Forecast confidence
Also known as Delivery forecast confidence, Forecast certainty
Definition
Forecast confidence expresses how strongly available evidence supports a predicted delivery outcome. It should reflect variation, sample quality, work definition, and changes that could invalidate historical patterns.
Confidence is conditional
A forecast is credible when its population, workflow boundaries, and uncertainty are visible. Confidence should fall when work is novel, the sample is small, the process changed, or a known dependency has no reliable estimate. A precise date can still be poorly supported.
A concrete example
An illustrative team has completed 30 similar changes with a median of four days and a long tail of ten. It can say that a new, comparable item is likely to finish within a range, but it should lower confidence if the new item depends on a system the team has never changed.
Show the evidence
Report the historical window, percentile or probability used, item class, and exceptions. Revisit confidence as new information arrives. A range with clear assumptions is more useful than a single date presented as certainty.
Limitations
Confidence is not the same as probability unless a defined statistical method supports it. Teams also differ in how they use the label. Prefer explicit probabilities or ranges when the decision depends on risk.
How this relates to Weave
Weave can help teams ground forecast discussions in observed delivery timing and the changes behind the distribution. It cannot guarantee an outcome or know every future dependency, so confidence should remain an explicit statement of assumptions rather than a product score.
Explore Engineering intelligence