# Your Next Three Useful Ad Tests

Use this after reviewing creator content or before commissioning another batch. The output is a short test plan that connects a customer question, a creative change and the next decision. It does not supply universal budgets or winner thresholds.

## Worked example

**Fictional app:** a meal-planning feature is being demonstrated to busy households. No customer result or ad-library performance is assumed. These are proposed tests, not winning ads.

| Question | Proposed concept or change | What to keep comparable where practical | What to observe | Interpretation limit / next action |
|---|---|---|---|---|
| Does the buyer understand the problem? | Concept A: show the repeated planning task before the shortcut | Offer, destination, conversion event and relevant delivery conditions | Agreed business outcome plus attention and task-understanding diagnostics | If promising, inspect whether screen detail is clear; do not credit the opening alone if the whole package changed |
| Can viewers follow the demonstration? | Variation within A: clearer interface close-up and narration | Core argument and intended task; record other changes | Comprehension feedback and observed campaign response | If the feature remains unclear, revise the demonstration before changing the creator |
| Is the message useful but the offer weak? | Package test: same use case with a revised offer | Record offer change explicitly; keep destination/event consistent where appropriate | Business outcome and full acquisition cost | Treat as package evidence, not a pure creative test; investigate offer/destination before another production batch |

These tests need not run simultaneously. Choose the most consequential uncertainty that your available data can help answer. The right next step can be research or a product/offer correction rather than another ad.

## Write one test before expanding to three

| Planning field | Test 1 | Test 2 | Test 3 |
|---|---|---|---|
| Buyer and situation |  |  |  |
| Decision / uncertainty |  |  |  |
| Concept or variation? |  |  |  |
| Exact demonstration / change |  |  |  |
| Supported claims and sources |  |  |  |
| Creator skill required / work sample |  |  |  |
| Comparison and known differences |  |  |  |
| Primary outcome and diagnostic signals |  |  |  |
| Observation and review plan |  |  |  |
| Full cost scope |  |  |  |
| Decision criteria and responsible reviewer |  |  |  |

## Review without inventing certainty

| Observed result | Evidence source and period | Alternative explanation / missing data | Decision | Next brief |
|---|---|---|---|---|
|  |  |  | Retest / revise / stop / investigate |  |
|  |  |  |  |  |
|  |  |  |  |  |

Choose decision criteria using your economics, current baseline, data and test design. If several inputs change, record the result at package level. A library ad's presence or longevity is not proof of profit. Low data volume may justify uncertainty rather than a confident winner.

AI can help organise source material and propose alternatives. A person still checks product claims, whether the proposed change answers the question and what the result supports.

**Your final output:** “We need to learn ___. We will change ___ and record ___. The comparison cannot establish ___. After review, we will ___.”
