with. 504 00:22:26,560 --> 00:22:30,610 So even if you're rolling all 6's or precisely on 3 and 1/2, 505 00:22:30,610 --> 00:22:33,490 you're still not being able to perform because of the way 506 00:22:33,490 --> 00:22:34,870 the system is hooked together. 507 00:22:34,870 --> 00:22:39,190 And this is really the key-- every time anybody rolls a 6 508 00:22:39,190 --> 00:22:42,190 and doesn't have enough chips, that capacity is wasted 509 00:22:42,190 --> 00:22:44,690 and it never comes back. 510 00:22:44,690 --> 00:22:45,860 You can never get it back. 511 00:22:45,860 --> 00:22:47,925 And that's where the "law of averages" fails. 512 00:22:50,565 --> 00:22:54,280 In mechanical engineering, we call it a ratchet effect. 513 00:22:54,280 --> 00:22:56,640 It can go one way, but it can't go the other. 514 00:22:56,640 --> 00:23:01,190 We can lose capacity, we never get it back. 515 00:23:01,190 --> 00:23:04,760 So that's why we're seeing these lower utilizations even 516 00:23:04,760 --> 00:23:06,710 though we did 20 rounds, which you'd think 517 00:23:06,710 --> 00:23:09,200 would be enough time for things to even out. 518 00:23:09,200 --> 00:23:12,170 OK, now we're going to run a computer simulation. 519 00:23:12,170 --> 00:23:14,330 The first thing that any person when 520 00:23:14,330 --> 00:23:17,480 confronted with tangled data like this should think is we 521 00:23:17,480 --> 00:23:18,690 need more data. 522 00:23:18,690 --> 00:23:20,300 And in fact, we do. 523 00:23:20,300 --> 00:23:23,400 Although as we will see in this particular case, 524 00:23:23,400 --> 00:23:25,842 even a very large amount of data does not help us. 525 00:23:25,842 --> 00:23:27,800 We're going to run a computer simulation, which 526 00:23:27,800 --> 00:23:32,180 does this little exercise for 20, but then also 200 527 00:23:32,180 --> 00:23:34,100 and something days. 528 00:23:34,100 --> 00:23:35,540 Looks like this. 529 00:23:35,540 --> 00:23:40,160 What's shown here is days and this is the inventory level 530 00:23:40,160 --> 00:23:41,270 at the different stations. 531 00:23:41,270 --> 00:23:45,800 The notations here are of the AP case, but again, 532 00:23:45,800 --> 00:23:48,500 it's the same simulation for the medical case-- 533 00:23:48,500 --> 00:23:52,700 the four stations, the inventory at the four stations, 534 00:23:52,700 --> 00:23:54,168 versus time. 535 00:23:54,168 --> 00:23:55,710 There's a couple of things down here. 536 00:23:55,710 --> 00:23:58,910 This is our processing capability. 537 00:23:58,910 --> 00:24:01,760 This is the WIP that's on the table, the total WIP. 538 00:24:01,760 --> 00:24:02,750 This is the cycle time. 539 00:24:02,750 --> 00:24:04,820 This is how long it takes the average chip 540 00:24:04,820 --> 00:24:06,680 to move through the system. 541 00:24:06,680 --> 00:24:08,180 And this is just a little luck thing 542 00:24:08,180 --> 00:24:10,880 to just see if maybe some of this variation 543 00:24:10,880 --> 00:24:12,890 is just due to luck. 544 00:24:12,890 --> 00:24:17,090 Maybe our average die roll is better or worse, 545 00:24:17,090 --> 00:24:18,980 but we're rolling an awful lot of die, 546 00:24:18,980 --> 00:24:23,750 so we do expect the actual average of the die rolls 547 00:24:23,750 --> 00:24:29,750 to kind of even out even if the system itself doesn't. 548 00:24:29,750 --> 00:24:33,590 OK, so first thing we can notice about the system is OK, 549 00:24:33,590 --> 00:24:34,940 we're running it 200 times. 550 00:24:34,940 --> 00:24:38,300 Is it settling down? 551 00:24:38,300 --> 00:24:39,440 It's not. 552 00:24:39,440 --> 00:24:43,790 Just waiting for it to settle down is not going to work. 553 00:24:43,790 --> 00:24:48,880 And let's see, did that do anything? 554 00:24:48,880 --> 00:24:50,190 Oh, there we go. 555 00:24:50,190 --> 00:24:51,660 Select that. 556 00:24:51,660 --> 00:24:54,290 Boink, I just ran it again. 557 00:24:54,290 --> 00:24:56,010 Came out different. 