Build the test card before the ad. Put the buyer situation, hypothesis, approved claim, execution change and next decision on one page. That gives production and growth the same question to work from.
The card should be short enough to use during a review and specific enough that someone who missed the meeting can explain the asset.
Use the card to separate evidence from interpretation and production acceptance from result evaluation. Preserve context through the brief, revisions and test review.
Field one: The situation and its source
Describe a relevant buyer in a specific context. Attach the authorized evidence and keep its limits visible. Record whether the finding comes from a review, interview, support question or another inspected source.
Here is a fictional example: a pet-product buyer is uncertain about how to choose the appropriate configuration. The scenario is for planning; it contains no real customer quote or product claim.
The situation gives you a possible communication task. It does not establish how common the concern is or whether the proposed angle will work.
Identify the intended segment and inspect other evidence where available. Preserve the difference between a vivid observation and a validated priority. The card should make that difference easy to see before production starts.
Field two: The hypothesis
Write what the execution is intended to explore. Use language that fits the evidence rather than turning the proposed response into a promised result.
For the invented example, a verified configuration explanation might help the intended audience understand which choice fits its situation. That is a hypothesis about communication, not a claim that the creative will improve acquisition cost.
Name the uncertainty and why it deserves a production slot. If several interpretations are plausible, record the one you are prioritizing and the alternatives that matter.
Avoid retrofitting the hypothesis after launch. An unexpected result can suggest another question, but the record should preserve what the team actually intended to investigate when it made the asset.
Field three: The approved claim and proof
List the statement the product evidence supports, including important conditions. Identify who verified it and where the supporting information can be found.
Then specify what the creative must show or explain. A configuration guide would need accurate product information. A demonstration would need an approved setup. A comparison would need a fair basis and verified labels.
The card should connect the proof to the actual claim. It should not ask an editor to infer product truth from general footage or a competitor reference.
If a necessary answer remains unresolved, assign that dependency. You can change the creative task to fit available evidence, but document the change before production treats the promise as settled.
Field four: The execution change
Describe the opening, sequence, demonstration or format you are changing. Write the intended constants too.
A narrow variation can explore one opening while keeping much of the asset consistent. A broader concept change can explore a different communication treatment. Both can be useful; they support different interpretations.
Record practical production requirements. State which assets exist, which scenes need recording and what the reviewer must inspect.
Keep the test card linked to the brief. When the execution changes during revision, update the record if the change affects the hypothesis or claim. The card is useful only if it describes the version that actually enters distribution, rather than an earlier plan that nobody follows anymore.
Field five: The next decision and its limits
Agree the objective and relevant evidence with the responsible growth team. Record the delivery, offer, audience and measurement context needed for review.
Specify what could leave the result inconclusive. There is no universal threshold supplied by this framework. The criteria need to fit the account and question.
Name the possible actions: continue, revise, stop or investigate. Explain what each would mean for this test. A practical stop may reflect resources rather than a demonstrated concept failure.
I want this connection between research and the next creative decision to remain visible. The test card organizes the work; it does not guarantee what the eventual result will be.
Run a handoff check before launch
Ask production to explain the intended piece from the card and brief. Ask growth to explain the question and review criteria. Ask the relevant reviewer to verify the claim and proof.
Resolve differences while the asset is still easy to change. Keep acceptance focused on whether the execution meets the approved assignment. Evaluate performance after distribution with the appropriate context.
Then write the result in three parts: observation, interpretation and next action. Link the exact asset and preserve uncertainty. The next strategist should be able to recover the reasoning without repeating the original meeting or inventing a new explanation for why the ad existed.
A worked example: make the handoff concrete
Consider a fictional buyer question about the size of a storage container. The research supports uncertainty about dimensions; the product specification supports the published measurements. Those are separate evidence items. The card should not convert the research into a claim that a particular demonstration will improve conversion. That is a question for the proposed execution and its later evidence.
The hypothesis might be that showing the verified measurements beside an appropriate reference will make the sizing explanation easier to inspect. The production instruction identifies the approved specification, measurement setup, shot and qualified wording. The decision field says what the team will consider after distribution and what missing measurement or limited delivery would leave unresolved.
| Decision field | Practical instruction | Check before handoff |
|---|---|---|
| Situation and source | Who has the problem, and where was it observed? | Link the authorized research |
| Hypothesis | What should this execution help you investigate? | Keep it distinct from a fact |
| Claim and proof | What may the ad say and show? | Attach approved wording and evidence |
| Execution and decision | What changes, and what happens after review? | Name limits and the next owner |
Try the card at the production handoff
Give the card and its linked evidence to the person creating the asset. Ask them to explain the buyer question, permitted claim, proof scene and review decision in their own words. Any answer that depends on a meeting nobody recorded identifies an incomplete field. Resolve that field with its owner before multiplying versions.
Keep the accepted card beside the final asset. If the proof scene changes, update the card and obtain the relevant review. If the offer changes, update the comparison context. The card is a working agreement about the question and execution; it stays useful only when it describes the asset that actually reaches the test.
Current reference: record the actual experiment conditions
Google Ads’ current custom-experiment documentation explains how a trial compares with an original campaign and how traffic and budget allocation form part of setup. It also warns that changes to the original do not automatically carry into the experiment. That supports recording the actual comparison context. The test-card and review methods here are original operating guidance, not a universal Google Ads testing recipe or a claim about another platform. Read Google Ads’ custom-experiment guidance (checked October 16, 2026).
Expert Q&A
Does every card need a new format?
No. A focused opening or proof variation can have a clear test question. State what changes and what remains consistent. A new format is appropriate when it serves a distinct communication job and the available evidence can support it. The card should explain the reason for the choice.
What if the product proof is not ready?
Record the missing evidence and name its owner before production begins. A customer concern can justify investigating a question, but it does not verify the product answer. Either obtain approved proof, narrow the claim to what existing evidence supports, or hold that execution until the required input is available.
What if the card changes during production?
Update the affected fields and obtain the relevant review. A change in proof, message or claim may change the hypothesis. Preserve the version history so the result can be linked to what actually ran. Do not evaluate a revised execution as if the original plan remained unchanged.
Who should maintain the card?
Assign an owner who can connect strategy, production and growth input. Product or qualified claim review keeps its own authority. The owner should preserve the exact asset, version and next action. Shared access helps collaboration but does not replace accountability for a usable record.
Write a five-field card for the next ad. Have production, growth and the reviewer explain it before launch. Resolve the first different answer while it is still a planning decision.