How to set a go/no-go threshold before you're emotionally attached to the answer
If you decide what counts as success after you see the number, it isn't a test anymore. Here's how to set the bar in advance.
Jul 28, 2026 · 5 min · KPIndicator
The single most common failure mode in demand testing isn't a bad test. It's a good test with no threshold set beforehand, so a mediocre result gets reframed as encouraging after the fact.
If "success" gets defined once you already know the number, you're not running a test — you're building a case for a decision you'd already made. Here's the fix.
Set the number before you see any data
Before a single visitor hits the page, write down, specifically:
- What visitor-to-lead conversion rate counts as a real signal (not "good," a number — e.g. "above 4%")
- What cost per qualified lead is acceptable given your eventual unit economics
- For anything past a landing page test, what commitment level counts — email signup, waitlist, refundable deposit, booked call
- How much traffic or how many days constitute enough sample to trust the number at all
Put it in writing, share it with whoever else has a stake in the answer, and don't revisit it once data starts coming in.
Anchor the threshold to something real, not a round number
"5% feels like a good number" is not a threshold, it's a guess wearing a threshold's clothes. Anchor it to something with teeth:
- Category benchmarks — what does a decent landing page in your space typically convert at, from ad platforms' own reported ranges or comparable public case studies
- Your own economics — if your product needs a $150 CAC to be viable and your test is producing $400 leads, that's not close, regardless of how it feels
- A control — if you're testing more than one idea or angle, the relative ranking is often more trustworthy than any single idea's absolute number
Decide what "inconclusive" looks like too
Binary go/no-go thinking breaks down when a result lands in a genuine gray zone — not clearly good, not clearly bad. Decide in advance what that zone is and what you'll do if you land in it: usually a short, cheaper follow-up test isolating the single biggest confound (channel quality, one weak variable, a small sample), not an instant pivot to gut feel.
The uncomfortable part
Setting a real threshold means accepting, before you know the answer, that the answer might be no — and that a no is a valid, useful, complete result. If that's uncomfortable, that discomfort is informative. It usually means the idea has a sponsor whose judgment is more attached to the outcome than the evidence is.