558 00:24:56,010 --> 00:24:59,020 If I run it again, different. 559 00:24:59,020 --> 00:25:02,550 Even the gross behavior of the system is not predictable. 560 00:25:02,550 --> 00:25:06,430 It doesn't like, get bigger and then kind of even off. 561 00:25:06,430 --> 00:25:08,770 If I ran it a billion times and took the average, 562 00:25:08,770 --> 00:25:11,145 I'd actually get sort of a logarithmic curve 563 00:25:11,145 --> 00:25:12,520 where it would go up and steadily 564 00:25:12,520 --> 00:25:15,550 flatten out, although never actually reach an equilibrium. 565 00:25:15,550 --> 00:25:18,760 But the behavior even over a large number 566 00:25:18,760 --> 00:25:20,800 of rounds of any given instance of this system 567 00:25:20,800 --> 00:25:22,540 is extremely unpredictable. 568 00:25:22,540 --> 00:25:25,510 So perfect system except for variability. 569 00:25:25,510 --> 00:25:30,090 The variability is making the system behave very badly. 570 00:25:30,090 --> 00:25:31,530 OK, what can we do about this? 571 00:25:31,530 --> 00:25:33,530 The other nice thing about a computer simulation 572 00:25:33,530 --> 00:25:36,000 is we can change it relatively easy. 573 00:25:36,000 --> 00:25:40,300 We could, for example, lower the customer input variability. 574 00:25:40,300 --> 00:25:42,630 That's a favorite of bad processes. 575 00:25:42,630 --> 00:25:44,427 People say, it's our customer's fault. 576 00:25:44,427 --> 00:25:46,260 There's too much customer input variability. 577 00:25:46,260 --> 00:25:50,850 Like all patients show up Friday night, so it's not our fault. 578 00:25:50,850 --> 00:25:53,580 Let's lower the customer input variability some. 579 00:25:56,143 --> 00:25:57,060 Didn't help very much. 580 00:25:57,060 --> 00:25:58,680 Let's lower it a lot. 581 00:25:58,680 --> 00:26:00,180 Didn't help very much. 582 00:26:00,180 --> 00:26:01,290 Why? 583 00:26:01,290 --> 00:26:03,880 That appeared to help a little bit. 584 00:26:03,880 --> 00:26:05,310 Yeah, that's a little better. 585 00:26:05,310 --> 00:26:09,420 That's with a lot less customer input variability. 586 00:26:09,420 --> 00:26:12,210 But basically, both count. 587 00:26:12,210 --> 00:26:14,640 The process variability and the customer input 588 00:26:14,640 --> 00:26:17,270 variability count a lot. 589 00:26:17,270 --> 00:26:19,160 Let's do this, though. 590 00:26:19,160 --> 00:26:24,260 Let's lower the customer demand just a little bit. 591 00:26:24,260 --> 00:26:29,770 Oh, it looks about the same as lowering the variability a lot. 592 00:26:29,770 --> 00:26:33,300 Let's lower the customer demand 20%. 593 00:26:39,440 --> 00:26:41,415 Doesn't look great, but what's changed? 594 00:26:48,560 --> 00:26:51,060 It looks like there's a lot less variability in the process. 595 00:26:53,700 --> 00:27:00,090 And if we lower the customer variability by 30%, 596 00:27:00,090 --> 00:27:03,750 there's still variability, but the system can handle it. 597 00:27:03,750 --> 00:27:05,970 Essentially we're destressing the system, 598 00:27:05,970 --> 00:27:10,200 running it at a little bit less than 100% capacity, 599 00:27:10,200 --> 00:27:12,340 and it behaves a lot better. 600 00:27:12,340 --> 00:27:14,550 And this was brought up in the AP system 601 00:27:14,550 --> 00:27:17,740 where they were running at 100% capacity. 602 00:27:17,740 --> 00:27:19,020 It's the same in this system. 603 00:27:19,020 --> 00:27:23,850 Basically, if we try to stress it, we only get about 70%. 604 00:27:23,850 --> 00:27:26,970 If we say, you know, it's only going to run at 70% anyway, 605 00:27:26,970 --> 00:27:29,010 let's just give it 70% of the load. 606 00:27:29,010 --> 00:27:31,380 It actually behaves pretty well. 607 00:27:31,380 --> 00:27:36,540 This is a nonintuitive effect unless you're 608 00:27:36,540 --> 00:27:37,540 one of those workers. 609 00:27:37,540 --> 00:27:40,140 What happens if you have to work all the time 610 00:27:40,140 --> 00:27:41,220 and something goes wrong? 