Put your comparison in numbers.

Compare measured clicks and impressions. See CTR, percentage-point difference and uncertainty intervals.

Enter measured impressions and clicks for two variants. The calculator reports the click-through rate of each one, the B-minus-A difference in percentage points and the relative change when the A rate is nonzero. It uses Wilson 95% intervals for each rate and a Newcombe interval for the difference between independent proportions.

The default result is descriptive. Mark the comparison conditions only if impressions were fairly allocated to independent, comparable groups. A difference interval spanning zero does not separate the rates within this calculation. Even a separated interval does not establish a better video: watch time, audience quality and context remain outside CTR. Before/after changes, uneven audiences and repeated checking can affect interpretation. This is not YouTube’s native thumbnail experiment decision method.

How to use

  1. Enter the actual impressions and clicks for variant A.
  2. Enter matching measurements for variant B.
  3. Mark the comparison conditions only when they apply.
  4. Calculate and inspect the observed rates and difference interval.
  5. Copy the comparison with its method note to your experiment log.

How to use the result

Uses measured impressions/clicks and Wilson/Newcombe intervals. Statistical comparison requires independent, fairly allocated groups.

  1. Confirm the counts cover comparable groups and periods.
  2. Review the uncertainty, not just the larger CTR.
  3. Consider watch time separately from clicks.

Frequently Asked Questions

What is the difference between percent and percentage points?

A change from 4% CTR to 5% is one percentage point and a 25% relative increase. The calculator reports both. When A has zero clicks, the relative percentage is undefined and is shown as unavailable instead of infinity.

Does a higher CTR prove a winner?

It shows more measured clicks per impression in the entered data. The interval and comparison conditions help judge uncertainty, but CTR alone does not measure watch time or video quality. Different audiences and periods can also explain a difference.

What assumptions do the intervals use?

They use a binomial proportion model and assume independent, fairly allocated observations for the A/B difference. Without comparable groups, treat the result as a descriptive calculation. Repeated peeking and choosing when to stop can invalidate a simple fixed-sample interpretation.