TL;DR
Black Friday falls on the day after US Thanksgiving, the fourth Thursday of November, which puts it on November 27 in 2026 with Cyber Monday on November 30. Every audience you plan to use that weekend is a membership list that fills as events arrive, and it cannot be backfilled from your Shopify order history the week before. A store that repairs its tracking on November 24 has a functioning pixel and an empty pool. The work that matters in September is unglamorous: confirm purchase and funnel events are actually reaching Meta, check that they carry enough customer information to match, and build the exclusion audiences now so they are populated when they start saving you money. The freeze comes later, and it is a different task.
Key Takeaways
- Set your calendar from November 27. A 60-day audience wanted on Black Friday started filling on September 28.
- Per Meta's advertiser notice of May 18, 2026, purchase-event website and app-activity custom audiences support retention up to 730 days, with existing 180-day purchase audiences migrated automatically unless an advertiser opted out [1]. Non-purchase events remain capped at 180 days [1].
- Build your buyer exclusion audience in September, so that by late November it holds a full season of purchasers you stop paying to re-acquire.
- Events that arrive without enough customer information are matched poorly, and event match quality is scored only for Conversions API website events on a 0 to 10 scale over the trailing 48 hours [3].
- Meta's Conversions API documentation describes server events as processed in the same way as pixel events at the dataset level [2]. The practical consequence for audiences is that a server path can keep pools filling when the browser path fails.
- Freezing changes and building membership are separate jobs on separate dates. September is for membership, November is for the freeze.
- Meta tracking is on the WeltPixel Conversion Tracking Plus plan at $39/mo; the free plan is GA4 only [4].
Count backwards from November 27
Work the calendar in reverse and the September deadline stops sounding dramatic.
| Audience you want on Nov 27 | Membership starts filling | What has to be working by then |
|---|---|---|
| 7-day cart abandoners | November 20 | Add-to-cart and checkout events |
| 30-day product viewers | October 28 | View content events with product IDs |
| 60-day engaged browsers | September 28 | The full browse funnel |
| 180-day non-purchase pools [1] | May 31 | Everything, already |
| Purchase-based buyer and exclusion pools | Up to 730 days back [1] | Purchase events with customer data |
The first row is the only one you could still repair in full in November, and it is the least valuable of the five. The rows that carry your Black Friday prospecting and your exclusion logic are already accruing, from whatever your store is sending today. That is the whole argument for treating September as the deadline: the pixel you fix in November is correct, and the pool behind it is thin.
There is a second reason to move now that has nothing to do with windows. Late November is the worst possible time to change tracking configuration, because verification competes with everything else and a mistake is expensive on the busiest day of your year. Doing this in September means the change has eight weeks of ordinary traffic to prove itself on.
How far back do Meta audiences remember a buyer?
Further than most retargeting playbooks assume, and the answer now splits by event type.
Per Meta's advertiser notice of May 18, 2026, audience retention for website and app-activity custom audiences built on purchase events can be set to a maximum of 730 days, and existing purchase audiences set to 180 days were migrated to the longer window automatically unless the advertiser opted out [1]. Audiences built on non-purchase events, page views, product views, add-to-carts, stay capped at 180 days [1].
For BFCM planning that split changes two decisions. A buyer-based audience, whether you use it as a seed or as an exclusion, can reach back two years, so your list of past customers is far deeper than your browse pool. And every browse-based audience is bounded by six months, so the browse data supporting a November campaign was collected no earlier than late May. If the browse funnel on your store has been quietly incomplete since the spring, that pool is smaller than the number in Ads Manager suggests, because the size figure counts people regardless of how well each one was matched.
Check your own retention settings before you plan against them. In Ads Manager, open Audiences → your custom audience → Edit and read the retention value rather than assuming it inherited anything.
Which audiences actually depend on September data?
Three, and they fail differently, which is why they are worth naming separately rather than grouping into "retargeting".
Prospecting seeds come first. A lookalike is only as good as the source list it was built from, so a purchase-based seed assembled from weeks of clean events produces a different audience than one assembled from a partially blocked event stream. You do not see this failure. You see a lookalike that performs slightly worse than last year and no obvious cause.
Retargeting pools are the visible ones. Cart abandoners, checkout initiators, and product viewers all need funnel events with product context, and they are the audiences most exposed to browser-side loss, since none of those events has an order record behind it to fall back on.
Exclusion audiences are the ones operators skip and then pay for. If your purchasers are not reliably in an audience, you spend BFCM budget re-acquiring people who already bought, at BFCM prices. That is a direct cash cost rather than a reporting inconvenience, and it is the strongest argument in this article for getting purchase events right in September. Build it now and it is fully populated by the time the auction gets expensive.
Do server-side events feed audiences the same way?
Meta's Conversions API documentation describes server events as processed in the same way as pixel events at the dataset level [2]. That is the accurate statement, and it is worth keeping at that level of precision rather than promising that a server event builds an audience identically in every case.
What follows from it in practice: a purchase event that arrives from your store's order webhook lands in the same dataset that your browser events land in, carrying the customer information the order has, which is exactly the information audience matching depends on. A browser-only setup drops the events of shoppers who run a blocker or leave before the page finishes, and those shoppers are missing from the audience, not only from your reports.