611 00:27:41,220 --> 00:27:42,180 Chaos, right? 612 00:27:42,180 --> 00:27:43,980 Crap, I have to stay up all night, 613 00:27:43,980 --> 00:27:48,040 but I can't, I'm too tired, so things just go to hell. 614 00:27:48,040 --> 00:27:51,240 What happens if you have six hours of work 615 00:27:51,240 --> 00:27:53,220 in an eight-hour day and something goes wrong? 616 00:27:55,970 --> 00:27:58,900 You can fix it. 617 00:27:58,900 --> 00:28:00,820 From an individual worker's point of view, 618 00:28:00,820 --> 00:28:03,310 this is very easy to understand. 619 00:28:03,310 --> 00:28:05,383 The system effect seems kind of mysterious, 620 00:28:05,383 --> 00:28:08,050 but actually, if you think about what's happening on the ground, 621 00:28:08,050 --> 00:28:09,970 if the system is not fully stressed, 622 00:28:09,970 --> 00:28:11,740 it can handle variability. 623 00:28:11,740 --> 00:28:15,010 And you don't have to back off on the system all that much 624 00:28:15,010 --> 00:28:19,170 to handle quite a lot of variability. 625 00:28:19,170 --> 00:28:21,420 Yeah, I understand that we're tight on time. 626 00:28:21,420 --> 00:28:26,890 We're coming to a close here fortunately. 627 00:28:26,890 --> 00:28:28,900 So that's kind of magic. 628 00:28:28,900 --> 00:28:31,720 The other thing we can do is attack variability, 629 00:28:31,720 --> 00:28:34,810 and if we attack variability heroically-- 630 00:28:34,810 --> 00:28:40,210 let's reduce variability 30% across the board-- 631 00:28:40,210 --> 00:28:41,680 it helps some. 632 00:28:44,530 --> 00:28:48,760 Yeah, it's better, but it's still behaving kind of badly. 633 00:28:48,760 --> 00:28:52,030 If we reduce variability 70% across the board, 634 00:28:52,030 --> 00:28:55,037 this is truly heroic variability reduction. 635 00:29:00,030 --> 00:29:02,550 That didn't work well. 636 00:29:02,550 --> 00:29:05,050 Oh, there we go. 637 00:29:05,050 --> 00:29:08,070 Notice when it was all but one, it was still no good. 638 00:29:08,070 --> 00:29:10,920 We have now reduced variability heroically across the board, 639 00:29:10,920 --> 00:29:14,520 we've got about the same level of effect as reducing the-- 640 00:29:14,520 --> 00:29:19,680 so apparently, according to the simulation, what we need 641 00:29:19,680 --> 00:29:26,850 is either heroic variability reduction or a modest decrease 642 00:29:26,850 --> 00:29:29,230 in the load on the system. 643 00:29:29,230 --> 00:29:33,900 And this is actually a very powerful practical lesson 644 00:29:33,900 --> 00:29:37,170 and it comes from that same equation 645 00:29:37,170 --> 00:29:40,180 that Earll was showing. 646 00:29:40,180 --> 00:29:44,110 If we're mathematically inclined here, 647 00:29:44,110 --> 00:29:45,490 what happens when our utilization 648 00:29:45,490 --> 00:29:48,680 is very high on this term? 649 00:29:48,680 --> 00:29:51,870 Just put a 1 in there and see what happens mathematically. 650 00:29:51,870 --> 00:29:53,750 It goes to infinity, right? 651 00:29:53,750 --> 00:29:55,320 If the system is completely loaded 652 00:29:55,320 --> 00:29:57,630 and there's any variability at all, 653 00:29:57,630 --> 00:30:01,430 it will just continue to get worse and worse in its behavior 654 00:30:01,430 --> 00:30:04,130 if it's 100% loaded. 655 00:30:04,130 --> 00:30:06,960 The variability was actually squared, 656 00:30:06,960 --> 00:30:10,410 so if we can reduce variability, that's good, 657 00:30:10,410 --> 00:30:13,080 but this effect is actually even more powerful. 658 00:30:13,080 --> 00:30:17,880 And if we plot the two effects, here's 659 00:30:17,880 --> 00:30:22,020 a plot of what happens if we control variability 660 00:30:22,020 --> 00:30:23,370 for various utilizations. 