Using Confidence Levels in Estimation
Should we add confidence levels to estimates? Like "8 points, 70% confidence" vs "5 points, 95% confidence"? Or does this complicate things too much?
Use confidence implicitly through estimate padding. Low confidence = estimate higher. We teach: if you'd bet $100 on completing in X points, that's your estimate. Uncertainty is built in.
Adding explicit confidence percentages just creates analysis paralysis. Keep it simple - the Fibonacci gaps already encode uncertainty.
We DO use confidence and it helps. Stories get tagged: HIGH (well understood, simple), MEDIUM (some unknowns), LOW (many unknowns). Sprint planning pulls HIGH confidence stories first. LOW confidence stories become spikes. Makes uncertainty visible to PO.
Three-point estimation: optimistic/likely/pessimistic. "3-5-13" means best case 3, likely 5, worst 13. We estimate the "likely" and track actual vs all three. Gives stakeholders realistic ranges instead of false precision.
Spike first if confidence is low. Don't estimate high-uncertainty stories, timebox a spike to reduce uncertainty. 4-hour spike, then estimate. Way better than wild-ass guessing and calling it "13 points low confidence".