Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 1 00:00:01,380 --> 00:00:03,800 Let's look at another estimation method, 2 2 00:00:03,800 --> 00:00:05,750 the Delphi technique. 3 3 00:00:05,750 --> 00:00:08,233 Named for the Oracle of Delphi in Ancient Greece. 4 4 00:00:09,490 --> 00:00:11,900 The Oracle of Delphi lived in a cave, 5 5 00:00:11,900 --> 00:00:13,190 and had the magic ability 6 6 00:00:13,190 --> 00:00:15,233 to answer any question about the future. 7 7 00:00:16,130 --> 00:00:17,770 It had a cute trick though. 8 8 00:00:17,770 --> 00:00:19,500 It would phrase the answer in such a way 9 9 00:00:19,500 --> 00:00:21,430 that no one could understand it, 10 10 00:00:21,430 --> 00:00:23,540 until the future event arrived. 11 11 00:00:23,540 --> 00:00:24,630 When suddenly it was obvious 12 12 00:00:24,630 --> 00:00:26,230 the prediction had been correct. 13 13 00:00:27,570 --> 00:00:29,080 The modern Delphi technique was 14 14 00:00:29,080 --> 00:00:31,950 invented by the RAND corporation in 1948, 15 15 00:00:31,950 --> 00:00:34,210 and the chief feature is the gathering 16 16 00:00:34,210 --> 00:00:36,970 of a group of experts to get a better result 17 17 00:00:36,970 --> 00:00:38,023 than a single guess. 18 18 00:00:39,800 --> 00:00:43,570 Delphi can be used for estimating deliverable time and cost. 19 19 00:00:43,570 --> 00:00:45,910 However, it's primarily used to estimate 20 20 00:00:45,910 --> 00:00:48,730 risk probability and time, 21 21 00:00:48,730 --> 00:00:51,353 where there's often little other information to go by. 22 22 00:00:52,300 --> 00:00:54,283 The process works like this. 23 23 00:00:55,720 --> 00:00:57,730 Gather your core project team, 24 24 00:00:57,730 --> 00:01:00,020 or a special group of subject matter experts 25 25 00:01:00,020 --> 00:01:02,240 if assessing a domain-specific item 26 26 00:01:02,240 --> 00:01:04,100 such as a group of construction personnel 27 27 00:01:04,100 --> 00:01:06,300 if assessing a construction risk, 28 28 00:01:06,300 --> 00:01:08,823 and provide them with all relevant information. 29 29 00:01:09,779 --> 00:01:12,670 Each member then writes down their estimate 30 30 00:01:12,670 --> 00:01:13,670 on a piece of paper. 31 31 00:01:14,870 --> 00:01:17,230 Here's a key value of the technique, 32 32 00:01:17,230 --> 00:01:20,890 everyone then reveals their estimate at the same time. 33 33 00:01:20,890 --> 00:01:23,240 Removing the significant influence of early estimates 34 34 00:01:23,240 --> 00:01:25,913 on later ones, if you had gone around the table. 35 35 00:01:27,070 --> 00:01:29,260 You might think you wouldn't change your estimate 36 36 00:01:29,260 --> 00:01:30,564 if you heard others first, 37 37 00:01:30,564 --> 00:01:33,543 but studies show the effect is very large. 38 38 00:01:34,720 --> 00:01:36,600 This is why when supreme courts 39 39 00:01:36,600 --> 00:01:38,630 first meet to discuss a case, 40 40 00:01:38,630 --> 00:01:40,330 the first time they get the justices in a room 41 41 00:01:40,330 --> 00:01:41,890 to exchange views, 42 42 00:01:41,890 --> 00:01:44,347 the most junior justice speaks first, 43 43 00:01:44,347 --> 00:01:46,220 then the next most junior, 44 44 00:01:46,220 --> 00:01:47,370 then the next, 45 45 00:01:47,370 --> 00:01:50,003 until the most senior judge speaks last. 46 46 00:01:51,580 --> 00:01:53,680 By all having all members write down their estimate 47 47 00:01:53,680 --> 00:01:55,680 and then turn them over all at once, 48 48 00:01:55,680 --> 00:01:58,780 Delphi estimation gets everyone's real views 49 49 00:01:58,780 --> 00:01:59,630 out on the table. 50 50 00:02:01,550 --> 00:02:04,240 Each member then discusses their reasoning 51 51 00:02:04,240 --> 00:02:05,523 behind their estimate. 52 52 00:02:06,530 --> 00:02:08,580 Others ask them questions as appropriate, 53 53 00:02:09,830 --> 00:02:11,290 then you iterate, 54 54 00:02:11,290 --> 00:02:14,050 each member writes down their next estimate, 55 55 00:02:14,050 --> 00:02:16,540 everyone reveals them all at once, 56 56 00:02:16,540 --> 00:02:17,960 and everyone discusses the reasons 57 57 00:02:17,960 --> 00:02:19,253 for their updated number. 58 58 00:02:20,780 --> 00:02:23,060 Usually less than five rounds are needed 59 59 00:02:23,060 --> 00:02:24,940 to get a consensus estimate. 60 60 00:02:24,940 --> 00:02:27,773 The process works quite well in practice. 61 61 00:02:28,940 --> 00:02:33,100 If the estimates are all within 10% or after five rounds, 62 62 00:02:33,100 --> 00:02:35,033 you can then average the numbers. 63 63 00:02:36,700 --> 00:02:39,330 This technique really works well. 64 64 00:02:39,330 --> 00:02:41,700 However, if the team is new to it, 65 65 00:02:41,700 --> 00:02:43,750 you should brief them on the process first, 66 66 00:02:43,750 --> 00:02:46,430 and emphasize the goal is not to win 67 67 00:02:46,430 --> 00:02:49,220 by talking the others into your estimate, 68 68 00:02:49,220 --> 00:02:51,460 but to share knowledge and factor it 69 69 00:02:51,460 --> 00:02:53,703 into your updated estimates for each round. 70 70 00:02:55,610 --> 00:02:57,600 Delphi produces a much better 71 71 00:02:57,600 --> 00:03:00,150 and much more defensible estimate 72 72 00:03:00,150 --> 00:03:01,563 than an individual guess. 73 73 00:03:02,780 --> 00:03:04,030 As a rule of thumb, 74 74 00:03:04,030 --> 00:03:05,130 it gives you a number with about 75 75 00:03:05,130 --> 00:03:07,373 plus or minus 33% accuracy. 76 76 00:03:09,580 --> 00:03:11,440 Delphi is often the best technique 77 77 00:03:11,440 --> 00:03:14,870 for estimating risk time and percentage. 78 78 00:03:14,870 --> 00:03:17,740 But once again, if the project really counts, 79 79 00:03:17,740 --> 00:03:19,280 use activity breakdown 80 80 00:03:19,280 --> 00:03:21,023 for deliverable estimating. 6580