661 00:30:23,370 --> 00:30:27,330 If we have a heavily loaded system-- that's this one-- 662 00:30:27,330 --> 00:30:29,670 and we reduce variability, once we've 663 00:30:29,670 --> 00:30:31,710 driven it down really low-- 664 00:30:31,710 --> 00:30:33,513 this is the cue time or inventory, 665 00:30:33,513 --> 00:30:34,680 they're basically the same-- 666 00:30:37,560 --> 00:30:40,710 it will start to behave. 667 00:30:40,710 --> 00:30:43,260 But if there's noticeable variation 668 00:30:43,260 --> 00:30:47,040 on a heavily loaded system, it's going to behave badly. 669 00:30:47,040 --> 00:30:50,585 If it's not so heavily loaded, it almost doesn't matter. 670 00:30:50,585 --> 00:30:52,710 This is a lightly loaded system and the variability 671 00:30:52,710 --> 00:30:54,793 doesn't matter, this is a moderately loaded system 672 00:30:54,793 --> 00:30:57,510 and reducing variability is good. 673 00:30:57,510 --> 00:31:00,090 And that's not really linear, but basically, 674 00:31:00,090 --> 00:31:02,610 sort of proportionally, if we reduce variability, 675 00:31:02,610 --> 00:31:03,960 we get better behavior. 676 00:31:03,960 --> 00:31:05,610 But in a very heavily loaded system, 677 00:31:05,610 --> 00:31:07,350 we really have to drive it to 0. 678 00:31:07,350 --> 00:31:09,430 This is Six Sigma. 679 00:31:09,430 --> 00:31:12,070 If we have a system that we want to load really heavily, 680 00:31:12,070 --> 00:31:15,550 we want to make billions of ICs, then 681 00:31:15,550 --> 00:31:18,010 we have to drive the variability very close to 0 682 00:31:18,010 --> 00:31:20,570 to get the system to behave. 683 00:31:20,570 --> 00:31:22,690 This is the converse thought. 684 00:31:22,690 --> 00:31:25,900 If we ease off on utilization for various variabilities, 685 00:31:25,900 --> 00:31:27,490 what we see is that blow-up effect, 686 00:31:27,490 --> 00:31:30,730 that divided by 0 as we get close to 1. 687 00:31:30,730 --> 00:31:33,160 And it's almost independent of the variability level. 688 00:31:33,160 --> 00:31:39,190 At any level, if we back off some from 100% utilization, 689 00:31:39,190 --> 00:31:41,890 the behavior actually gets a lot better. 690 00:31:41,890 --> 00:31:42,980 Question? 691 00:31:42,980 --> 00:31:47,050 AUDIENCE: I'm just wondering how organizational management deals 692 00:31:47,050 --> 00:31:49,510 with this fact if they know that they 693 00:31:49,510 --> 00:31:51,728 need to operate significantly under their capacity-- 694 00:31:51,728 --> 00:31:53,395 HUGH MCMANUS: That's a tough one, right? 695 00:31:53,395 --> 00:31:55,228 AUDIENCE: --in order to control their system 696 00:31:55,228 --> 00:31:59,275 versus having a perfect system without variation. 697 00:31:59,275 --> 00:32:00,400 HUGH MCMANUS: That's right. 698 00:32:00,400 --> 00:32:00,942 That's right. 699 00:32:00,942 --> 00:32:03,760 Managers like Six Sigma because if your system is perfect, 700 00:32:03,760 --> 00:32:06,220 you can load it at 100%. 701 00:32:06,220 --> 00:32:07,450 They want to hear that. 702 00:32:07,450 --> 00:32:09,490 That's a message they want to hear. 703 00:32:09,490 --> 00:32:11,980 In systems like health care or engineering, 704 00:32:11,980 --> 00:32:16,840 getting the variability very, very low is very, very hard. 705 00:32:16,840 --> 00:32:21,580 And this is a very rich field. 706 00:32:21,580 --> 00:32:23,470 In the interest of time, I'm going 707 00:32:23,470 --> 00:32:28,010 to jump to a couple of useful tips. 708 00:32:28,010 --> 00:32:32,205 One is that one of the purposes of Lean, 709 00:32:32,205 --> 00:32:33,580 and this was actually illustrated 710 00:32:33,580 --> 00:32:35,650 very nicely by the AP case study, 711 00:32:35,650 --> 00:32:38,140 is to free up those resources. 712 00:32:38,140 --> 00:32:42,800 Not in order to lay people off or necessarily save money even, 713 00:32:42,800 --> 00:32:48,910 but if you can take 10% or 20% of your non-value-added time 714 00:32:48,910 --> 00:32:51,640 out of your day, what does that get you? 715 00:32:51,640 --> 00:32:54,190 That gets you that headroom that allows 716 00:32:54,190 --> 00:32:56,800 you to deal with variability. 