Last updated: November 19, 2026. Facts checked against the Meta Business Help Center and the cited studies in November 2026.
TL;DR
To bring back past customers with Facebook ads, build an audience from your customer list, exclude people who bought recently, and judge the campaign by whether it creates purchases that would not have happened anyway, measured against what a repeat customer is worth to you. Nova, AdAdvisor's approval-first AI media buyer, reads your Meta account and Shopify store, judges every campaign against your break-even, and changes nothing until you approve.
Quick answer
The Split-List Holdout is AdAdvisor's method for judging Facebook ads to past customers: hold back a random half of your lapsed buyers and judge the campaign only by the margin gap between the two halves.
Split the list.
Export lapsed customers from Shopify and split them at random into two halves.
Advertise to one half.
Upload that half as a customer list custom audience, used as a hard inclusion, and keep the other half out of every campaign.
Judge the gap.
Compare margin per customer in Shopify. Keep the campaign only if the extra margin is larger than the ad spend.
Past customers are the people most likely to buy again anyway, so to win back customers with Facebook ads profitably, measure what the ads add rather than what Ads Manager credits them with. Meta interface details below were checked in the Meta Business Help Centre on November 19, 2026.
How do I advertise to past customers on Meta?
Upload your buyers as a customer list custom audience, then put that list in its own ad set as a hard inclusion. That is the core of customer list Facebook ads (see our Facebook ads targeting guide for the other options): customer file, hashed and matched audience, ad set with Use as a suggestion off, then orders counted in Shopify. Meta's customer list formatting guidelines ask for a CSV or TXT file with at least 1,000 customers and at least one main identifier, such as email or phone number. Phone numbers need a country code, even if every customer is in the US.
Export buyers from Shopify.
Email, phone, name and country give Meta more to match on. Meta hashes them before matching.
Create the audience.
In Ads Manager, go to Audiences, then Create audience, Custom audience, Customer list. Accept the custom audience terms, map the identifiers and select Import and create. The audience shows Populating until it is ready.
Make the list a hard control.
In the ad set, choose Further limit the reach of your ads, include the list and untick Use as a suggestion. That is the setup Meta itself describes for spending on existing customers in its guide to replicating the existing customer budget cap.
Refresh it monthly.
Meta marks a list Outdated after 30 days without an update, which may affect delivery.
Step 3 matters because under Advantage+ audience an inclusion is only a suggestion (see Facebook custom audiences: suggestion or control), and Meta's Advantage+ audience page says it isn't recommended for retargeting. Meta has also retired the existing customer budget cap in Advantage+ sales campaigns; Meta's replacement is two ad sets: one limited to existing customers with a spending limit, one that excludes them.

Drowning in Meta Ads?
Put your campaign on autopilot with Nova.
Read moreShould I exclude recent buyers from win-back ads?
Yes, in most cases. Repeat purchase Facebook ads pay off on customers who are drifting away. Someone who ordered last week is likely to reorder on their own schedule, so ads shown to them mostly claim orders. Our Facebook retargeting guide covers purchaser exclusions for site-visitor campaigns; for win-back, the same logic sets who goes in the list.
The table below is our editorial position from AdAdvisor's account-review work (8+ years in media buying, $60M+ in managed ad spend, and an ex-Meta engineer on the team). The rows are judgment calls for illustration; Meta sets no such rule.
| Who | Include or exclude | Why |
|---|---|---|
| Customers whose last order is older than your usual reorder interval | Include | They are drifting away, which is the job win-back ads exist for |
| Customers who ordered within your usual reorder interval | Exclude | They likely reorder anyway, so ads mostly take credit |
| Customers who opted out of marketing | Exclude, always | Meta's terms require you to keep opted-out people out of customer list audiences |
| The random holdout half of your lapsed list | Exclude from every campaign | It is your baseline for measuring what the ads add |
How long is "recent"? Check the typical gap between a customer's first and second order in Shopify. Fixed windows such as 0-30, 31-90, 91-180 and 180+ days appear in many AI answers but in no Meta documentation we found; your reorder cycle is a better guide.
Do win-back ads create extra sales or just claim sales that would happen anyway?
Some likely do, and the only reliable way to tell is a control group that sees no ads. Ads Manager credits a win-back ad set with purchases from people who saw or clicked an ad, including many who would have reordered anyway. The more loyal the audience, the more that attributed number likely overstates what the ads caused.
