Creative testing template and tracker

Download the creative testing template

The file is creative-testing-template.csv: a header row, two marked example rows and 20 columns. In Google Sheets, use File, Import, Upload and keep “Convert text to numbers, dates, and formulas” ticked. In Excel, open the file directly. The formulas use only functions both spreadsheets share, so no macro or add-on is needed.

Columns H, I, J, K, N, Q and R compute. The rest you type. Replace both example rows before you plan anything with the sheet.

H  spend per variant
   = F x G
I  total spend
   = E x H
J  pairs compared
   = E x (E - 1) / 2
K  smallest detectable gap
   = 1 - 1 / EXP(
     (NORM.S.INV(1 - 0.05 / (2 x J))
     + NORM.S.INV(0.8))
     x SQRT(2 / G))
Q  observed gap
   = 1 - O / P

The sheet states no CPA, conversion count or budget of its own. Apart from the two marked example rows, every number in it is an input you supply.

Fill in one row: a worked test

Row T01 shows the arithmetic with example inputs, never a benchmark: a CPA of $30, 40 conversions per variant and three hooks.

A two-variant row at the same inputs detects about 47 per cent, and five variants push it to about 56. Both figures match the creative testing budget calculator and the framework. The three-variant 51 comes from the same formula with z = 2.39 (method).

So read the result before you write the hypothesis. If you expect the new hook to cut CPA by 20 per cent, 40 conversions per variant cannot show it. The framework’s table puts that at about 315 conversions per variant for two variants.

T01’s example result reads 24 against 31: a 23 per cent gap, inside the 51 per cent the row can detect. Column R prints “Inside the gap”. The decision cell records what follows, such as a rerun of the best two hooks with more conversions.

The creative testing matrix: what to test first

A creative testing matrix lists the variables you could change, and each row of the sheet takes exactly one. Meta’s page on testing more than one variable gives the reason: with several, “you won’t know exactly which variable led to the winning ad set”.

Format rows need care, since Meta’s “dynamic” options mix assets instead of comparing them. Static vs dynamic ads separates them. AI hook generators compares the tools that write hook variants.

The creative testing tracker: logging results and making the call

The tracker half of the sheet, columns L to S, records what happened against what the row promised. Log the start date, then fill columns O and P with the best and runner-up CPA when the stop rule in M is met.

Call the row at its stop rule, never earlier. Meta’s A/B test best practices recommend “a minimum of 7-day tests” and cap A/B tests at 30 days, which is why the example stop rules end at 30 days.

A creative test inside a live campaign gives no confidence figure: Meta’s creative test page says “A confidence level is not included.” Column K is the yardstick then.

Log fatigue in the notes column too. Meta’s creative fatigue page shows “Creative limited” when cost per result rises above your past ads but stays under twice as much, and “Creative fatigue” at twice or more. It covers ad sets with one creative only, so a test with several ads in one ad set never gets the status.

A creative testing roadmap for a quarter

A creative testing roadmap is the sheet’s rows in launch order, with column I summed by month. That sum is what the quarter commits, and it decides how many rows fit.

Order rows by two things. First, what the last called row decided: a winning hook fixes the hook for the next format row. Second, cost: at a CPA of 30, a two-variant row at 20 conversions commits $1,200, and a five-variant row at 40 commits $6,000.

Status in column B (Planned, Live, Called) filters the sheet into the roadmap view. A spreadsheet filter on B equal to Planned, sorted by L, gives the quarter’s queue.

Naming ads so the tracker matches Ads Manager

Column N builds an ad name prefix from the test ID and the variable, such as T01_hook. Add the variant after it, T01_hook_v2, when you name each ad. Ads Manager exports carry the ad name. A filter on the prefix pulls every variant of a row into one view, and the CPA columns paste straight into O and P.

Use the same prefix on every platform so the exports match.

Survey and pre-testing templates: the other kind

Several pages that rank for this query, such as aytm and Attest, offer survey templates that ask a panel about an ad before any spend. They measure stated reactions such as recall or appeal, where this sheet measures CPA in live delivery. Tools that predict ad performance before launch sit in AI creative testing tools.

The ad creative brief template sets out every field, one by one.