717 00:32:56,800 --> 00:33:00,640 And it's one of the sort of root causes of why Lean seems 718 00:33:00,640 --> 00:33:04,330 to have a magic effect of increasing quality 719 00:33:04,330 --> 00:33:07,720 and throughput time when that's not even what you're aiming at. 720 00:33:07,720 --> 00:33:09,370 If you just take waste out of a system, 721 00:33:09,370 --> 00:33:12,130 suddenly everybody in the system has a little bit of extra time 722 00:33:12,130 --> 00:33:17,050 to deal with variation, to deal with things going wrong. 723 00:33:17,050 --> 00:33:21,860 So that's sort of the hint, number one. 724 00:33:21,860 --> 00:33:26,260 The other thing is that in terms of variability reduction, 725 00:33:26,260 --> 00:33:28,670 although that's an art, there's no one answer. 726 00:33:28,670 --> 00:33:34,200 There are things that help a lot. 727 00:33:34,200 --> 00:33:36,870 Reducing something like health care variability to 0 728 00:33:36,870 --> 00:33:41,370 is obviously not even possible, but one 729 00:33:41,370 --> 00:33:44,400 of the things about actually both health 730 00:33:44,400 --> 00:33:46,170 care and engineering-type processes 731 00:33:46,170 --> 00:33:48,960 is that the distributions aren't really normal. 732 00:33:48,960 --> 00:33:51,160 It tends to be like, a bunch of standard cases 733 00:33:51,160 --> 00:33:54,150 and then some wacko ones. 734 00:33:54,150 --> 00:33:56,162 Well, what happens to the variability 735 00:33:56,162 --> 00:33:58,620 in terms of those equations if you just take the wacko ones 736 00:33:58,620 --> 00:33:59,300 off the table? 737 00:34:02,490 --> 00:34:03,960 Gets a lot better, right? 738 00:34:03,960 --> 00:34:05,260 It gets a lot better. 739 00:34:05,260 --> 00:34:10,080 So essentially, special case handling for the outliers. 740 00:34:10,080 --> 00:34:12,162 In the simulation game, we originally 741 00:34:12,162 --> 00:34:13,620 designed the health care simulation 742 00:34:13,620 --> 00:34:15,495 to be part of a three-day seminar where you'd 743 00:34:15,495 --> 00:34:19,110 play that last round at the end, like after this lesson, 744 00:34:19,110 --> 00:34:21,500 and there's something in there called a patient advocate. 745 00:34:21,500 --> 00:34:23,830 Anybody even look at that card? 746 00:34:23,830 --> 00:34:25,680 What did it do? 747 00:34:25,680 --> 00:34:28,679 AUDIENCE: It helped those patients that 748 00:34:28,679 --> 00:34:31,025 would take the most amount of time and helped 749 00:34:31,025 --> 00:34:32,400 shepherd them through the system. 750 00:34:32,400 --> 00:34:34,199 HUGH MCMANUS: Right, exactly. 751 00:34:34,199 --> 00:34:37,739 In terms of the simulation, it took the 6's off, right? 752 00:34:37,739 --> 00:34:40,350 And if you looked at the distributions, 753 00:34:40,350 --> 00:34:41,639 they were highly non-standard. 754 00:34:41,639 --> 00:34:43,199 The 6's were horrible. 755 00:34:43,199 --> 00:34:48,090 And so that's actually a real example 756 00:34:48,090 --> 00:34:52,500 of taking the really difficult cases out of the standard flow 757 00:34:52,500 --> 00:34:54,929 of the system, and that makes the standard flow 758 00:34:54,929 --> 00:34:55,679 work a lot better. 759 00:34:55,679 --> 00:34:57,554 Of course, you have to devote some resources, 760 00:34:57,554 --> 00:35:00,000 some special resources, to the difficult people. 761 00:35:00,000 --> 00:35:03,420 But there's ways you can kind of knock the top off 762 00:35:03,420 --> 00:35:06,660 of high variabilities that can help a lot on that side. 763 00:35:06,660 --> 00:35:09,090 On the capacity side, the real key 764 00:35:09,090 --> 00:35:13,020 is to use the benefits of any kind of Lean 765 00:35:13,020 --> 00:35:15,210 to essentially plow them back into making 766 00:35:15,210 --> 00:35:20,070 people's lives easier so they have a little bit of extra time 767 00:35:20,070 --> 00:35:22,880 to absorb the variability.