Meta Advantage+ Sales Campaigns: When to Use Them, When to Run Manual, and What the Incrementality Data Says (2026)

meta's pitch for advantage+ sales campaigns is that automation beats your media buyer. the independent data says something more specific: automation wins the first two weeks, loses the second two, and reports a number about 12 points better than it earns. both can be true, and the structure that works is built on both. this guide is when to run advantage+, when to run manual, and how to keep the two honest.

---

what advantage+ sales campaigns actually are

meta's automated sales campaign type: audience, placements, budget, and destination decided by the model, creative supplied by you, optimized against the pixel and conversions api signal. meta describes it as "our most advanced ai marketing tool" that "optimizes your audience, placements and budget automatically" and claims a "20% lower cost per result on average" compared to traditional campaigns (meta for business, advantage+ sales campaigns). the earlier advantage+ shopping campaign product was folded into a unified advantage+ campaign experience in 2025, with controls returning to the ad set level through 2025 and 2026 according to practitioner reporting.

here is what most guides get wrong: they frame the choice as advantage+ versus manual. under andromeda, both campaign types run on the same retrieval and ranking stack (see the algorithms behind meta ads); the difference is how much of the input you hand over. advantage+ is a harvester: it finds the people closest to buying, fast. manual is where you decide the audience shape and the sequence. the question is not which one, it is how much of the budget each one gets and what you measure them on.

---

the fundamental shift: automation is the default, and the default over-reports

before: advantage+ shopping was an opt-in experiment next to a manual structure. most spend stayed manual.

after: advantage+ sales became the default path in ads manager, and by 2026 industry estimates put roughly 62% of meta ecommerce spend inside it (see the andromeda guide). meta's own performance claims for advantage+ sales run from "9% lower cost per action" (april 2025) to a "22% increase in roas" and "$4.52 return on ad spend for every dollar spent" (december 2025) (meta for business, demystifying creative diversification).

and the independent data arrived. haus's analysis of 640 meta incrementality tests found that "58% of brands saw higher iroas on manual campaigns than advantage+," that advantage+ "performed 9% better than manual campaigns at the experiment midpoint but 12% worse by the end," that the post-treatment lift was "+17% vs +32% for manual campaigns," and that "advantage+ over-reported relative to manual campaigns by 12 percentage points on average." advantage+ did win on new-customer share (70% versus 65%) and on omnichannel halo (51% versus 43%) (haus, the meta report).

in plain english: advantage+ finds the buyers who were closest to buying, converts them fast, and reports it generously. manual builds the demand that advantage+ harvests. run both, size them deliberately, and measure them outside the platform.

---

what this broke in existing accounts

1. accounts that moved everything into advantage+ and watched it decay

the midpoint win became the end-of-test loss. with no manual prospecting feeding new demand, advantage+ kept harvesting a shrinking pool, cpa rose after week three, and the reported roas kept looking fine because the reporting gap did not shrink with the performance.

2. budgets scaled on the reported number

a 12-point over-report relative to manual means the campaign that looked best in ads manager was, on average, the one whose reported and incremental numbers diverged most. accounts that moved budget toward the highest reported roas moved it toward the biggest gap.

3. existing customers harvested as "sales"

without an existing-customer list and a budget cap, advantage+ serves people who were going to buy anyway. meta's own help notes that running without the cap "can result in lower cost per purchase," which is true and is exactly the problem: the cheapest purchases are the least incremental.

4. creative starvation

advantage+ handles audience and placement. it does not handle creative volume. accounts that expected the campaign type to compensate for six tired ads ran into the same entity-id grouping as everyone else (see the creative testing system).

5. signal below the floor

advantage+ needs about 50 optimization events a week to stabilize. accounts under that floor split spend between advantage+ and manual and starved both.

6. manual campaigns kept as a museum

interest stacks, ten ad sets, three ads each. the manual half of the structure was the 2022 structure, and it lost to advantage+ because it was built wrong, not because manual loses.

7. the same conversion counted by both

advantage+ and manual campaigns in the same account share the pixel. an order touched by both is attributed by both in ads manager. campaign-level roas comparisons inside the account are comparing overlapping claims.