Two studies show both sides. In a randomized test of a retailer's retargeting campaign, Johnson, Lewis and Nubbemeyer found the ads lifted website visits by 17.2% and purchases by 10.5%. Across 15 large Facebook experiments, Gordon and colleagues found that common observational methods often failed to recover the true effect. Neither study looked at past purchasers, so use them as a sign of direction rather than a benchmark.
There are two ways to build the control group on Meta.
| Meta Conversion Lift | Split-List Holdout (DIY) | |
|---|---|---|
| Who randomizes | Meta splits people into test and control | You split the customer list before upload |
| Entry bar | As a guide, a campaign from the past year with USD 5,000+ spend and 500+ conversions, plus Conversions API with event match quality above 5 | A lapsed list large enough for two halves of 1,000+ customers each |
| Recommended setup | At least 28 days; for a well-powered test, USD 10,000+ spend, a 10% holdout and 500+ conversions during the test, per Meta | At least one full reorder interval |
| What gets counted | Conversions Meta matches to each group | Orders per customer in Shopify for each group |
| Main weakness | Many small stores won't meet the entry bar | Only some of your list matches to Meta profiles, so the measured effect is diluted |
Sources: About Conversion Lift, How to set up a conversion lift test and Conversion Lift best practices. In Meta's design the test group is eligible to see ads, the control group is not, and the difference in conversions is the lift.
If you don't qualify for Conversion Lift, the Split-List Holdout is a rougher substitute. It measures only the customers you uploaded, and anything that reaches the holdout half weakens it, so treat its result as an estimate rather than proof:
Split at random.
Export lapsed customers and assign each one to group A or B with a random number column, not by name or order date.
Advertise to A only.
Upload group A as the win-back audience. Add the full list (A and B) as a custom audience exclusion in your Advantage+ sales, prospecting and site retargeting campaigns, so group B isn't reached through another door. Meta's own lift guidance warns that competing ads seen during a test dilute the measured lift.
Treat both groups the same everywhere else.
Same emails, same SMS, same offers. Otherwise you measure your email program, not the ads.
Wait a full reorder interval.
Fix the end date in advance and don't stop early.
Count orders and margin in Shopify.
Ads Manager can't show the gap. Before acting, check the difference with a two-proportion significance test, which any A/B test calculator runs.
Incremental orders =
(A orders / A customers
- B orders / B customers)
x A customers
Incremental margin =
(A margin / A customers
- B margin / B customers)
x A customers
- ad spendUse contribution margin after discounts. A win-back discount also lands on reorders that would have happened anyway, and comparing margin rather than orders captures that cost.
An illustrative example (made-up numbers):
| Illustrative 60-day read | Group A (ads) | Group B (holdout) |
|---|---|---|
| Customers | 10,000 | 10,000 |
| Orders in Shopify | 600 | 500 |
| Purchases Ads Manager credits to the win-back ad set | 420 | n/a |
| Ad spend | $2,500 | $0 |
At a $60 average order, Ads Manager would show about $25,200 in attributed revenue, roughly 10x ROAS. The holdout shows 100 extra orders: $6,000 in incremental revenue, about 2.4x. At a $30 contribution margin per order and no discount, the ads earned $3,000 in margin for $2,500, a thin profit. Now add a 15% win-back discount: it takes $9 off each $60 order, and because product costs don't fall, that comes straight out of margin, about $5,400 across group A's 600 orders, which turns the campaign into a loss. Same campaign, two very different stories, and only the second one tells you whether to keep paying. For a deeper method guide, including lift-adjusted break-even ROAS, see AI incrementality testing for Meta ads.
What is a past customer worth when I set the bar?
For win-back ads, the bar is the margin on the extra orders the ads cause, not the customer's full lifetime value. Their earlier orders are already banked, and many of their future orders are likely to happen with or without ads. Judging win-back ads against full LTV counts revenue you would have earned anyway.
Our LTV-adjusted ROAS floor is built for acquisition, where a new customer's whole future value depends on winning the first order. For someone who already bought, only the incremental part is at stake.
LTV still matters in one way: if a win-back order restarts a lapsed customer's habit, the later orders count too. Capture them by reading the holdout for one more reorder interval, not by adding an assumed lifetime figure. For LTV alongside your other scaling metrics, see seven key metrics to track before scaling ads.
What privacy rules apply to customer lists?
