Pinterest self-serve A/B testing: build creative variants with Sume

Pinterest added self-serve A/B testing for creative, targeting and Performance+ on Sept 17, 2026. Build clean creative variants, one change each, with Sume.

5 min readSume
All posts

Pinterest's September 17, 2026 announcement adds self-serve A/B testing "across creative, targeting, and within Pinterest Performance+" (read 2026-10-03). The test tool compares what you give it, so the useful job for Sume is making variants that differ by exactly one thing.

A bulk run can produce that set: one brief, one varied field per item, a spend cap on each. Sume does not run or read the Pinterest test.

What did Pinterest announce?

The newsroom post bundles several launches. Only the first two rows change how you build creative; the rest are context so you do not cite them for the wrong product.

  • Pinterest's holiday guide (published July 30, 2026) advises testing creative approaches at pre-holiday moments such as Halloween.
  • The guide also says to build on proven signals instead of starting from scratch.
  • Neither page gives a minimum budget or sample size, so none is stated here.
Pinterest performance ads announcement, read 2026-10-03
LaunchPinterest says
Self-serve A/B testingTest campaign strategies across creative, targeting and within Performance+
Priority ProductsPrioritize specific products during key moments, seasonal launches, promotions and category pushes while keeping automation
App PromotionGlobal beta
Visual Search AdsBeta, coming in the weeks after the post, in all Pinterest ad markets
RestyleConsumer feature, beta in the U.S. and Canada

How do I keep variants comparable?

The rule is one variable per test. If variant B changes the headline, the product angle and the background together, a win tells you nothing you can reuse. Write the invariant part of the brief once and put only the changing line in each item.

What does the queue look like?

The queue below holds three items. Fix the product and format, vary the opening line. Each item carries its own generation_spend_cap_usd, and the queue needs a fresh Idempotency-Key per batch.

The sume-editorial-product-set Format produces stills for collections and gift sets; use a video Format from the catalog if your test is on video Pins.

curl -sS -X POST https://api.sume.com/v1/formats/sume/sume-editorial-product-set/bulk-runs \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"concurrency":3,"items":[
    {"instruction":"Gift set, 2:3 Pin, variant A: warm wood table","generation_spend_cap_usd":10},
    {"instruction":"Gift set, 2:3 Pin, variant B: white marble","generation_spend_cap_usd":10},
    {"instruction":"Gift set, 2:3 Pin, variant C: wrapped in paper","generation_spend_cap_usd":10}]}'

What happens after the queue finishes?

After the queue reaches completed, check counts.failed; a missing variant breaks the comparison. Requeue only the failed item and keep its variant id.

Upload the finished files in Pinterest Ads Manager and create the test there. Name each file with the test id and variant letter so results map back.

What Sume does not do

Sume cannot create a Pinterest test, read its result, or enable Priority Products. It also has no statistical test of its own, so declare a winner by Pinterest's reporting.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume