Skip to main content

General guidelines

For an upcoming Apex test, we’ve seen the most success when customers run some variation of this setup:
  • Create an A/B test in Meta Experiments with two separate campaigns:
    • Apex campaign: optimize using Custom Attribution.
    • BAU campaign: continue optimizing using Meta’s native attribution.
      • Note: we recommend a fresh duplicate of your BAU campaign to compare apples to apples.
  • Keep the campaigns as identical as possible (audience, budget, creative, etc.). As long as the key variable you’re testing is the attribution setting, you’re set up for success.
  • Meta recommends running the test for at least two weeks to allow for the algorithm to reconfigure itself. Because your Meta account has never optimized to this attribution model before, we typically see slower performance at the start as a result.
  • Use Northbeam as the source of truth, with GA as a secondary reference, and monitor CPM, CPC, and conversion rate for tradeoffs.
    • Apex campaigns will typically see increased CPM, CPC’s, etc. with increased CVR to make up the difference in media cost.
      • This is a watchout for you on GA4 as well, traffic will likely dip for the Apex campaign but CVR should see an increase.

Step 1: Build campaign for testing

Build out the two ad sets outlined below: Navigate to Ads Manager table
Copy the reference ad set into two new campaigns: Meta PBA test – BAU and Meta PBA test – TEST
  1. Control - Create a new Sales objective campaign and, within the new campaign, create a duplicate L2 (ad set) from the reference L2 (ad set). Name this “Meta PBA test – BAU {campaign name}”.
  2. Test - Repeat step 1 and name this “Meta PBA test – TEST {campaign name}”
  3. Campaign and ad set need to satisfy the following eligibility criteria (for both Test and Control)
    1. Select Sales as your campaign objective, then click Continue.
    2. At the campaign level, select Campaign budget as your budget strategy, and Highest volume or Highest value as your Campaign bid strategy.
    3. Create and edit an ad set to configure these conversion settings:
    4. Conversion location: Select Website or Website + App.
    5. Performance goal: Select Maximize number of conversions or Maximize value of conversions
    6. Dataset: Select an eligible dataset
    7. Conversion event: Select Purchase
    8. Attribution model:
      1. Select Custom for Test
      2. Select Standard for BAU
    9. Select the custom attribution source:
      1. If using Web only, custom attribution source will be auto-populated.
      2. If using both Web and App datasets, select the custom attribution source for each.
  4. Save and stop for QA
Reminder: Please make sure creatives and targeting match for both test and control.

Step 2: A/B test setup

Set up a split test using the Experiments tool.
  • Set up the A/B test in Experiments by clicking ‘Create an A/B test’ Create an A/B test in Meta Experiments
  • Add campaigns to test and control cells: Add campaigns to test and control cells
    • Campaign A (Test): Name: “Meta PBA test – TEST {campaign name}”
    • Campaign B (BAU): Name: “Meta PBA test – BAU {campaign name}”
    • For more complex designs (e.g., 3-cell), add additional cells and campaigns as needed
  • Set the schedule
    • Start date: one day BEFORE the campaign start date
    • End date: one day AFTER the campaign end date
    • DISABLE ‘End test early if a winner is found’ to preserve statistical validity End test early if a winner is found toggle, turned off
  • Test Name
    • Name your study clearly (e.g., Custom Attribution vs BAU - Q3 2026)
  • Determining the winner:
    • Select ‘Cost per purchase’ as the key metric for winner declaration
    • You may add metrics for additional business-specific KPIs
    • Use your own Source of Truth to determine the campaign winner

Step 3: Launch

Things to monitor during the test flight

Continue to monitor the following key metrics and reach out to Meta with any questions:
  • Ensure CAPI integration & campaign quality for changes in Events Manager/Ads manager
  • Ensure your match key coverage remains consistent before and during the test
Reminder: Don’t make manual changes to the test or control campaigns while the test is running. If you make any changes, alert Meta immediately.

Step 4: View results

Custom Attribution campaigns are tagged with a “Custom” label in Ads Manager. Rather than displaying Meta’s default attributed conversions, the Results column reflects the advertiser’s externally attributed data.
Expect up to a 48-hour delay in Ads Manager. Custom Attribution results in Ads Manager are delayed by up to 48 hours, so you won’t see conversions there for up to 48 hours after launch. When data is delayed, Ads Manager shows a data delay indicator; hover over the asterisk in the Results column to see why. The delay doesn’t affect optimization, and results are available in Northbeam in real time.
Judge the test in Northbeam, not Events Manager. The Custom Attribution Source panel in Meta Events Manager shows the raw API calls Meta receives from Northbeam and counts every event as 1, even when Northbeam credited Meta with a fraction of the conversion. It confirms data is flowing but isn’t a measurement view.