---

the data

• meta: advantage+ sales campaigns deliver "20% lower cost per result on average" versus traditional campaigns (sales campaigns page); "9% lower cost per action" (april 2025) and a "22% increase in roas" with "$4.52 return on ad spend for every dollar spent" (december 2025). self-reported.
• haus, 640 meta incrementality tests since early 2024: 58% of brands saw higher incremental roas on manual; advantage+ +9% versus manual at the midpoint, -12% by the end; post-treatment lift +17% versus +32%; advantage+ beat manual 42% of the time; incremental roas 12% lower at 18% lower daily spend; over-reported relative to manual by 12 points; new-customer share 70% versus 65%; omnichannel halo +51% versus +43%. (haus, the meta report)
• haus, july 2025 to june 2026: meta's incremental attribution setting outperformed standard attribution by a pooled 1.26x, 1.38x for dtc-only brands. (haus)
• confect, 115.7 billion impressions across 3,014 advertisers in 2025: overall ecommerce roas down 7% post-andromeda, top-third advertisers down 31%. (see the andromeda guide)
• industry estimate: about 62% of meta ecommerce spend runs through advantage+ sales campaigns in 2026. (as reported in the andromeda guide's sources)

read the platform and the independent numbers together: advantage+ is cheaper per reported result, converts new customers at a slightly higher share, and delivers less incremental return per dollar than a well-built manual campaign over a full test window. that is a description of a harvester.

---

what changed in 2026 specifically

structure returned. the unified advantage+ campaign experience brought ad sets back to advantage+ sales campaigns, with per-ad-set creative limits, so a single advantage+ campaign can be segmented by offer or product line without leaving automation. practitioner reporting through 2026 also describes the existing customer budget cap and audience controls being restored after their early-2025 removal; verify the controls available in your account rather than assuming.

incremental attribution became worth turning on. the setting flipped from underperforming to outperforming standard attribution in haus's pooled tests over the last year. it is now the default to test on advantage+ campaigns, whose reported numbers need it most.

andromeda finished rolling out. the retrieval engine that rewards creative diversity is the same one advantage+ runs on. automation did not remove the creative requirement; it raised it.

unchanged: the 50-event weekly floor, the need for an existing-customer list, and the fact that in-platform roas comparisons between advantage+ and manual are comparisons of two attributed claims on the same orders.

---

what actually works now

1. decide the split from the account's signal

under 50 purchases a week: one campaign, and it should be advantage+ sales only if that gets the whole signal in one place; otherwise one broad manual campaign. above 50 a week: advantage+ sales at roughly half to two-thirds of budget as the harvester, manual prospecting at a quarter to a third to build demand, and a small remarketing or retention layer if it is not already inside advantage+.

2. give advantage+ a ceiling, not a blank check

set the campaign budget as a ceiling that the manual layer can keep feeding. when advantage+ cpa rises past two weeks of stable performance, the fix is more prospecting and more creative, not more advantage+ budget.

3. upload the existing-customer list and cap it

connect the customer list, set the existing customer budget cap where the control is available (a quarter to a third is the common working range), and read new-customer share as a primary metric. if the cap is not exposed in the account, exclude existing customers through the audience controls or run existing customers in their own campaign with its own budget.

4. build the manual layer for 2026, not 2022

one campaign, one or two broad ad sets (location and age minimum only, advantage+ audience on), 10 to 20 genuinely different creatives per ad set, campaign budget. this is the andromeda structure, and it is what beat advantage+ over full test windows in haus's data. the manual layer's job is prospecting angles and sequences that advantage+ then harvests.

5. share creative, not audiences

the same creative concepts run in both layers; new concepts launch in manual first because you control the sequence there, and winners move into advantage+. audiences are not shared: advantage+ finds its own.

6. measure outside the platform

turn on incremental attribution and compare it with standard on a holdout. once a quarter, run a geo holdout with advantage+ and manual in separate treatment cells so each gets its own multiplier (see incrementality testing). report reported roas, incremental roas, and new-customer cac side by side; budget on the second and third.

