Read a Monte Carlo forecast
Metriq forecasts the future from your team's measured past, not from anyone's guess. It replays thousands of simulated futures using your real throughput history and reports the spread of outcomes. This guide reads that forecast.
Two questions, two modes:
- When? — given a number of items, when will they be done?
- How many? — given a date, how many items will be finished by then?
For the reasoning behind forecasting instead of estimating, see How forecasting replaces estimation. For the technique itself, see Monte Carlo.
Before you can forecast
A forecast needs at least 4 weeks of completed work as throughput history. With less, Metriq refuses rather than guess: you will see "Not enough data for forecast. Need at least 4 weeks of throughput history." That refusal is the honest answer — finish more work and come back.
Read a "when" forecast
A "when" forecast answers "when will these N items be done?" with four dates, not one:
| Line | Meaning |
|---|---|
| 50% likely by | A coin flip. Real life beats this date half the time. |
| 70% likely by | More likely than not. |
| 85% likely by | The date to commit to. Conservative enough to usually hold. |
| 95% likely by | Near-certain, and the latest. |
The card highlights the 85% likely by date because that is the one worth promising. The earlier dates are optimistic; the 95% date is your safety margin. Quote a single date and you are hiding the uncertainty — the whole point of the spread is to show it.
The footnote tells you what the forecast was built from: how many weeks of data, how many items remain, and how many simulations ran. More weeks of data means a more trustworthy spread.
Read a "how many" forecast
A "how many" forecast flips the question: "by this date, how many items will we finish?" The same percentiles apply, now as item counts:
- p50 — you'll finish at least this many half the time.
- p85 — the conservative count: you'll almost always finish at least this many. Plan scope around it.
- p95 — the floor you can nearly guarantee.
Read it the safe way round: promise the p85 count, not the p50 count. The lower the count, the more confident the forecast.
Interpret the spread
The gap between the optimistic (p50) and conservative (p95) numbers is the real signal:
- A narrow spread — your flow is consistent, so the future is predictable. You can commit tightly.
- A wide spread — your throughput swings week to week, so any single promise is shaky. Tighten the spread by reducing variability (see the cycle-time scatterplot), not by picking the optimistic number and hoping.
A Monte Carlo forecast never gives you one number because the future is not one number. It hands you the odds; you choose how much certainty to buy.
Forecasts vs. your SLE
Your SLE is a standing promise about a single item's cycle time. A forecast is a batch projection about many items finishing. They are complementary: a healthy SLE is a good sign your forecasts will be tight, because both are driven by the same consistency in your flow.