Most sprint forecasts miss because they're built on the wrong number. The fix isn't a fancier method — it's a different statistic. Forecast at p85, not the mean. Build the tail into the budget.
Cycle times have long tails. A few stories take 5-10× the median because of blocked dependencies, hidden complexity, scope ambiguity. Those don't move the median much, but they wreck individual dates.
Forecast at the mean and 86% of sprints miss. Forecast at p85 and dates start landing.
10-min guide. Why the mean lies, the math, the stakeholder message, four anti-patterns to avoid.
Paste raw cycle times. Get count, mean, p50/p85/p95, min/max, histogram, and a tail-shape diagnosis.
Simulate the sprint 10,000 times. Get the p85 completion date. Defensible numbers for stakeholders.
Once committed, track in-flight. SVG burndown with ideal line, completion %, and pattern diagnosis.
Velocity, throughput, cycle time, WIP — what each tells you, what it doesn't, and which to skip. The Goodhart trap and how to avoid it.
ReadWhen a dip is noise vs signal. Seven real causes in order of probability. The one-meeting planning fix. What NOT to do under pressure.
ReadWhat burndowns actually show, four patterns to read at a glance, and when to ignore the chart and look at cumulative flow instead.
ReadWhat story points actually measure, the Fibonacci convention, and why estimates drift from reality (and what to do about it).
ReadSprintFlint records cycle time per ticket, runs Monte Carlo forecasts on every sprint, and surfaces p85 in dashboards. The toolkit, but built into your sprint workflow.