All answers

The Prevailing
Answers for
Business model:
Business model, pick up to two

Should we send a third-party attribution tool's results to Meta as the signal it optimizes for?

The PrevailingUpdated Oct 20, 2026

Our answer

Try Custom Attribution only through a two-cell test in new campaigns, against the Meta setting you would otherwise run. Do it only if you pay for a supported tool whose numbers differ from Meta's. Otherwise stay on Meta's models.

Confidence
Low
Strongest proof
3 controlled tests
4 Meta rules, 1 dataset, 2 practitioners, 4 opinions or claims
Sources
9 sources
14 statements

New here? Take the 1-minute tour

Why this is the answer

3 reasons and our judgment

The only Custom Attribution results come from Northbeam, which sells the connection. Ka'Chava's Custom cell beat Standard by 43% on new-customer ROAS, and Kitsch's by 8.2% over three weeks, both read in Northbeam.

Meta's own Incremental Attribution has independent support. It earned 1.26 times standard attribution's incremental return from July 2025 to June 2026 in Haus holdout tests, and 1.38 times for brands that sell only on their own site.

The gain depends on your account. It depends on how much your model differs from Meta's (Rockerbox), and reads best at 15 or more Meta-attributed conversions a day (Triple Whale). Jon Loomer, a Meta ads educator, advises that most advertisers shouldn't bother.

Our judgment

  • Require a two-cell test, and only with a tool you already pay for. The only results are Northbeam's vendor-graded tests and the partners' own documentation.
  • Judge the test on new-customer cost in your own numbers. This joins Northbeam's and Triple Whale's advice to judge on cost per purchase in your own source of truth with Ka'Chava's new-customer goal.
  • Compare the Custom Attribution campaign with the setting you would otherwise run. That means Incremental Attribution if you have moved to it, because a Custom cell that beats standard may still trail it in Haus's results.
  • Build each test cell as a new campaign with campaign budget and the same bidding. Custom Attribution can be enabled only on a new campaign and needs campaign budget (CustomerLabs, Triple Whale), and Northbeam's Ka'Chava test and Triple Whale's test guide both use identical cells.
  • Re-check Incremental Attribution against a holdout every few months. Brad Ploch (Scalability School) plans to retest it in three to six months after it drifted.

Our judgment, built from Northbeam, Rockerbox, Triple Whale, Haus, CustomerLabs and Brad Ploch (Scalability School): our reading of these sources, not a finding from any one of them.

What the sources say

14 statements from 9 sources
  • Bryan BumgardnerNov 5, 2026Caveat

    Northbeam's Bryan Bumgardner says that across 1,000 Northbeam customers, Meta conversion rates hit record lows while ROAS improved. He gives no figures or method, and Northbeam sells the fix he recommends.

    “ROAS is better than ever, but conversion rates are tanking.”

    Read it on x.com

  • NorthbeamOct 29, 2026Supports

    Northbeam, the tool's seller, says Kitsch's three-week A/B test of its attribution fed into Meta gave 8.2% higher new-customer ROAS and 8.7% lower cost per new-customer order. Spend not given.

    Read it on themediabuyer.northbeam.io

  • Nik SharmaOct 21, 2026Supports

    Maxx Blank says feeding Triple Whale's custom attribution model back into Meta for optimization shows big results for brands. He is praising his own company's product and gives no figures.

    “take an attribution model and put it back in the meta and it will optimize off that.”

    Read it on podcasts.apple.com

  • RockerboxOct 8, 2026Supports

    Rockerbox says Meta's Custom Attribution unlocks after 14 days of passed-back data, so allow about four weeks from setup to a readable two-week test; it helps most when your model and Meta's disagree.

    “If your own model closely tracks Meta's reported results, there is less room for the model to improve delivery.”

    Read it on help.rockerbox.com

  • Triple WhaleSep 16, 2026Supports

    A vendor's setup rules for Meta Custom Attribution, with no results: Sales campaigns optimizing for Purchase, campaign budget, about two weeks of signal, and testing with at least 20% of budget or a 50/50 split.

    “at least 20% of your Meta budget”

    Read it on kb.triplewhale.com

  • CustomerLabsSep 14, 2026Supports

    A vendor explainer says Meta's setup uses Sales campaigns optimizing for Purchase and can be enabled only on a new campaign. It may not be available to every account, and passback must arrive within 48 hours.

    “can be enabled on a new campaign only; it cannot be added to an existing campaign”

    Read it on customerlabs.com

  • NorthbeamAug 27, 2026Supports

    In a Northbeam-run test at one brand, Ka'Chava, Meta ad sets optimized to Northbeam's signal beat Meta's standard attribution by 43.2% on new-customer ROAS, read in Northbeam. Spend and duration are not given.

    “ran a controlled A/B test at the Meta ad-set level with identical campaigns.”

    Read it on themediabuyer.northbeam.io

  • Triple WhaleAug 19, 2026Caveat

    Triple Whale says Meta's profit optimization needs a profit field sent through the Conversions API, about 200 purchases in 28 days, at least 3 distinct profit values and Meta enabling it. It helps only if margins vary.

    “confirm that your margins actually vary across products”

    Read it on kb.triplewhale.com

  • Triple WhaleAug 12, 2026Supports

    Triple Whale's docs for Meta's Custom Attribution test: two cells, Custom and Standard, run at least 2 weeks after about 14 days of data, on Sales campaigns. Asks for 15+ Meta-attributed conversions a day.

    “Judge on cost per purchase in your own source of truth.”

    Read it on kb.triplewhale.com

  • Jon LoomerAug 1, 2026Disagrees

    Jon Loomer says Meta's Custom attribution setting needs granular data from a partner such as Northbeam or Triple Whale. It could help if you think Meta overstates results, but most should use Incremental when unsure.

    “That said, most advertisers shouldn’t bother with this.”

    Read it on jonloomer.com

  • HausJul 16, 2026Supports

    Haus's randomized tests show Meta's Incremental Attribution returning 1.11x the incremental return of standard optimization overall: 0.80x in July 2024 to June 2025 and 1.26x the next year. Test counts are not given.

    “A year ago, Haus data showed Meta’s standard attribution performing better than their Incremental Attribution setting.”

    Read it on haus.io

  • Scalability SchoolJul 2, 2026Caveat

    Brad Ploch says for one apparel brand at $150-200K a month, Incremental Attribution passed a lift test but drifted after Meta's March click change, so they went back to one-day click and plan to retest.

    Read it on scalabilityschool.com

  • NorthbeamMay 28, 2026Supports

    Northbeam, a vendor, says dozens of identical campaigns on Meta's base algorithm versus Apex averaged +21.43% ROAS and -23.78% cost per order. It gives no advertiser count or spread.

    “A/B tested dozens of identical campaigns across a set of advertisers”

    Read it on themediabuyer.northbeam.io

  • Northbeam2026Supports

    Northbeam, which sells the integration, says Custom Attribution runs on Sales campaigns at the ad-set level and needs about 50 Northbeam-attributed orders a day. Ads Manager then shows Northbeam's credit.

    “The Results column in Ads Manager for Custom Attribution campaigns reflects Northbeam-provided credit and value at the ad ID level”

    Read it on docs.northbeam.io

In the full answer

For readers who sign in
  • What to do this week3 steps
  • The numbers behind it3 figures from cited sources
  • Where each source standsA chart placing each source on the question
  • The strongest disagreementThe best case against the answer, from a cited source
  • How sure we areWhat its confidence rests on, and what would change the answer

Read next

More in Measurement