Cost per acquisition (CPA) is spend divided by conversions. To get it per creative variant, every variant needs a name, a tagged link and a row in a scorecard. Without those three, you'll know a campaign worked, but not which image, hook or format did the work.
TLDR: Name variants with a fixed pattern, put that name in a UTM tag, and record spend, clicks and conversions per variant in one scorecard. Only judge once each variant has 20 to 30 conversions, and change one element per variant.
Change one thing per variant
A test only tells you something if the variants differ in one respect. If variant B has a new image, a new headline and a new offer, you can't tell which change helped. Pick one element:
- the hook (first line, or the first 2 seconds of video)
- the visual (product vs. person)
- the format (static, carousel, short video)
- the offer (the call to action)
Keep everything else identical. It also keeps your briefs simple, as we explain in AI-assisted creative briefs.
Use a naming rule that survives reporting
A name like "final_v3_new" tells you nothing in a report. Use a pattern you can sort and filter: campaign_element_variant_date, for example springsale_hook_B_1004. Use it in four places: the file name, the post title in your calendar, the UTM campaign or content tag, and the scorecard.
| Piece | Example | Purpose |
|---|---|---|
| Campaign | springsale | Group all variants |
| Element tested | hook | Know what changed |
| Variant | B | Compare against A |
| Date | 1004 | Spot fatigue later |
Tag every link
Add UTM parameters so each click arrives labelled: source (the platform), medium (social or paid-social), campaign (your campaign name) and content (the variant name). One tagged link per variant per platform is enough. Check that the analytics tool on your website or shop shows the content tag, because a rule nobody can read back is worthless.
Build the one-row scorecard
One row per variant. Keep it short enough that the team will actually fill it in.
| Variant | Spend | Clicks | Conversions | CPA | Conversion rate |
|---|---|---|---|---|---|
| springsale_hook_A | |||||
| springsale_hook_B |
CPA is spend divided by conversions. Add the conversion rate (conversions divided by clicks) so you can tell a weak creative from a weak landing page: few clicks point at the creative, plenty of clicks with few conversions point at the page.
Decide with a minimum, not a hunch
Early numbers lie. A variant with 3 conversions and a low CPA may just have been lucky. Set a rule before the test: don't judge until each variant has 20 to 30 conversions or two full weeks, and only then compare CPA. If the gap between variants is small, call it a tie and test a bigger change.
For organic posts, where there's no spend, swap CPA for orders or leads per 1,000 reach. Our guide to predicting which posts will convert covers that version.
Keep the plan and the posts in one place
The scorecard only works if the variant you scheduled is the variant you measured. Schedule each variant with its name in the content calendar, and use Mydrop's performance analytics for the organic engagement side. Mydrop doesn't replace your ad platform's spend data, so pull spend from there into the scorecard.
Read the result and decide what to scale
When a variant wins on CPA, don't just raise its budget. First check that it also won on conversion rate, so you know the creative did the work and it wasn't a lucky audience. Then write down the winning element (for example, "hook that names the price") and use it as the starting point for the next test. Retire any variant whose CPA climbs for two weeks in a row, because audiences tire of the same creative. Keep a short log of winners, so new team members don't repeat old tests.
Run one clean test next week
Pick one campaign. Create two variants that differ in one element, name them with the pattern, tag the links and open the scorecard. In two weeks, you'll have your first CPA per creative. If you run several brands, a multi-brand workspace keeps each client's names and calendars separate. Start free at app.mydropai.com.
Google review
Trustpilot review
Product Hunt comment
