How Meta Advantage+ Shopping Picks Creatives Explained

Quick Answer

Advantage+ Shopping (ASC) runs a campaign-level creative bandit rather than separate ad-set bandits. All ads in the campaign compete in a single pool, with budget flowing to whichever creative has the highest eAR for whichever audience segment is in play. This means ASC tests creative across the entire eligible population, but only if the creative survives the first 50-200 impressions.

The Mechanism Explained

ASC's architecture differs from standard CBO in three ways:

1. One pool, no ad sets — all creatives compete directly. There's no per-ad-set learning.
2. Pooled signal — every conversion across the campaign improves the eAR for all creatives, not just the one that delivered it.
3. Audience is fully algorithmic — no audience targeting, just the broad population.

The creative bandit does the heavy lifting. When you upload 30 creatives to an ASC campaign, the bandit:

1. Allocates a cold-start budget to each new creative — typically enough for 50-200 impressions
2. Computes early eAR estimates from the cold-start data
3. Promotes creatives with above-median eAR to the main pool
4. Demotes below-median creatives to a maintenance pool that gets occasional impressions to detect drift

Creatives in the main pool fight for impressions via the standard ranker. Creatives in the maintenance pool typically get 0.5-2% of total spend as monitoring traffic. Creatives that fail cold-start get effectively zero further delivery unless you pause others.

A subtle behaviour: ASC will not test new creatives on weekends or during high-cost auction windows. The bandit defers cold-start to lower-cost periods to minimise wasted budget. This is why dropping new creatives into ASC on Friday often results in delivery only starting Sunday morning.

The cap on creatives in ASC has historically been 150 ads per campaign. The practical sweet spot is 8-15 ads — enough variety for the bandit to find winners, few enough that cold-start budgets aren't fragmented.

Practical Implication

Don't dump 100 ads into ASC. Start with 8-15, let the bandit converge over 5-7 days, then add 2-3 new ads at a time as challengers. Adding ads in waves preserves the learning the existing ads have done while still testing new variants.

Real Numbers

• ASC creative cold-start budget: ~50-200 impressions per ad
• Maintenance pool spend share: 0.5-2% of campaign
• Optimal ad count per ASC campaign: 8-15 for most accounts
• Cold-start success rate for new creatives in ASC: ~30-40%

FAQs

Q: Can I prioritise specific creatives in ASC?
No — the bandit decides.

Q: Does ASC respect my audience inputs at all?
Mostly no, except for excluded existing customers.

Q: How many creatives should I add to ASC at once?
8-15 to start, then 2-3 at a time after.

Q: Why don't some of my creatives get any impressions?
They failed cold-start; bandit decided not to invest more.

Q: Does ASC use Dynamic Creative?
No — ASC uses static ads competing in a single pool.

Pix-Vu

ASC's bandit only finds winners if you give it real creative diversity. Pix-Vu generates 10-15 distinct product image variants quickly so you can populate ASC properly — at https://pix-vu.com.