Lesson 21 of 22

The wall4 minWe were wrong

Views scored blind to who saw them

One account was 77% Algeria. We'd been comparing its numbers to everyone else's.

We judged content by views for a long time before checking who those views were. When we finally looked at audience composition per account, the numbers were not comparable at all.

  • One account: 77% of its audience in Algeria.
  • Another: 14% US.
  • Another: 62% of viewers aged 35+.

All of them sitting at roughly 98% For You traffic — so by the usual health metrics they looked fine. A thousand views from a mismatched audience and a thousand views from your actual buyers are the same number and completely different outcomes.

Why this compounds

This is the same confound as the deck-versus-video error, one level down. If accounts differ this much in WHO sees them, then any comparison across accounts is measuring audience composition as much as content. We'd made a lot of those comparisons.

The practical version: before you conclude a piece of content underperformed, check whether it was shown to anyone who could plausibly want it. Frequently the content was fine and the audience was wrong — and that is a distribution problem, not a creative one.

Takeaway

A view is not a view. Check audience composition before concluding anything about content.

Every law on one page

The compiled reference — framing laws, the negative prompt block and the pre-ship checklist. One email, no sequence you can't leave.

The pack, and a note when a new teardown goes up. Unsubscribe whenever.