Meta Ads

Meta Ads account structure for e-commerce: testing, scaling, budgets

An online store’s Meta account doesn’t need many campaigns, but each one needs a reason. This 2026 guide lays out a structure by monthly spend, when to use ad set budgets or Advantage+ campaign budget, what to do when one ad takes all the spend, and how to judge results without trusting ad-level ROAS. It draws on Meta’s Help Center and on the practice of one e-commerce agency, labelled as such.

By the Dokiri Studio team · November 6, 2026 · 7 min read

In this article

Start with one campaign and justify every other

Meta’s own advice leans toward fewer, bigger units. An ad set usually leaves the learning phase after about 50 results in the week following its last significant edit (Meta Help Center), and Meta’s page on managing ad volume suggests combining ad sets so they get there sooner.

Bluesense Digital, an e-commerce agency whose presenter publishes long Meta Ads breakdowns on YouTube, adds a reason from practice: in his experience, conversion data is largely kept per campaign, so four campaigns sharing 100 purchases a month each learn from about 25. Start from one prospecting campaign and write down the purpose of every extra one. He is just as clear that putting every ad in one ad set is wrong for most brands. Split when:

  • Unit economics differ: high- and low-margin ranges, or stock you need to clear.
  • Audiences don’t overlap: menswear and womenswear, say.
  • Volume is low: expensive products with few purchases a month.

Inside a campaign, he makes each ad set one concept: one persona, angle and offer, with several ads in different formats and openings. Results then read by concept, which is what your next brief needs. Why concepts matter explains the reasoning.

Ad set names that read by concept (persona_angle_offer_format_date)
night-nurses_warm-water_none_mixed_2026-11-06
gift-runners_daily-use_giftwrap_mixed_2026-11-06
comparers_proof_15off-sunday_offer_2026-11-13

Structure by monthly spend

Bluesense’s templates by monthly spend (video). Treat the thresholds as approximate.

Monthly spendCampaignsInside eachNew ads a month
Under $50kOne testing campaign; a scaling campaign only for ads already spending about $400 a dayOne ad set per concept, 3 to 5 ads each, past buyers excludedAbout 1 per $1,000 of spend
$50k to $250kTwo or three prospecting campaigns, each with a business reason: a persona with its own landing pages, new arrivals, a high-value categoryConcept ad sets as above100 to 300
$250k and upSpecialised campaigns (promo, advertorial, catalogue) and, in his practice, a second ad account on the same pixel to try other biddingConcept ad sets; winners copied into scaling by post ID250 and up, by the same rule

Below $50k, read results over at least five days, and 10 to 14 for small budgets or long buying cycles. Retargeting is optional at every tier in his view: most accounts he audits don’t need a separate website-visitor campaign, because broad prospecting already reaches warm people.

The scaling campaign is not where winners go to be relaunched. Pausing a winner to restart it elsewhere throws away its learning; he leaves it running and adds it to scaling by its existing post ID, so the likes and comments come along.

Ad set budgets or Advantage+ campaign budget

Meta now calls campaign budget optimisation (CBO) Advantage+ campaign budget, also shown as Campaign budget. One budget flows in real time to the ad sets with the best opportunities, and Meta warns it may spend most of it on one ad set, so read results at campaign level (Meta Help Center). Ad set budgets (ABO) fix what each ad set spends. In between, ad set budget sharing lets up to 20% of a daily budget move between ad sets, and a campaign budget accepts minimum and maximum spend limits per ad set (Meta’s comparison).

Bluesense treats the choice as risk management (video). Delivery concentrates spend, and a campaign budget leans into it: in the accounts he has analysed, a few percent of ads take most of it. Efficient, but fragile, since a new concept can end up with no spend at all. Ad set budgets force spend into each test and build a bench of winners.

  • Ad set budgets for testing, for high prices or low purchase volume, and in quiet months such as January.
  • Campaign budget for scaling proven ads, and for peak season from Black Friday to Christmas.

When one ad takes all the spend

It happens in every account, and it is often fine: delivery starves the other ads for a reason. Bluesense’s options, in order (video):

  1. 1

    Launch diverse new creative

    New concepts in a new ad set, not more copies of the leader.

  2. 2

    Otherwise, leave it alone

    If the leader hits your target, the other ads lost auctions you can’t see.

  3. 3

    Keep the zero-spend ads on

    In his experience they support the sequence of ads people see and take over when the leader fades.

  4. 4

    Duplicate only as a last resort

    A copy in a new ad set starts learning from scratch.

Meta’s rules explain the first step: adding a new ad to an ad set is a significant edit and sends the ad set back into learning. Leave an ad set that hits its target alone and put new ads in a new one. If one winner carries the account, he either shifts production toward it or caps its ad set at around 40% of spend and keeps new concepts rotating.

Judge by spend and new customers, not ad-level ROAS

Bluesense ranks signals by trust (video): first the amount spent, then ROAS or cost per purchase on 7-day click or incremental attribution, then cost per click and click-through as early warnings, then hook and hold rates to find which part of an ad to fix. Spend comes first because delivery funds what works: in his view, an ad at 1.7× on twice the spend of one at 2.2× is probably the better prospecting ad.

Ad-level ROAS misleads because people see several of your ads and the purchase is credited to one of them. His trust grows with the level: ad, ad set, then campaign on 7-day click with past buyers excluded, which tracks the bank account best. Know your floor first with the break-even ROAS calculator.

Attribution settings, as Meta defines them

In the standard model, Meta counts conversions within 1 or 7 days of a link click (click-through), within 1 day of an impression (view-through), and within 1 day of another interaction, such as a non-link click or a video watched for 5 seconds (engage-through). The incremental model optimises for conversions it predicts the ad caused. Results from different models don’t compare side by side; use Compare attribution settings (Meta Help Center). Meta notes that some accounts still run older definitions while the change rolls out.

Bluesense’s current default is 7-day click plus 1-day engage-through, without 1-day view, which in his experience over-credits people who would have bought anyway. Then check the business itself: spend ÷ new customers, and new-customer revenue ÷ spend.

Settings people forget

  • Exclude past buyers from prospecting. Custom audience exclusions hold even with Advantage+ audience, per Meta’s developer documentation. He excludes roughly the last two years of buyers, and checks that the customer list still syncs.
  • Review Advantage+ creative enhancements. Meta can adjust your media and text; some enhancements may be on by default, and you can switch them off (Meta Help Center). The list includes AI music, overlays and video subtitles: on a video with captions already burned in, automatic subtitles can double them. Bluesense switches enhancements off on every ad.
  • Keep one attribution setting across ad sets you compare.
  • Watch the Page ad limit. Ads running or in review count: 250 per Page whose highest-spending month is under $100,000, as of November 2026.
  • Decide how many creatives go in one ad. Meta suggests fewer ads with up to 10 creative assets each; Bluesense avoids it for testing, because it hides which creative did the work.

With Dokiri Studio

Dokiri doesn’t build or run campaigns. What it gives this structure is ads ready for concept ad sets: write one persona, angle and offer in the angle field, run a batch of 5 or 10, and you get up to that many 9:16 videos in different formats and openings, each with a Headline and a Primary text. The ZIP’s ads.csv names every ad like product_c01-concept_format_v1, so concept and format stay visible in Ads Manager reports. See export to Meta and the account structure reference.

Frequently asked questions

Should a small store use ad set budgets or a campaign budget?

For testing on a small budget, ad set budgets keep each concept funded. Move proven ads to a campaign budget when you scale. Practitioners disagree, so test both on your account.

Do I need a retargeting campaign?

Often not at first: broad prospecting already reaches warm people. If you run one, keep it small; Bluesense caps it at about seven impressions per person per 30 days.

How long before I turn off a test?

A common rule is to let a test spend about three times your target cost per purchase before deciding; his stricter version is five times, over at least five days. Killing a prospecting ad after one day hides the sales it set up for other ads.

Which attribution setting should I pick?

The standard model offers 1- or 7-day click-through, 1-day view-through and 1-day engage-through; there is also an incremental model. Bluesense judges prospecting on 7-day click plus 1-day engage-through. Whatever you choose, keep it the same across ad sets you compare.

Can Dokiri set up my campaigns?

No. Dokiri makes the videos and the copy files; you build and run the campaigns in Ads Manager.

Try it

Tools in this article

Turn your product into ads today

Drop a product photo or a store link: the scripts are written for free, and you pay only for the videos you make.

Make my first ad

Keep reading

All articles