The intervention
Name the exact action. “Optimized the funnel” conceals the work; “removed two required checkout fields” explains it.
For people who sell optimization
If you cannot explain the change, you cannot sell the result.
Use your own intervention, numbers, and evidence. The workbench does the arithmetic, grades the claim, and exports what you can actually defend.
The credibility test
Replace the pitch with proof.
No generated expertise. No mystery uplift. Enter what you changed and the numbers you observed; this tool makes the comparison explicit and refuses to call weak evidence causal.
Your inputs
What did you actually do?
Describe the concrete change, not the hoped-for effect.
A claim worth buying
Four parts. Nothing hidden.
The movement
Show baseline and result together. Relative lift alone can make tiny changes look enormous.
The evidence
A controlled comparison can support causality. A before/after screenshot can only establish that two values differ.
The boundary
State where the finding applies and what else could explain it. Credibility grows when uncertainty is explicit.
Pressure-test the language
Three phrases that reveal the quality of the work.
“We changed…”
Can the operator name the intervention with enough precision that another person could reproduce it?
Action“We observed…”
Are the result, baseline, unit, and sample visible, or is the strongest number doing all the persuasion?
Measure“This suggests…”
Does the certainty match the evidence design? Explanation earns trust when it stops exactly where the data stops.
BoundaryThe service is the understanding