7. keep the signal clean

pixel plus conversions api, deduplicated, with a good or great event match quality on purchase (see the meta capi guide). advantage+ automates on this signal; a weak one automates the wrong thing.

8. re-run the split every quarter

as the manual layer finds new angles and advantage+ harvests them, the right split moves. the holdout multipliers tell you which way.

---

what to stop doing immediately

• stop moving the whole budget into advantage+ and calling it a structure.
• stop scaling advantage+ on its reported roas.
• stop running advantage+ without an existing-customer list and a cap or exclusion.
• stop expecting automation to fix creative starvation.
• stop keeping a 2022 manual structure as the comparison and concluding manual loses.
• stop comparing advantage+ and manual roas inside ads manager as if they were independent.
• stop deciding the split once and leaving it.

---

the orbit agent instructions

{
  "guide": "meta-advantage-plus-shopping",
  "agent_tasks": {
    "audit_budget_split_and_signal": {
      "check": "read weekly purchase events for the account, budget share by campaign type (advantage+ sales, manual prospecting, remarketing), and the 14-day cpa trend on each",
      "flag": "advantage+ above 75 percent of budget with no manual prospecting layer, any campaign under 50 weekly events while another runs the same objective, or advantage+ cpa rising for 14 days while its budget rises",
      "recommend": "set advantage+ at half to two-thirds with a ceiling, fund a broad manual prospecting campaign, and consolidate any sub-floor campaigns"
    },
    "audit_existing_customer_controls": {
      "check": "confirm an existing-customer list is connected to advantage+ campaigns, whether a budget cap or exclusion is set, and the new-customer share of purchases over 30 days",
      "flag": "no customer list, no cap or exclusion, or new-customer share below 60 percent on an advantage+ campaign meant for acquisition",
      "recommend": "connect the buyer list, cap existing-customer spend at a quarter to a third or exclude existing customers, and track new-customer cac as a primary metric"
    },
    "audit_manual_layer_structure": {
      "check": "for manual prospecting campaigns, count ad sets, targeting constraints beyond location and age, creatives per ad set, and budget type",
      "flag": "more than 2 ad sets per objective, interest or lookalike stacks as primary targeting, fewer than 10 creatives per ad set, or ad set budgets instead of campaign budget",
      "recommend": "rebuild to one campaign, one or two broad ad sets, campaign budget, and 10 to 20 diverse creatives"
    },
    "audit_creative_flow_between_layers": {
      "check": "compare active creatives in advantage+ and manual campaigns by concept, and list winners in manual not yet running in advantage+",
      "flag": "advantage+ running fewer than 10 distinct creatives, or manual winners past the spend floor not promoted into advantage+ within 14 days",
      "recommend": "promote cleared winners into advantage+ and launch new concepts in manual first"
    },
    "compare_reported_vs_incremental": {
      "check": "read reported roas, incremental attribution roas, and the latest geo holdout multipliers for advantage+ and manual separately, plus new-customer cac for each",
      "flag": "budget decisions made on reported roas alone, no holdout with separate cells in 120 days, or incremental attribution never tested against standard",
      "recommend": "budget on incremental roas and new-customer cac, schedule a two-cell holdout, and test incremental attribution on advantage+ campaigns"
    },
    "audit_signal_quality_for_automation": {
      "check": "read event match quality on purchase, deduplication coverage, and the share of purchases arriving from both pixel and conversions api",
      "flag": "event match quality below good, deduplication under 90 percent, or purchase coverage from both sources under 90 percent",
      "recommend": "fix the conversions api setup before adjusting the advantage+ and manual split"
    }
  }
}

---

the bottom line

advantage+ sales campaigns are a harvester: fast, cheap per reported result, and generous in what they report. manual is where demand gets built. give advantage+ a ceiling and an existing-customer cap, build the manual layer the andromeda way, move winning creative from manual into advantage+, and measure both with incremental attribution and a holdout.

the platform's number is the start of the conversation. the holdout ends it.

---

built by orbit - the performance marketing intelligence layer. questions? jose@orbitllm.com