Skip to main content
FHAVI

Free resource

Creative tests that produce answers, not screenshots

Most creative testing produces an opinion. Two ads run for four days at different budgets, one wins, nobody knows whether it won because of the hook, the audience, or the day of the week — and the result is forgotten in a fortnight. The fix is not more creative. It is one variable per test and a rule written down before the test starts.

Design

One test, one variable

01

Before the test starts, decide

  1. The variable. Hook, visual, format, offer, or audience — exactly one, or the result explains nothing.
  2. The metric that decides it: cost per add-to-cart, cost per purchase, or thumb-stop rate. Choose one; a test with three possible winners has none.
  3. The budget and the duration, both fixed before launch. Changing either mid-test invalidates it.
  4. The winner rule, in numbers: 'variant B needs 20% lower cost per ATC over seven days to replace A'.
  5. The minimum volume below which you will call it inconclusive rather than guess. Say the number out loud; it prevents false confidence.

The log

Columns for the record

02

One row per variant

  1. Test id and date range.
  2. Variable under test and what was held constant.
  3. Variant description in one line a stranger could understand.
  4. Spend, impressions, clicks, add-to-carts, purchases, revenue.
  5. Derived: CTR, cost per ATC, cost per purchase, ROAS — computed, not typed.
  6. Verdict against the pre-written rule: won, lost, inconclusive.
  7. What you will do differently next time. One sentence; it is the only part that compounds.

Traps

The four ways creative tests lie

03

Check for these before believing a result

  1. Different audiences or placements. You tested the creative and the delivery at the same time.
  2. Learning phase interference — a variant that never exited learning has no reliable cost per purchase.
  3. Small numbers dressed as percentages. Forty clicks does not produce a 30% difference; it produces noise.
  4. Seasonality: a Friday-to-Sunday test against a Monday-to-Wednesday control measures the calendar, not the creative.
04

What to do with a winner

  1. Replace the loser only after the winner has held for a second budget level; some creatives only work cheap.
  2. Write the winning hook pattern into the log's summary — patterns across tests are worth more than any single result.
  3. If three tests in a row are inconclusive, the problem is usually volume, not creative.

Keep going

The rest of the library

Every checklist here is free and ungated. The research pages carry the reasoning behind them.

All resources · Research library

Testing without a winner rule?

Send your last four creative tests and their spend. We will tell you which produced a real answer and which produced noise.

Prefer email? Write to md@fhavi.com