How to Build GA4 Predictive Audiences From Shopify Purchases Before Black Friday

TL;DR

Predictive audiences such as "Likely 7-day purchasers" only exist for GA4 properties that clear Google's purchaser threshold [1], and that threshold turns into a store-side test you can run in Shopify in two minutes. If your busiest recent week had well under 1,000 orders, GA4 has nothing to train on. Stores over the line build two audiences, let them share into Google Ads, and judge them by the Google Ads size estimate. Stores under it build the same kind of list by hand, in the same screen.

Key Takeaways

• Count orders by week in Shopify first. If every week of the last 28 days sits well under 1,000 orders, plain arithmetic from Google's purchaser rule says GA4 can't build purchase predictions yet.
• Confirm in GA4 at Admin → Data display → Audiences → New audience, where predictive suggestions that meet the prerequisites carry a Ready to use label [2][3].
• If eligible, build "Likely 7-day purchasers" (purchase probability above the 90th percentile) and "Predicted 28-day top spenders" (predicted revenue above the 95th percentile) [2].
• Judge each audience by its Google Ads size estimate. Google Ads needs at least 100 active users in the last 30 days before it serves a list [4][5].
• Purchases the browser never reported reach GA4 without a browser cookie behind them. They count in GA4's member total, and in the Google Ads estimate only where Google can match them without a cookie [4].
• Build in October. GA4 prepopulates a new Google Ads list with up to 30 days of data it already collected [3], but the model's quality has to be "sustained over a period of time" first [1].
• Under the threshold, "started checkout, didn't buy" needs no model and takes five minutes to set up [3].

Does your store clear GA4's bar?

Google's prerequisite, in full: "In the last 28 days, over a seven-day period, at least 1,000 returning users must have triggered the relevant predictive condition (purchase or churn) and at least 1,000 returning users must not. Model quality must be sustained over a period of time to be eligible." [1] The purchase events also have to carry value and currency [1].

A thousand purchasers in seven days takes at least a thousand orders in seven days. That floor is arithmetic from Google's rule, not a number Google publishes, and it only works in one direction. A store under it can stop here. A store over it still has to pass the returning-user and model-quality conditions, which you can't check from outside GA4, and repeat buyers push the real requirement higher, because ten orders from one customer are one purchaser.

The Shopify side of the check is a data exploration, which Shopify offers on all plans [6]. Go to Analytics → Reports, click New exploration, paste this and click Run:

FROM sales
  SHOW orders
  TIMESERIES week
  SINCE startOfDay(-28d) UNTIL today
  ORDER BY week ASC

orders counts each order once [7]. If every week sits well under 1,000, the answer is no for now. If one week comes close, let GA4 decide: Google's window is any seven days, and a rolling week can straddle two of these calendar buckets.

Only orders whose purchase reached GA4 count toward the thousand. WeltPixel Conversion Tracking sends the GA4 purchase from the server with value and currency on every plan, and your plan and settings change which orders arrive. On the free plans, admin and API orders go to no destination, and the monthly order allowance is far below a thousand a week anyway. On Plus, admin, POS and subscription renewal orders are sent unless you've turned on Exclude Admin/API Orders for GA4. And if you use the optional "Skip purchase events for visitors who declined analytics" setting (GA4 only, every plan, off by default, and active only while Respect Customer Privacy Settings is on), those purchases never reach GA4, so they don't count either.

Then confirm in GA4. Open Admin → Data display → Audiences → New audience [3] and go to the predictive section of the suggested audiences. Google's wording: "Suggested predictive audiences that meet prediction-modeling prerequisites are labeled as Ready to use" [2]. Each prediction has its own status there [1], so purchase probability can be ready while predicted revenue isn't. Google also recommends turning on the Modeling contributions & business insights data-sharing setting [1].

The test runs per property. A business that splits its stores across several GA4 properties needs a thousand purchasers in each one.

Which two GA4 predictive audiences pay off on November 27?

Google offers five predictive templates [2]. Two of them suit a sale week.

Purchase probability is "the probability that a user who was active in the last 28 days will log a specific key event within the next 7 days" [1], and only purchase, ecommerce_purchase and in_app_purchase feed it [1]. The app sends GA4's standard purchase, so nothing needs renaming. Membership moves daily: GA4 recalculates predictions for each active user once a day [1].

Eligibility can also lapse. "If the model quality for your property falls below the minimum threshold, then Analytics will stop updating the corresponding predictions and they may become unavailable" [1]. A property that is Ready to use in October can lose it in a quiet week. Open the audience builder again in the week of November 23 and check the label before you build a campaign around it.

Sharing needs no export step. "Predictive audiences are automatically shared with any advertising accounts you have linked to your property" [2], as long as the Google Ads link keeps Enable Personalized Advertising on [3]. Meta audiences run on their own calendar, which the September article on retargeting audiences lays out.

Read the Google Ads estimate, not the member count

GA4 shows two sizes for every audience. Measurement size is everyone who met the criteria. The Google Ads estimate is "the portion of your Google Analytics audience that is actually eligible for remarketing", because "a user who is part of your Google Analytics audience might not be remarketable due to privacy consent choices, cookie availability, browser limitations, or lack of Google identifier matches" [4]. You've probably seen this already: a healthy audience in GA4 and a much smaller list in Google Ads, with nothing broken on either side.

The estimate appears in the audience builder once the property is linked to Google Ads with ads personalization on [4]. It forecasts list size 30 days ahead as a 60% confidence range, with a colour for the risk of being too small: green below 8%, yellow from 8% to 20%, red above 20% [4]. The floor behind the colours is Google Ads' own: at least 100 active users in the last 30 days for Search, Display and YouTube lists [4][5].