This is not legal advice. Meta's Customer List Custom Audiences Terms require you to have the rights, permissions and a lawful basis to use the data, to leave out people who opted out of this use, and to remove anyone who opts out later. Lists can't be based on sensitive information such as health or financial data, and since September 2025 Meta has been flagging and blocking audiences whose names or criteria suggest it.
Under the GDPR, Meta's setup page says advertisers act as the data controller for uploaded lists, and California customers can be marked for Limited Data Use, a flag Meta says must be set on each audience separately. Keep opted-out people in an Only exclusions list so nobody can include them by mistake, and check the details with your counsel.
Where this guide stops
Meta publishes no recency thresholds for win-back audiences. Lists under Meta's 1,000-customer guideline may not deliver well, and a small holdout is likely too noisy to show a difference of a point or two in repeat rate. Ad tests are noisy even at scale: across 25 large retailer experiments, Lewis and Rao found the median confidence interval on ad ROI was over 100 percentage points wide. We have no measured data on how incremental Meta win-back ads are for past purchasers, from our accounts or published research; the studies above looked at site visitors. Results likely depend on your product, reorder cycle and offer, and Meta's features keep moving. Nothing here guarantees a result.
Where Nova fits
A holdout tells you once whether win-back ads pay; keeping each campaign tied to your margin every day after that is what slips. Nova, AdAdvisor's AI media buyer for Shopify stores, does that daily read, and in Copilot mode every change waits for your approval: see how Nova runs Facebook ads for Shopify stores.
FAQ
Summary
Facebook ads can bring back past customers, but the setup is the easy part. Build the audience from your customer list as a hard inclusion, keep recent buyers and opted-out customers out, and hold back a random half to see what the ads add. Judge the result by the margin on extra repeat orders, not attributed ROAS or full lifetime value. If that daily margin check is where your week runs out, Nova can take it on for your Shopify store, with every change waiting for your approval.
Sources
- Meta Business Help Centre: Customer list formatting guidelines for custom audiences (checked November 19, 2026)
- Meta Business Help Centre: How to create a customer list custom audience in Ads Manager (checked November 19, 2026)
- Meta: Customer List Custom Audiences Terms (checked November 19, 2026)
- Meta Business Help Centre: How to replicate existing customer budget cap (checked November 19, 2026)
- Meta Business Help Centre: About Advantage+ audience (checked November 19, 2026)
- Meta Business Help Centre: About Conversion Lift (checked November 19, 2026)
- Meta Business Help Centre: How to set up a conversion lift test (checked November 19, 2026)
- Meta Business Help Centre: Best practices to set up a conversion lift test (checked November 19, 2026)
- Meta for Developers: Customer File Custom Audiences (checked November 19, 2026)
- Johnson, Lewis and Nubbemeyer, "Ghost Ads", Journal of Marketing Research (2017) (checked November 19, 2026)
- Gordon, Zettelmeyer, Bhargava and Chapsky, "A Comparison of Approaches to Advertising Measurement", Marketing Science (2019) (checked November 19, 2026)
- Lewis and Rao, "The Unfavorable Economics of Measuring the Returns to Advertising", Quarterly Journal of Economics (2015) (checked November 19, 2026)
Related reading

Performance Optimization
Facebook Retargeting Ads: How to Set It Up and Make It Work (2026)
How Facebook retargeting works in 2026: audience size floors, pixel plus Conversions API, exclusion audiences that stop wasted spend, and a step-by-step setup.
Read more
Performance Optimization
Optimize Meta Ads to LTV, Not First-Order ROAS
Most Meta ads accounts are still judged against a single first-order ROAS number. For any brand with real repeat purchase or subscription revenue, that number can be too conservative. Here is the calculation that fixes it.
Read more
Strategy & Planning
Facebook Custom Audiences in 2026: Suggestion or Control Under Advantage+?
In 2026 a Facebook custom audience is a suggestion by default under Advantage+ audience. See which settings Meta still enforces, how to make delivery strict, and what the new 730-day purchase window changes.
Read more
Analytics & Reporting
AI Incrementality Testing for Meta Ads: How to Measure True Lift (2026)
Incrementality testing shows which Meta ad conversions your ads actually caused. Learn how to run Conversion Lift and geo tests, and turn lift into a break-even ROAS rule.
Read more
Strategy & Planning
Facebook Ads Targeting: The Complete 2026 Guide
Every Facebook ads targeting option in 2026: broad, interest, custom audiences, and lookalikes, what Meta removed, and a framework for choosing the right targeting for each goal.
Read more



