Sprint Backlog
Completed in the sample
- Coherent account-count dataset by period and segment
- Four documented calculated rates and source-count display
- Exact current/prior cohort and observation dates
- Percentage-point comparison chart
- Saved scoped insights, empty state, storage warning, and reset
- CSV export and copied stakeholder summary
- Sample observation-age label and impossible-count validation
- Explicit no-data and action-error states
- Hash navigation, case study, responsive layout, and logic/browser tests
Prioritized next investments
1. Test comprehension of the existing metric contract
Decision supported: whether the visible numerator, denominator, dates, and unavailable state are sufficient for an operating review. Evidence needed: five fictional-data sessions from the discovery plan, including successful explanation of a rate, its observation window, and the Early access empty state. Dependency: consenting participants and an accessible test script. Defer any new visualization until this evidence exists.
2. Test whether the existing aggregation explanation resolves confusion
Decision supported: whether inline segment detail is worth its visual cost. The PRD now includes the exact 313/452 weighted example; test that explanation first. Evidence needed: repeated failure to reconstruct All from the available counts, with case and assistance recorded. Dependency: the first comprehension study. Prefer improving the explanation over adding a new visualization if it resolves the problem.
3. Establish definition ownership only with a real governance context
Decision supported: whether the prototype could move beyond a fictional, single-owner sample. Evidence needed: an accountable data and product owner, agreed change-review process, and production data-contract requirements. Dependency: a governed environment; therefore no prototype metadata feature is scheduled.
Explicit deferrals
Confidence intervals, alerts, a larger KPI catalog, live data connections, and automated priority scoring are deferred. They would imply provenance, operating ownership, or causal confidence that this portfolio sample does not have.
Admission and closure rules
A candidate enters implementation only with a recorded problem observation, a specific acceptance scenario, and a named proposed decision owner. The PM compares the smallest explanation change, a workflow change, and leaving the sample alone. No staffing, sprint dates, or engineering capacity is implied.
Close a candidate when its acceptance scenario passes software checks and, where it concerns comprehension, fresh participant evidence addresses the original confusion. Passing the current automated suite cannot close the human-comprehension question. If the study shows no recurring need beyond the existing spreadsheet/review process, stop product expansion rather than filling the backlog. See risk response priorities.