Estimation & Planning — Forecasting in a Fog

INVEST heuristic: stories should be Independent, Negotiable, Valuable, Estimable, Small, Testable.

Acceptance criteria live alongside (often "Given / When / Then" — Gherkin format).

The point isn't the number — it's the conversation that surfaces hidden complexity. If a senior says 3 and a newer dev says 13, you've found a gap in shared understanding before the work starts.

• Definition — story points completed per sprint, averaged over a few sprints.
• Use it for — capacity planning, longer-range forecasting, rough sprint loading.
• Don't use it for — comparing teams, performance reviews, KPI dashboards.
• Goodhart's Law: "When a measure becomes a target, it ceases to be a good measure." Velocity-as-KPI = points inflation.
• Range, not single number. Forecast "5–8 sprints" with confidence bands, not "exactly 6."

• At project start, an estimate has roughly 4× spread in either direction (0.25× to 4× actual).
• By detailed design, the spread shrinks to about 1.25×.
• Early "estimates" are guesses with extra steps. Use ranges; re-estimate as you learn.
• The Three-Point Estimate: best-case, most-likely, worst-case → use a weighted average for forecasts.

• Estimates as commitments — the original sin. They're forecasts; treat them so.
• Estimating to look busy — story-point grooming sessions that produce no decisions are pure waste.
• Big stories. If a story is > 13 points or larger than half a sprint, split it before estimating.
• Ignoring throughput. Historical data beats new estimates — once you have it, lean on it.
• Roadmaps as Gantt charts. Now-Next-Later beats month-by-month for products with real uncertainty.