On a Shopify store, reachability starts with whether a browser stands behind the purchase. On a normal web order, the app's server purchase reuses the GA4 client ID the shopper's browser already had, so the buyer lands on the same GA4 user that browsed, cookie included. When the browser never reached the app, the purchase still goes to GA4, under a stand-in identity built from the customer or the order. That user has no browsing history and no browser cookie. The client ID article explains the stand-in. These purchasers add to measurement size, and they reach the Google Ads estimate only where Google can match them without a cookie [4]. Whether they count as returning users for the model, Google doesn't say.

Consent shrinks the reachable share too, and the June 2026 consent change covers what changed for GA4 data headed to Google Ads and what to audit. Google's own list of fixes for a small estimate: widen the conditions, extend membership duration up to 540 days, implement user-provided data collection, and make sure consent mode is set up correctly [4].

When does the work have to happen?

Sooner than the audience itself needs. GA4 doesn't start from zero: "When you create an audience, Analytics adds any users who have met the audience criteria during the last 30 days. This applies to Google Ads audiences only." [3] New members then take "24-48 hours" to show up [3]. An audience built on November 2 starts its Google Ads list with users who met its criteria in October, when GA4 has that data.

That prepopulation draws on what GA4 already collected. It never reaches into Shopify's order history, so a purchase that didn't reach GA4 in October can't be added in November.

Three things can't be hurried. The model needs quality "sustained over a period of time" [1]. The Google Ads link and its personalization switch have to be on before any audience shares. And a list needs its 100 active users before it serves [5].

So build the audiences this month, re-check the Ready to use label late in November, and leave the purchase stream alone once the tracking lock arrives on November 16. The tracking freeze calendar and the BFCM readiness hub set out what locks when.

If GA4 says no, build these by hand

Rule-based audiences use the same builder with no model behind them. Google Ads still applies its 100-user serving minimum [5]. Three are worth having before the sale:

1. Started checkout, didn't buy. Add a condition on the begin_checkout event. Click Add group to exclude, choose Exclude from audience temporarily, and set the condition to the purchase event [3]. Membership of 30 days fits a sale campaign. The app sends begin_checkout from the browser when that event is on in its settings, and browser events respect Shopify's Customer Privacy API signals, so this list holds visitors whose analytics consent let the event fire.
2. Bought before, not lately. Include users with a purchase event at any point. In an exclude group set to Exclude from audience temporarily, with the scope Across all sessions, add purchase with event_count above 0, turn on Time Period, choose True at the user's last interaction and enter 60 days [3]. Set membership up to the 540-day maximum [3].
3. High lifetime value. Use a metric condition on LTV with your own threshold, for example twice your average order value. Google's builder accepts time-windowed metric conditions like "LTV > 5 in any 7 day period" [3].

Stand-in purchasers land in the second and third lists like anyone else, and they face the same cookie gap in the Google Ads estimate. If the estimate comes back red, widen the condition before you lower your expectations of the channel.

FAQ

How many orders does a Shopify store need for GA4 predictive audiences?

At least 1,000 orders inside some seven-day window of the last 28 days, because Google requires 1,000 returning purchasers in that window [1]. That's a floor worked out from Google's rule. Clearing it doesn't guarantee eligibility, since the model-quality and returning-user conditions still apply.

Do GA4 predictive audiences fill with past data?

Partly. GA4 adds users who met the criteria in the last 30 days, and Google says this "applies to Google Ads audiences only" [3]. It uses data GA4 already collected, never Shopify's order history.

Why is my Google Ads list smaller than my GA4 audience?

GA4 counts everyone who met the criteria, while Google Ads counts only remarketable users, after consent, cookie availability, browser limits and identifier matching [4]. Purchases that reached GA4 with no browser cookie behind them sit in the first number, and in the second only where Google can match them without a cookie.

Can a predictive audience stop working?

Yes. If model quality falls below Google's threshold, GA4 stops updating those predictions and they may become unavailable [1]. Check the Ready to use label again in the week before you rely on it.

Feed GA4 the purchases it predicts from

WeltPixel Conversion Tracking sends Shopify purchases to GA4 from the server with value and currency, on the shopper's own GA4 client ID when the browser shared it. Run the weekly order count, then look for the Ready to use label in the audience builder.

Sources

1. Google Analytics Help, "[GA4] Predictive metrics", support.google.com/analytics/answer/9846734, accessed October 6, 2026
2. Google Analytics Help, "[GA4] Predictive audiences", support.google.com/analytics/answer/9805833, accessed October 6, 2026
3. Google Analytics Help, "[GA4] Create, edit, and archive audiences", support.google.com/analytics/answer/9267572, accessed October 8, 2026
4. Google Analytics Help, "About Google Ads list size estimates in Google Analytics", support.google.com/analytics/answer/17456279, accessed October 8, 2026
5. Google Ads Help, "About audience segments in Audience manager", support.google.com/google-ads/answer/7558048, accessed October 6, 2026
6. Shopify Help Center, creating a new data exploration, help.shopify.com/en/manual/reports-and-analytics/shopify-reports/report-types/custom-reports/create-custom-explorations, accessed October 6, 2026
7. Shopify Developer Docs, "sales" schema and "SINCE, UNTIL, DURING" syntax (API version 2026-10), shopify.dev/docs/api/shopifyql/latest/schemas/sales_revenue/sales, accessed October 8, 2026