Principle
High confidence
Commonly Used Metrics Are Poor Predictors of Business Impact
3 out of the 5 most commonly used test metrics have the lowest correlation with actual business outcomes. Search optimization shows 2.3% expected impact but appears in only 1.3% of tests — a missed opportunity.
Hypothesis
Teams that optimize for CTR or bounce rate may be running tests that don't move revenue.
Why This Works
Key Learnings
Viewed 12+ times
Related Entries
Research Finding
Principle
Test Quality Peaks at 1–10 Annual Tests Per Engineer; Drops 87% After 30
testing_strategyany
High confidence
Research Finding
Principle
4+ Variation Tests Win 2.4x More Often With 27% Higher Lift
testing_strategyany
Expected lift: 10%-30%
High confidence
Best Practice
Loser
Ghost / Outline Buttons Lose to Filled Buttons — Consistently
ctaany
Expected lift: -25%-0%
High confidence
Research Finding
Principle
Personalized Experiences Generate 41% Higher Impact Than Generic
personalizationany
Expected lift: 20%-50%
High confidence
Test Idea
Inconclusive
Pattern #39: Product Size References
otherany
Low confidence