A talk by David Lowe and James Wyllie delivered at Agile on the Beach, Falmouth, on 3 September 2015. The talk covered a variety of approaches to estimate and forecast, including planning poker, Kanban metrics, calibration, Student's t-statistic and Monte Carlo simulations.
2. Story Points & Velocity
Story Points
“They are a unit of measure for expressing the overall size of a user story or feature. They tell us how big a story is, relative to
others, either in terms of size or complexity”
- Mike Cohn
5. Story Points & Velocity
Velocity
total story points / no. of sprints
Last 5 sprints completed points were
30, 40, 50, 60, 70 = 250 points / 5 sprints = velocity of 50 points
How big is my project?
Total estimate (points) / velocity = number of sprints
e.g. 500 story points for project / 50 points = 10 sprints
How long will it take?
Number of sprints x length of sprint (weeks)
e.g. 2 weeks sprints = 10 sprints x 2 weeks = 20 weeks
https://www.mountaingoatsoftware.com/tools/velocity-range-calculator
8. Kanban Metrics
Little’s Law
Average Lead Time (ALT) = Average WIP / Average DR
Average WIP (AW) = Average Lead Time x Average DR
Average Delivery Rate (DR) = Average WIP / Average Lead Time
“Slap anyone who uses little’s law to create forecasts”
- Daniel Vacanti
https://leanpub.com/actionableagilemetrics
Basic Forecast
Forecast = (WI(n) / DR) + ALT
9. Monte Carlo simulators
(5.99 x 127,583) - 312,270 = 451,952
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(5.99 x 127,583) - 312,270 = 451,952
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.49 x 26,715) - 305,923 = -159,255
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(8.49 x 45,202) - 228,423 = 155,341
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(6.99 x 117,371) - 245,352 = 575,073
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.00 x 14,054) - 329,609 = -217,180
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.49 x 73,265) - 369,820 = 178,935
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.49 x 7,904) - 207,013 = -139,911
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(5.99 x 148,909) - 206,799 = 685,166
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(7.99 x 23,805) - 375,232 = -185,034
(5.99 x 40,272) - 300,293 = -59,064
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(7.99 x 23,805) - 375,232 = -185,034
(5.99 x 42,630) - 253,040 = 2,314
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(7.99 x 23,805) - 375,232 = -185,034
(5.99 x 42,630) - 253,040 = 2,314
… and on …
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(7.99 x 23,805) - 375,232 = -185,034
(5.99 x 42,630) - 253,040 = 2,314
… and on … and on …
(5.99 x 151,186) - 228,727 = 676,880
(8.99 x 35,272) - 333,723 = -16,618
(7.99 x 124,891) - 368,288 = 629,595
(8.99 x 76,506) - 388,655 = 299,138
(7.99 x 23,805) - 375,232 = -185,034
(5.99 x 42,630) - 253,040 = 2,314
… and on … and on … and on …
10. “the future of forecasting in knowledge work”
~ Dan Vacanti
12. Calibration I’m 80% confident
it will be between
0mm and 100mm
Estimate the
average
rainfall in
September
13.
14. Student’s t-statistic
William Sealy Gosset (13 June 1876 – 16 October 1937)
https://upload.wikimedia.org/wikipedia/commons/thumb/4/42/William_Sealy_Gosset.jpg/766px-William_Sealy_Gosset.jpg
A small amount of data can be
used to forecast the future … if
you take the appropriate level of
error into account
15. Conclusions
Measurement
“A quantitatively expressed reduction of uncertainty based on one or more observations”
- Douglas Hubbard
4 useful measurement assumptions:
● Your problem is not as unique as you think
● You have more data than you think
● An adequate amount of data is more accessible than you think
● You need less data than you think
17. Conclusions
“It is better to be roughly right than precisely wrong.”
~ John Maynard Keynes, British Economist
Can Guinness help you estimate?
18. Conclusions
“It is better to be roughly right than precisely wrong.”
~ John Maynard Keynes, British Economist
Can Guinness help you estimate?
forecast?
@jamespwyllie & @bigpinots
Notes de l'éditeur
Everyone has heard of story points
Planning Poker is the mechanism a team can use for coming to a shared understanding of that relative size
Cards with the fibonacci sequence (i.e. each number is a product of the previous 2). Mike and mountain goat has trademarked his own sequence
PRACTICAL EXERCISE HERE
Discuss the story, everyone chooses a card they think represents the size … everyone declares them at once. This way no-one is led by anyone else.
Try to gain consensus by having outliers justify why they think it’s a certain number.
Agreement made when we have a definitive size and a shared understanding of the work to be done.
Sprint Planning
Size stories (or work items) and team agree what will be brought into the sprint. The Scrum Master uses team velocity as a guide to ensure they don’t over-commit
Not going to talk about ideal hours
Mike’s velocity calculator
Nice touch with the confidence level. Will be a running theme of this talk
URL will be on the last slide
Little’s Law
Defines the relationship between Lead Time, WIP and Delivery Rate
As WIP and Lead Time increase, Delivery Rate goes down.
We get more, more quickly by doing less at the same time
If you have a team of 5, and you have 10 WIs in progress, that’s an average of 2 per person. How can they each be doing two things at once?
Work items relatively sized. How big doesn’t matter as we are averaging. Clearly you don’t want work items too big though.
Application of Little’s Law to produce a basic forecast
G(n) = (WI(n) / DR) + ALT
Is this the right thing to do though? Can we do better than using averages?
Vacanti: “Slap anyone who uses little’s law to create forecasts”