Epic Estimation 2019
Download as PPTX, PDF3 likes1,088 views
The document presents an agile approach for estimating epics, incorporating techniques like affinity mapping and t-shirt sizing to improve product backlog burnup and manage demand effectively. It emphasizes the principle of not estimating epics unless necessary and suggests relative estimation using geometric series instead of traditional point systems. Additionally, the methodology allows for adjustments based on historical performance and team velocity, aiming to streamline planning processes and enhance story mapping.
Epic Estimation 2019
- 1. Epic Estimation An Agile Approach for Estimating Epics David Hanson November 2019 Version 2.1
- 2. An Agile Approach for Estimating Epics This presentation incorporates the various Agile approaches often mentioned for epic estimation, like affinity mapping and t-shirt sizing, with one novel concept, such that epic estimation can be used to create an effective product backlog burnup, helping to guide decisions for number of teams or anticipated count of sprints, and creating an effective tool for demand management.
- 3. Epic Estimation Outline Don’t Estimate Affinity Sizing Outliers Relative Estimation Spot Check Yesterday’s Weather Adjustments Analysis & Research Backlog Burnup Conclusion
- 4. Don’t Estimate Epics Unless Required Never Accurate Early estimates are not accurate Precise estimation is expensive Better to invest in story mapping Better to start and learn Sometimes Useful How many teams? How many sprints or releases? Helps flush out risk Demand management
- 5. Affinity Sizing Review a dozen or more epics with a half-dozen members Group them roughly on a scale of smaller to larger smaller larger
- 6. T-Shirt Sizing Divide them into five relative groups: XS, S, M, L, XL smaller larger XS S M L XL
- 7. Outliers XXS: convert to story XXL: split into multiple epics smaller XS S M L XL larger
- 8. Relative Estimation Instead of estimating points, estimate number of stories per epic Instead of Fibonacci series start with geometric series XS: 2 stories S: 4 stories M: 8 stories L: 16 stories XL: 32 stories Use median story point size to convert to points (1, 2, 3, 5, 8, 13, 21) XS: 10 points S: 20 points M: 40 points L: 80 points XL: 160 points smaller XS S M L XL larger 2 stories 10 points 4 stories 20 points 8 stories 40 points 16 stories 80 points 32 stories 160 points
- 9. Spot Check & Adjust New or unsure Pick an epic from each size and break down into stories Still unsure point the new stories Use this learning to make adjustments and regroup smaller XS S M L XL larger 2 stories 10 points 4 stories 20 points 8 stories 40 points 16 stories 80 points 32 stories 160 points
- 10. Yesterday’s Weather As epics are replaced with stories, replace the simplified counts with actual counts As stories are pointed, replace the estimated 5 points per story with average story points If small averages 6 stories and stories average 4 points then use 6 stories or 24 points If medium averages 12 stories and stories average 4 points then use 12 stories or 48 points Note: adjustments may be warranted if sizes converging or diverging (e.g., L & XL above) smaller XS S M L XL larger 3 stories 12 points 6 stories 24 points 12 stories 48 points 19 stories 72 points 23 stories 92 points
- 11. Historical Adjustments If completed epic has fewer or more stories than relative neighbor then recategorize size If two sizes converging closer than Fibonacci-equivalent then consider regrouping epics to maintain minimum 50% difference (e.g., L = 1.5 * M or L = M + S) If two sizes diverging beyond geometric-equivalent then consider regrouping epics to maintain maximum double size difference (e.g., M = 2 * S or M = S + S) smaller XS S M L XL 16 larger 3 stories 12 points 6 stories 24 points 12 stories 48 points 18 stories 72 points 30 stories 120 points
- 12. Estimating Analysis & Research Not required because velocity adjusts accordingly Getting to Ready Agile methodologies don’t estimate getting stories to ready Agile methodologies assume continuous grooming Teams might spend 10% bandwidth on grooming One PO or analyst per 5 team members (coders & testers) Velocity already accounts for this overhead Research Spikes Only stories are pointed, so how will time for spikes be estimated? Assume research will be continuous or decreasing Alternatively point spikes and timebox when tasking Consider reserving 10% time for learning Average velocity will adjust accordingly
- 13. Product Backlog Burnup Best Practice Recommendations Baseline Simplicity Set total scope using only epics Track burnup using only stories Credit only done stories Map all stories to parent epic Optional Complexity Add stories without parent epic to total scope Replace epic size with total stories or total points when epic closed Use maximum of epic size estimate or actual total stories
- 14. Product Backlog Burnup 0 200 400 600 800 1000 1200 1400 1600 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Epic Scope Benchmark Velocity Story Burnup Scope Increases Scope Decrease Early Finish Major Refactoring Slow Start Accelerating Delivery Stable Delivery
- 15. Integrated Approach for Epic Estimation Rapid estimation technique leveraging several existing Agile concepts, such as affinity sizing, T-shirt sizes, splitting, and yesterday’s weather Instead of estimating epics based on story points with Fibonacci series, estimate epics based on story count with geometric series Refined estimates based on sampling of epics, completed by breaking down stories and pointing, which will be required anyway Can be leveraged with new teams just getting started, and can be adjusted as teams perform work and improve abilities Just enough to set expectations for number of teams and sprints and create simple product backlog burnups useful for demand management Spend less time during planning on estimation, to spend more time on story mapping or design thinking