Debunking Risk Management

What works and what doesn't for software development project.

Transcript

1. alexey@kuznetsov.co
3. ☺ ▪ ▪ 3
4. ☺ ▪ ▪ ▪ 4
5. ☺ ▪ ▪ ▪ 5
6. ☺ 6
7. ☺ 7
8. Everyone wants to look like a hero 8
9. Manager of your (my) project normally looks this way 9
10. 10 PM is responsible for everything, but he/she is not a superman
11. CHAOS Report (Standish Group). Why they show this all the time? 11
12. Even inaccurate report doesn’t proove that Risk Management is helpful 12
13. Train wagon ☺ 13
14. Figures of cost and probability are both quite uncertain 14
16. Risk exposure is $400, which is not enough to cover any of these risks should they materialize. Risk exposure IS ALCHEMY. But some RM coaches are still pushing it. You can do better using technique called «They are not going to give me more anyway». Some risks have materialized? So be it, you have maximum reserves to deal with them. Simple and easy to understand. 16
17. Estimate and risk are like ice and water: not the same but you turn one into another. Estimate which is too low guarantees trivial circumstances to have high impact for your project delivery date. Approved exaggerated estimate means the opposite. Why don’t we exaggerate estimates? Because everyone looks at them ☺ Risks are catching much less attention, because at the end of the day they are product of your what-if fantasies and you can produce as many of them as you want. You have too big estimate! Okay, now we have too much risk. So... You want to produce minimal estimate and sleep well. You need to eliminate known risks. 17
18. г О Т Т Т δi = δ = sqrt(δ1* δ1 +..+ δn*δn) = 5.055 Tδ = E + δ = 17.66+5.055 ~ 22.7 T2δ = E + 2*δ = 17.66+10.11 ~ 27.8 T3δ = E + 3*δ = 17.66+15.165 ~ 32.8 δ ~ probability of success 68% 2δ ~ probability of success 95% 3δ ~ probability of success 99% 18
19. Sum of a large number of weakly dependent random values has a distribution close to normal (very strong statement, much stronger than it is presented in a high school): Normal distribution chart. Probability is a square below the chart 19
20. • • ≤ • • 20
22. Software development project is not a subject to physical randomness 22
23. Longer you wait, more complicated life becomes. 23
24. If you’re an external vendor add assumptions in your contract Add redundancies of various kinds Or don’t bother taking this project Vendors almost never follow #3 :) 24
25. • • • • • • • ☺ • … 25
26. • • • • • 26
27. • • • Software gets softer and more elastic. Java is simpler to refactor than C++. 27
28. Insurance business Your savings Strategic weapon systems Aircraft, submarines and spacecraft Banknotes attached to your driving license (don’t do this in N. America) Emergency exits Unlike expert opinions, redundancy has a precise cost and always works 28
29. Nassim Nicholas Taleb 29
30. Taleb, Nassim Nicholas The Black Swan: The Impact of the Highly Improbable New York: Random House, 2007. ISBN 978-1-4000-6351-5. 30
31. Spyros G. Makridakis, Robin M. Hogarth, Anil Gaba Dance with Chance: Making Luck Work for You Oxford: Oneworld Publications, 2009 ISBN: 978-1-85168-679-7 31