How to budget and document a creative test

A folder of new creatives is not yet a test. Before spending, decide what question the variants answer, what you can afford to learn and when the result will be reviewed. This worked example uses five background directions from our HeliosGen workflow. Every number below is illustrative.

1. Define a comparison you can explain

Keep the product, offer, audience definition, destination and measurement window consistent. Change the background direction: studio, commute, travel, home or detail. Keep an approved existing creative as a control where the experiment design supports it.

Write the question before producing more assets: “Does a commuter setting improve our chosen business outcome compared with the current approved creative?” Put the exact conversion event and reporting window in the brief. If the platform optimizes delivery unevenly, simply comparing ad rows does not establish a causal winner.

2. Calculate a spending scenario

Use the planner below, or open the standalone creative test budget planner. Enter five variants, an assumed CPA of 15, ten planned conversions per variant and a five-day duration. The arithmetic is:

5 variants × 15 assumed CPA × 10 conversions = 750
750 ÷ 5 days = 150 per day
15 × 10 = 150 per variant

This is an affordability exercise. Ten conversions per variant is an input, not a recommended statistical sample or a guarantee that the budget will buy that many conversions. Actual delivery, conversion rates and reporting delays can differ. Decide separately whether this design can answer your question and whether the business can afford the loss.

Run a second scenario with CPA doubled to 30. The envelope becomes 1,500. If that exceeds the approved limit, reduce the scope before launching instead of assuming the optimistic case will hold.

3. Prepare the results sheet

Download the creative test matrix and open the CSV in your spreadsheet application. Replace the five example directions with your approved asset names. Record production cost separately from media spend, including failed generations and editing time where relevant.

For each variant, keep spend, impressions, clicks, conversions and attributed revenue together. Use one reporting window and one conversion definition across rows. Add the date the export was taken so a colleague can reproduce the comparison after delayed conversions arrive.

4. Write the review rule before launch

Name the owner, maximum approved spend and scheduled review time. Record the minimum evidence your team requires and what makes the result inconclusive. Do not force a winner if tracking broke, delivery was heavily imbalanced or the test stopped early.

For a causal comparison, choose an appropriate controlled experiment. Google's explanation of campaign experiments describes a platform mechanism to compare changes. The worksheet organizes evidence; it does not turn ordinary delivery into random assignment.

5. Hand the evidence to campaign operations

Use the ROAS calculator for consistent arithmetic and the budget pacing tool for the remaining envelope. Attributed ROAS still does not tell you what would have happened without the ads.

For supported Meta and Google Ads work, review the AdRails workflow alongside the brief and results. Keep any proposed campaign change separate from the evidence and obtain the campaign owner's approval before applying it. There is no automatic CSV import or account connection in this worksheet.

The completed deliverable is a short decision record: what changed, what the data supports, what remains uncertain and the next approved action. A smaller test with an honest conclusion is more useful than a spreadsheet that labels every variation a winner.