The matching side has a measurable proxy. Event match quality is available for website events sent through the Conversions API, on a 0 to 10 scale calculated on the last 48 hours of data [3]. Read it before September ends, and if your purchase score is low, work the parameters using how to get event match quality to 8 or higher. The same recovery logic that applied after the iOS tracking prompt applies here, and what iOS 14 changed and the server-side fix frames what is actually recoverable rather than promising the pre-2021 world back.
What a September gap costs in November
Run the arithmetic on an illustrative store, and keep every number in it hypothetical: 2,000 orders a month, an average order value of $80, and 25% of BFCM revenue historically driven by retargeting and exclusion-shaped campaigns.
Say the browse funnel has been losing a fifth of its events since spring, from ad blockers and browser breakage. The 60-day browse pool going into Black Friday is 20% smaller than it should be, which does not reduce sales by 20%, but it does reduce reach and it does force a wider, colder targeting mix at the exact moment auction prices peak. Call it 500 weekend orders, so a $40,000 weekend, with $10,000 of it attributed to those campaigns. A fifth of the pool is not a rounding error against that.
The exclusion side is easier to size. If 15% of your BFCM ad spend goes to people who already bought this season, and the weekend budget is $12,000, that is roughly $1,800 spent re-selling to buyers. The fix costs nothing at BFCM time because it was made in September.
Neither figure is a measurement of your store. They are there to make the shape of the loss concrete, because "your audiences will be smaller" does not move anyone's calendar.
Your September checklist
Six checks, in order, and none of them takes a developer.
- In Events Manager, open Data sources → your dataset → Overview and compare the Purchase event count for a closed 30-day window against your Shopify order count for the same window and time zone. A gap here is a delivery problem, and it is the only item on this list that blocks the others.
- In the same view, confirm the browse funnel is present: view content, add to cart, initiate checkout. An empty funnel means your browse-based pools are filling from nothing.
- Read the event match quality score on the Purchase event [3]. Meta publishes no pass/fail number; in practice a Purchase score under 6 on a Shopify store points to missing hashed email coverage, which is what the match-quality guide works through.
- In Ads Manager, open Audiences and check the retention value on every audience you plan to use in November, now that purchase-based audiences can hold a longer window [1].
- Create the purchaser exclusion audience today if it does not exist. It needs the next eight weeks to fill.
- Write down what you changed and when. In November you will be comparing against a baseline, and an undocumented September change is how a healthy store gets misdiagnosed on Black Friday.
Then stop. The next thing on the BFCM tracking calendar is not another change, it is the freeze: locking configuration, theme edits, and app installs before the traffic arrives. That is a November job with its own dates, covered in the Black Friday conversion tracking freeze.
WeltPixel Conversion Tracking sends the Meta purchase from Shopify's order webhook with the hashed customer record attached, and delivers the browse funnel server-side as well, so the events feeding your audiences do not depend on a script surviving in the shopper's browser. If you want your November pools filling correctly from September, install WeltPixel Conversion Tracking before the traffic arrives.
FAQ
Can I build a BFCM audience from my Shopify customer list instead?
You can upload a customer list as an audience, and it is a reasonable complement for buyer-based targeting and exclusions. It does not replace website audiences, because a customer list has no browse behavior in it: no product viewers, no cart abandoners, no engaged non-buyers. Those exist only if the events were captured as they happened.
How long do Meta custom audiences keep someone in them?
It splits by event type. Per Meta's advertiser notice of May 18, 2026, purchase-event website and app-activity audiences support retention up to 730 days, with existing 180-day purchase audiences migrated automatically unless the advertiser opted out [1]. Audiences built on non-purchase events remain capped at 180 days [1].
When is Black Friday 2026?
November 27, with Cyber Monday on November 30. Black Friday is the day after US Thanksgiving, which falls on the fourth Thursday of November.
If I fix my tracking in November, is it too late?
Too late for the pools, not too late for measurement. A November fix gives you accurate reporting and a working 7-day retargeting window for the weekend itself, while the 30-day, 60-day, and 180-day pools you would have spent against were filled, or not filled, months earlier.
Do server-side events count toward custom audiences?
Meta documents server events as processed in the same way as pixel events at the dataset level [2], and they carry the customer information the order holds. The reason to care is coverage: shoppers who block scripts produce no browser event at all, and the purchase itself arrives only if something outside the browser sends it.
Sources
- Meta advertiser notice of May 18, 2026, on audience retention for purchase-event website and app-activity custom audiences (advertiser communication; Meta's corresponding Help Center page could not be captured for direct citation)
- Meta for Developers, "Conversions API", developers.facebook.com/docs/marketing-api/conversions-api/, accessed August 26, 2026
- Meta Business Help Center, "About event match quality", facebook.com/business/help/765081237991954, accessed August 26, 2026
- WeltPixel Conversion Tracking, Shopify App Store listing, apps.shopify.com/weltpixel-conversion-tracking, accessed August 26, 2026