Lesson 7 of 13 Module 7

A/B Testing

Three different hypotheses and useful measurement.

Join free to unlock all 13 lessons

Join AI Accelerator and get instant access to every coaching module, plus one short AI tip in your inbox each week.

Free. Unsubscribe anytime. Already joined? Unlock here.

Test hypotheses, not cosmetic variations

For a meaningful test, create three different reasons to click. Example A: proof and revenue (“$15K FROM CHATGPT SEARCH”). Example B: AI discovery curiosity (“AI KNOWS MY NAME”). Example C: perceived authority (“THEY THINK YOU'RE FAMOUS”). Three nearly identical blue thumbnails with slightly different font weights are not a useful experiment.

  1. Decide whether you are testing thumbnails only or title-plus-thumbnail combinations.
  2. Upload up to three eligible variations using YouTube Studio's native A/B test on desktop.
  3. Do not change the tested title or thumbnail mid-test unless restarting intentionally.
  4. Let the experiment run long enough to gather useful exposure; results can take days or up to two weeks.
  5. Review the winning variation by watch time, not CTR alone.
  6. Keep a creative log: hypothesis, hook, colors, subject crop, start/end date, result, and next lesson.

Read the numbers correctly

Look at impressions, CTR, watch time, average view duration, and traffic source. Search impressions can behave differently from Browse or Suggested impressions. Strong CTR with short viewing can mean the packaging overpromised. Low CTR may mean the idea is vague, the visual is lost among competitors, or the audience is mismatched.

Search ranking is not guaranteed by thumbnail changes. YouTube weighs query relevance and viewer response. The goal is to earn qualified viewing from an accurate, differentiated promise.

Coach's exercise

Write the hypothesis BEFORE seeing results. For example: “Specific dollar proof will beat generic fame language among personal-brand searchers.” Then compare actual metrics, record the outcome, and choose the next hypothesis.

Questions and comments

Stuck on something? Ask your question here. Brad and the community reply.

Loading comments...