Loading page
About this tool
Define a metric by numerator and denominator, check its 95% margin of error and sample size, then build the AI prompt that documents it.
The KPI Definition Prompt Builder takes a metric's numerator, denominator, type and reporting grain, evaluates a worked example, and — for percentage metrics — computes the 95% margin of error using the normal approximation z × √(p(1−p)/n) along with the denominator you would need to hit a target margin. It then writes an AI prompt that turns all of it into a definition with the edge cases decided rather than caveated. For analysts and PMs tired of two dashboards disagreeing about the same number.
Open KPI Definition Prompt Builder on AltFTool — it loads instantly in your browser.
Add your input to the workspace.
Adjust the options until the result looks right.
Copy or download the output and put it to work.
The definition starts from what is counted and what it is divided by, not from a dashboard name.
The 95% interval and required sample size show whether a movement is real before anyone reacts to it.
Ten common edge cases can be attached, and the prompt demands a rule for each, not a footnote.
State the numerator and denominator as specific events with their filters and time window, name the grain and direction, then decide every edge case as a rule. A definition is finished when two engineers reading it separately would write the same query.
For a proportion, the 95% margin of error is 1.96 × √(p(1−p)/n), expressed in percentage points. At p = 0.258 and n = 4,800 that is about ±1.24 points, so anything smaller than a 2.5-point swing is inside the noise.
Rearranging the formula, n = z²p(1−p)/e². To measure a 25.8% rate to within ±2 percentage points at 95% confidence you need roughly 1,841 in the denominator. The normal approximation also needs at least five in each outcome group before it is trustworthy at all.
Any metric with a denominator someone can shrink, or a numerator someone can trigger without creating value. The prompt asks explicitly for three gaming routes and the guardrail metric that would catch each, because pairing a goal metric with a guardrail is what stops the optimisation going sideways.