Write the question before producing the variants. Keep the comparison interpretable, define the outcome and review rule in advance, and use a suitable experimental setup. A platform delivering more impressions to one ad is not, by itself, a controlled test.
If an existing ad is declining, use the creative fatigue diagnosis guide before choosing what to test. A broken destination and a tired message need different work.
Name the uncertainty.
“Find a winning ad” is an ambition. “Does a product-use opening produce a different qualified-visit rate than a presenter opening?” is a testable question. State what changes, why you think it might matter, and which outcome would help you decide what to create next.
Choose an outcome that is close enough to the business goal to be useful and frequent enough to measure. Views can help diagnose attention, but they are not purchases. Define a landing-page view, lead or qualified signup consistently before comparing results.
Turn competitor research into a testable idea.
A long-running ad is worth investigating. Its age alone does not establish sales, profitability, or whether the same approach will work for your product. Treat it as a source of questions before spending money on a new creative.
Record the source URL or library ID, the date you inspected it, the advertiser, and the country and filters you used. Separate the message you can actually see from your interpretation. A product demonstration is observable; “this is their best-converting ad” usually requires evidence you do not have.
Meta’s Ad Library is a starting point for inspecting public creative. Its official disclosure documentation distinguishes information by region and ad category: some listings include additional estimated reach, impressions, or spending ranges. Check the fields available for the specific ad. Those disclosures still do not establish your competitor’s profit or your likely conversion rate.
A worked observation-to-brief example.
Fictional coffee example, not an observation of a named advertiser or a measured campaign result.
| Step | Research note |
|---|---|
| Observation | The opening shows a brewer and asks which grind to choose. |
| Unknown | Sales, spend, audience mix, attribution, and why the advertiser kept the creative active. |
| Hypothesis | Explaining grind choice may reduce uncertainty for our first-time buyers. |
| Original brief | Film our own product with our own verified grind guidance. Change the opening while keeping the offer and destination stable. |
| Decision | Use the planned outcome and guardrail from the test below. Record an inconclusive result when the evidence is insufficient. |
Look for a customer problem you can address, rather than copying the competitor’s script, imagery, or claims. If the idea depends on a feature your product does not offer, reject it. If it fits, combine it with your own customer questions before commissioning variants.
Download the competitor creative research worksheet (CSV). Keep source notes and unknowns alongside the hypothesis, then use the testing plan to define how you will evaluate the original creative.
Keep the comparison interpretable.
For a focused creative question, change one planned dimension and keep the other relevant elements stable: offer, destination, audience definition, measurement and the remaining creative. If the whole concept changes, label it a concept comparison; you will not know which individual change produced the difference.
Use the platform’s appropriate experiment tools when you need a controlled comparison. Two ads running in the same account may receive different delivery, placements or audience mixes. Report that limitation when interpreting ordinary campaign observations.
Google’s experiment guidance recommends focusing on one variable and choosing a metric for the decision. Setup and availability depend on the campaign type. Read Google’s campaign experiment guidance.
A worked coffee example.
| Plan field | Illustrative decision |
|---|---|
| Question | Does an opening about choosing a grind help shoppers understand the offer better than a general morning-routine opening? |
| Control | Morning-routine opening. |
| Variant | Grind-choice opening. |
| Kept stable | Product, offer, destination, remaining edit and measurement definition. |
| Primary outcome | A pre-defined qualified action on the destination, if tracking and volume support it. |
| Review rule | Use a planned evaluation window and the experiment’s uncertainty reporting; do not choose a winner after a few early events. |
If you change the voice, music, product shot and offer along with the opening, the result answers a broader question. That may be a useful test, but record it honestly.
Record enough context to interpret it.
Save the creative versions, dates, spend, event counts, measurement definition and any tracking changes. Record the market, currency and placements. Note interruptions, learning periods, promotions or stock issues that might affect the result.
Define a guardrail as well as the primary outcome. For example, a higher volume of low-quality leads may be undesirable even if the cost per lead falls. The appropriate guardrail depends on your business and the goal of the campaign.
There is no universal number of days or conversions that makes every creative test conclusive. Required sample size depends on baseline performance, variability, the effect you want to detect and the experiment design. Use the experiment’s statistical guidance or qualified analytical help for consequential decisions.
Keep “inconclusive” as a real outcome.
A useful result can be positive, negative, or uncertain. If there is insufficient evidence, record that rather than declaring the cheaper early result a winner. Decide whether to continue, revise the hypothesis, or stop because the test is not worth the additional cost.
When a result is useful, translate it into a specific next brief. “The product explanation deserves another test” is more actionable than “make more good ads.” Keep the original comparison available so later tests can be interpreted in context.
Build the plan before producing variants.
The ad creative testing worksheet captures the hypothesis, control, variant, outcome definition, guardrails and review rule. It is a planning tool, not a statistical calculator or a campaign launcher.
Aitoma can help prepare brand-informed creative for your review. The testing plan determines which versions are worth making and what you will learn from them.