Ad Creative Testing Plan + Free Test Log Template
Before comparing ad creatives, write down the question, the one thing you will change and the evidence needed for a decision. This plan helps an ecommerce team distinguish a useful test from a collection of unrelated ad results.
Download the free creative test log
The CSV has planning and readout fields for a two-variant test. It contains blank A and B rows, with no sample performance figures. Open it in a spreadsheet, duplicate the rows for the next test and keep actual platform exports alongside it. No signup is required.
Made by VideoAnt. Adapt it for your own or client work. If you share the original, credit this page; linking is optional.
This is a planning and recordkeeping template. It does not launch ads, assign audiences, calculate significance or collect campaign data.
Define the decision before running ads
Start with a question you can answer: “Does a demonstration opening produce more qualified product-page visits than a problem-led opening for this audience?” Choose the success measure before reviewing results, and make sure it connects to the business action you need.
- State the hypothesis. Name the audience, the changed creative element and the expected direction. Record why you expect it; an expectation is not a result.
- Choose one variable. For a hook test, hold the offer, benefit, body, CTA, duration, placement and destination constant as far as the experiment allows. Log any differences you cannot control.
- Define the primary metric. Record its numerator, denominator, attribution window, event definition and data source. If optimizing for purchases, include spend and purchase tracking; click-through rate alone cannot prove a sales improvement.
- Set a budget, observation window and stopping rules. Use the platform’s experiment guidance and your baseline volume. Decide in advance what counts as useful evidence, a tracking failure or a reason to stop for cost. No single spend or duration threshold fits every campaign.
- Confirm allocation and tracking. Prefer a platform-supported split test with comparable, separate audiences when available. Two ads receiving unequal algorithmic delivery are not automatically a controlled experiment.
- Freeze and log the setup. Save creative versions, campaign settings and the landing page. Record changes to pricing, inventory, audience or delivery during the run.
Google Ads’ experiment guidance recommends isolating variables and choosing success metrics in advance. TikTok’s split-testing documentation also specifies one variable per split test. Available experiment controls depend on the platform and campaign type.
A worked plan: problem-led vs demonstration opening
This is a fictional study design for a foldable desk lamp. No campaign was run and no performance result is implied. Assume the folding claim has been verified and you have approved real footage.
| Plan field | What the team would record |
|---|---|
| Question | Which opening better supports the chosen product-page action for people using a small desk? |
| Variant A | Problem-led opening: “Desk still crowded after work?” |
| Variant B | Demonstration opening: “Watch the lamp arm fold away.” |
| Changed variable | Opening treatment. Define whether that means copy only or the whole opening; changing copy and footage together cannot isolate copy’s effect. |
| Kept constant | Approved footage after the opening, benefit, offer, CTA, total duration, destination, audience and experiment settings |
| Primary measure | Choose and define the product-page action before launch. Fill in the actual event, denominator and attribution window; do not silently substitute clicks for purchases. |
| Window and budget | Set from actual baseline volume, platform guidance and acceptable cost; leave unset until those inputs are available. |
| Readout | Pending actual delivery and tracking data; “inconclusive” remains a valid outcome. |
Build these directions from a product ad brief with an evidence log. For more opening structures, use the free ad hook planner.
Record what happened, including uncertainty
In the CSV, enter spend, impressions, clicks, the primary metric’s numerator and denominator, and the platform’s reported uncertainty where available. Preserve the metric definition and attribution window from the plan. Compare the same reporting period and currency.
- Validate first: check event firing, destination availability, audience overlap, delivery balance and changes logged during the run.
- Assess the primary measure: use the preselected rule and the platform’s experiment readout where available. An early rate difference on a small sample is not enough to declare a winner.
- Check business guardrails: a higher click rate can coexist with fewer purchases or higher cost. Record cost and downstream outcomes relevant to your goal.
- Choose a status: adopt for a further rollout, retest, inconclusive or invalid due to a setup failure. Record the evidence and the next action, not just a winning label.
Sequential tests can be affected by seasonality, inventory or offer changes. If allocation was not controlled, describe an observed association rather than claiming the opening caused an uplift. Do not turn a platform-specific confidence threshold into a universal guarantee.
Prepare concepts in VideoAnt; run tests in your ad platform
VideoAnt helps organize the product brief, three hooks and an approximately 15-second storyboard for free. It does not currently publish campaigns, split traffic or import ad-platform results. Fill this log with actual data from the platform where you run the test.
Optional VideoAnt generation currently costs 30 credits for one 5-second, 480p single-shot concept clip in 16:9 or 9:16, after you review and confirm the quote. It is not a finished 15-second ad and does not automatically include voiceover or subtitles. Review product accuracy and complete editing before treating any concept as an approved test asset.
If generated clips differ in lighting, motion and product appearance as well as the opening, the comparison cannot isolate the hook. Use controlled editing and approved assets when you need to test only that element.
Sources and common questions
Method references checked October 24, 2026: Google Ads: experiment planning guidance and TikTok Ads: split-testing variables. This worksheet is VideoAnt’s own planning resource, not an endorsed platform tool.
How long should a creative test run?
There is no universal duration. Use your baseline volume, the outcome’s reporting delay, platform guidance and the plan agreed before launch. If the available data cannot support a decision, record an inconclusive outcome rather than shortening the rule after seeing results.
Can I test three hooks at once?
That depends on the platform’s supported design and your available traffic. This worksheet starts with two variants to keep the comparison explicit. Adding variants requires an appropriate allocation and analysis plan; it is not the same as choosing a favorite from three previews.
Does the CSV calculate a winner?
No. It stores the plan and actual figures. Use an appropriate experiment analysis or the platform’s supported readout, and keep the original report with the log.