Beyond Story Points - Forecasting with empirical data
Download as PPTX, PDF0 likes219 views
The document critiques the use of story points for estimation in agile development, highlighting both the benefits and drawbacks associated with traditional estimation methods. It suggests that data-driven forecasting techniques, such as queuing theory and Little's Law, can provide better predictability and effectiveness than sizing-based estimates. Additionally, it emphasizes the importance of managing cycle time and work-in-progress limits to enhance forecasting accuracy.
![“I was there when they [story points] were invented. I
may actually have invented Points. If I did, I’m sorry
now.”
Ron Jeffries](https://image.slidesharecdn.com/bb-beyond-story-points-180313051309/85/Beyond-Story-Points-Forecasting-with-empirical-data-7-320.jpg)
Beyond Story Points - Forecasting with empirical data
- 1. Beyond Story Points Using empirical data to forecast
- 2. “Treat your user stories like perishable goods” Mary Poppendieck
- 3. Question: why do we use story points?
- 4. The positives of estimation • Requirements discovery • Exploring complexity • Conversations • Scope, trade offs • Scheduling • Others?
- 5. The negatives of estimation • Time and effort • Loss of confidence in the process when they are wrong • Developing to the estimate, not the task • Pressure on team driving down quality • Down stream effects on dependencies • Constant revision when understanding grows • Others?
- 6. There’s also the PLANNING FALLACY
- 7. “I was there when they [story points] were invented. I may actually have invented Points. If I did, I’m sorry now.” Ron Jeffries
- 8. The positives of estimation All benefits can be gained in other ways without traditional sizing- based estimation
- 9. Predictability can make estimating unnecessary. We can replace sizing-estimate based forecasting with data-driven forecasting.
- 10. Firstly, some key definitions: • Cycle time • Throughput • Work-in-progress (WIP) limits • Planning • Forecasting
- 11. Planning and forecasting with data Using concepts such as cycle time and WIP limits we can use queuing theory and Little’s Law to forecast completion dates. Often with far more success than traditional estimation methods. Little’s Law is one technique but there are others (eg Monte Carlo)
- 12. Little’s Law Essentially, use average cycle time and WIP limits to derive a timeframe to deliver a set of stories. Throughput = WIP / CT Time = # Stories / (WIP / CT)
- 13. But isn’t that estimating?
- 14. Prerequisites • Understanding your work-in-progress limits • Understanding (and reducing) your cycle time variance • Understanding your system of work
- 15. For those who use physical walls, you don’t need digital walls for this
- 16. Cycle time!
- 17. The key is to keep cycle time as consistent as possible… …so how do we do that?
- 18. Reducing CT Variance • Get your rhythm first
- 19. Reducing CT Variance • Get your rhythm first • Slice your user stories continuously and consistently
- 22. Reducing CT Variance • Get your rhythm first • Slice your user stories continuously and consistently • User story kick offs
- 23. Reducing CT Variance • Get your rhythm first • Slice your user stories continuously and consistently • User story kick offs • Visualise cycle time on your wall
- 25. Reducing CT Variance • Get your rhythm first • Slice your user stories continuously and consistently • User story kick offs • Visualise cycle time on your wall • Use process control charts to retrospect on outliers
- 29. Challenges • Consistent user stories (slicing is not trivial) • Buy in from stakeholders • Cold start problem • Discipline to maintain the conversations • Others?
Editor's Notes
- #3: The more time we invest up front, the more we are invested in a solution and the harder it is to change.
- #4: Prompts: Costing. Project resourcing (time, money, people). Iteration planning. When I gave this talk for the first time, someone yelled out “because we’re told to!”
- #5: When we estimate, the first thing we ask is “what’s in scope for this user story?” We start to talk about requirements. Testing. Edge cases. These are all valuable conversations to have.
- #6: Then these estimated stories sit in the backlog, get stale and out of date, do we estimate them AGAIN?
- #7: a phenomenon in which predictions about how much time will be needed to complete a future task display an optimism bias and underestimate the time needed
- #11: Cycle time – the time taken to move a “unit of work” (eg a user story) from the backlog to done. For our context, we will track this as starting when the user story moves into the first build column (eg in development, doing) until moving to done. Note that lead time is slightly different but often used interchangeably. Throughput – the number of things done in a given time period. We are mainly concerned with “user stories per week”. WIP limits – capping the amount of work in any step in your workflow. This is a kanban concept. Planning – discussing what work will be done, in what order. Independent of WHEN. Forecasting – a prediction of when a thing may be completed (at multiple levels)
- #12: We can put metrics around how long it takes the average user story to move through the queue. There are mathematical theories that allow us to use known data to predict this duration.
- #13: It seems really simple when you look at it mathematically. We know in reality it doesn’t always fit within a nicely defined formula, but we can do things to ensure it does as much as humanly possible. The formula also allows us to adjust variables to determine the effect on the plan. For example, calculate an upper bound on your cycle time to reflect the “cone of uncertainty”
- #14: Kind of…. At a backlog level, not story level Constantly revised when the backlog changes, or average cycle time changes
- #16: Dot the card or date the card
- #17: Dot the card or date the card
- #19: Don’t even think about it yet. Start to get a feel for what “normal” feels like for your team
- #20: Continuously – at refinement, at planning, at card kick off Consistently – team will eventuallly develop muscle memory for this, to help use some kind of heuristic eg x-acceptance tests per story Techniques and exercises – hamburger method, carpaccio exercise, probing questions
- #23: Help with keeping story size consistent
- #24: As per next image. This will help the team to self correct.
- #26: Process control chart is an ugly sounding word for a very powerful thing
- #27: Explain the difference between this and the one in JIRA.
- #28: Show images of the wall and explain the done column
- #29: This is how I have used the forecasting data in the past