Build a creative testing matrix around arguments, proof and format

A customer obstacle is the starting point

Choose a customer decision to understand. People may hesitate because they doubt suitability, worry about implementation or cannot justify the price. Write each obstacle as a question without pretending it describes every member of the audience. For each question, propose an argument the business can substantiate. 'Easy to start' needs an honest explanation of the required steps; 'lower operating cost' needs a documented comparison with its conditions.

Separate argument, proof and format into different columns. An argument is the reason to consider the offer. Proof is the demonstration, specification or other verified material that supports it. Format is how the evidence is communicated: a demonstration video, a static comparison or a sequence of images. Changing format and argument simultaneously can be a valid exploration, but the result will not identify which component mattered.

Educational example: a fictional scheduling product has three proposed concepts. One addresses setup effort with a recording of the actual setup process. Another addresses coordination with a clearly labelled example workflow. A third explains price scope with a verified inclusion list. The team chooses to compare the first two arguments using the same format, then examine another format for the more promising argument. This stages the questions; it does not assume either concept will win.

Record the unit being evaluated. A named creative can be reused in several ads, while delivery mixes audiences, placements, text and destinations. An observed asset ranking is therefore not automatically a controlled creative test. Google documents experiment options for comparing original and test campaigns; check the applicable design and account capabilities if a platform experiment is appropriate. A production matrix alone provides no randomisation.

Three rows that make production useful

• Argument row: a specific customer question, the proposed answer and what would make that answer useful. It gives writers a stronger brief than 'make this more engaging'.
• Evidence row: source material, owner approval, limitations and expiration. It keeps unsupported claims from acquiring credibility through polished production.
• Learning row: the planned comparison, primary business outcome and alternative explanations. It determines which new piece of content is worth producing next.

Stage the questions before creating assets

1. Collect non-identifying themes from sales questions, product feedback and aggregate search tasks. Separate direct evidence from the team's assumptions about motivation.
2. Select a small set of genuinely different arguments. Avoid turning synonyms of one promise into supposedly independent concepts.
3. Attach proof that operations or product owners can verify. If the evidence is unavailable, change the argument or label the concept as requiring research.
4. Choose which dimension to compare first. Keep offer, destination and outcome definition stable where the question requires it, and log unavoidable differences.
5. Define evaluation dates, spending boundaries for any later authorised test and what counts as inconclusive. Do not promise that a fixed number of days will identify a winner.
6. After evaluation, store the conclusion at the level the evidence supports: an observed association, a test result or an unresolved question. Update the next brief rather than repeatedly reshooting the same untested assumption.

A matrix is not a controlled experiment

Sources and further reading

• Google Ads: experiment options and comparisons