<<

Estimating By The Zodiac

2017 NASA Cost and Schedule Analysis Symposium NASA Headquarters, Washington DC August 2017 Joe Hamaker, PhD Galorath Federal

© 2017 Copyright Galorath Incorporated 1 Introduction

• One day a senior cost estimator arrived at work early and as she was walking across the nearly empty parking lot from her car to her office building, a piece of paper blew across the pavement • Being a good citizen, she picked up the paper, planning to dispose of it properly once inside the building • To her surprise, she noticed that across the top of the piece of paper, in bold letters, it said “The Zodiac Model: World’s Best Spacecraft Cost Model” • The body of the paper contained an equation • Studying the equation a bit more, she determined that it purported to be a CER that would estimate the development cost of spacecraft based on several variables- - variables that didn’t make any engineering sense at all

© 2017 Copyright Galorath Incorporated 2 Zodiac Model Variables

• The astrological sign of the Project Manager coded as follows: • = 1 • = 2 • = 3 • Etc. • The percent of the project team who were vegetarians • For an indicator variable called Zeus, the model instructed the user to wear a tin foil hat to shield the brain from electromagnetic fields; then flip a coin: Heads assign 0, Tails assign 1 to the variable Zeus • Percent of PC vs Mac users on the team • Take Project Office cost estimate in millions, multiply by pi, and use the resulting number for Man Wearing a Tin Foil Hat the spacecraft mass in kilograms • The number of alphanumeric characters that were in the headlines of the morning edition of the New York Times on the first day of the Project PDR • The Chief Engineer’s shoe size in European units

© 2017 Copyright Galorath Incorporated 3 Zodiac Model CER

• Where… • Zodiac = Astrological sign of PM coded as Aquarius=1, Pisces=2, =3, etc. • The percent of the project team who were vegetarians • For an indicator variable called Zeus, wear a tin foil hat; then flip a coin: Heads assign 0, Tails assign 1 • Percent of PC vs Mac users on the team • Project Office cost estimate in millions multiplied by pi, for the spacecraft mass in kilograms • Number of alphanumeric characters in the headlines of the morning edition of the New York Times on the first day of the Project PDR • The Chief Engineer’s shoe size in European units

© 2017 Copyright Galorath Incorporated 4 Data To Test The Model

• While recognizing that was ridiculous, our estimator was working on a project for which she thought she could gather the data for the model • (Although obtaining the Zodiac sign of the PM and the shoe size of the Chief Engineer was a bit awkward) • The data was…

Variable Value Zodiac sign of the Project Manager, Aries 3 Percent of the team that are vegetarians 7% Zeus variable (from coin flip in tin foil hat) 1 Percent of PC users vs Mac users on the team 55% Mass variable (Project estimate of $370M X 3.14 = 1162 kg) 1162 Headline variable 11 Chief Engineer's shoe size in European Units 34

© 2017 Copyright Galorath Incorporated 5 Exercising The Zodiac Model • Next our intrepid estimator exercised the CER

Variable Value Zodiac sign of the Project Manager, Aries 3 Percent of the team that are vegetarians 7% Zeus variable (from coin flip in tin foil hat) 1 Percent of PC users vs Mac users on the team 55% Mass variable (Project estimate of $370M X 3.14 = 1162 kg) 1162 Headline variable 11 Chief Engineer's shoe size in European Units 34 • Cost (Then Year $) = =

=1210.3 * 0.55+((1162)^0.5)*11/34 = $677M

© 2017 Copyright Galorath Incorporated 6 Using The Zodiac Model On A Larger Sample

• Our estimator noticed that our result of $677M was nearly double the Project Office Estimate of $370M • That in itself proved nothing • So with greater effort over the next several days, as time permitted, she gathered the data for a dozen historical missions and ran them through the CER • The percent errors were all very small—all within a few percentage points of the actual cost and just about evenly distributed between positive and negative • Not a single percent error was over 5% • This CER was arguably better than anything she had ever seen before for a spacecraft cost model! © 2017 Copyright Galorath Incorporated 7 The Story Continues….

• Now as it happened, our estimator was teaching a beginning cost estimating class a few days later and she showed her students the results of the Zodiac model • She told the class exactly how she came by the CER and that she had no knowledge of the data base behind it • She asked the class to ruminate on this over lunch and when they returned, she wanted to lead them in a discussion of the efficacy, or lack thereof, of this model • After lunch, about 1/3 of the students objected to the CER totally, about 1/3 weren’t sure and about 1/3 thought that since it seemed to work, it should be used • Our senior estimator prepared to voice her opinion • She paused for dramatic effect…The students all leaned forward expectantly… • Then she said…..

© 2017 Copyright Galorath Incorporated 8 What do you think the teacher should have told her class?

© 2017 Copyright Galorath Incorporated 9 Conclusions

• As we all know, correlation does not imply causation • I submit that when using cost models, even if we don’t have firm knowledge about causation, we should not necessarily care • From https://blogs.scientificamerican.com/guest-blog/9-bizarre- and-surprising-insights-from-data-science/ • Female-named hurricanes are more deadly • Over 6 decades of data show that hurricanes with “relatively feminine” names killed almost three times the number killed by hurricanes with “relatively male” names • So if that is true, it seems to me we’d be foolish not to prepare more for female named hurricanes even though we can't explain the reason • (At that site above they make the conjecture that it is because people become passive when the coming is named Sue but take it more seriously if it is name Samson) • Everyday we use tools we don’t fully understand (e.g. the internet) • My assertion is, we’d be foolish NOT to use the Zodiac Model!

© 2017 Copyright Galorath Incorporated 10 Afterword… • So what Zodiac has been about is… • To have a little fun, and • Make the point that sometimes we stress too much over the statistics of our models • Tending to forget that, at the end of the day, it really matters… • That a model does a decent job predicting cost • After all, we all know that weather models include inputs regarding a butterfly flapping it’s wings in South America • So it shouldn’t bother us that the Zodiac Model

© 2017 Copyright Galorath Incorporated 11