GA4 Predictive Audience Eligibility Checker
GA4’s predictive audiences — likely 7-day purchasers, predicted top spenders, likely churners — are some of the most valuable targeting signals Google gives away for free. But the model only switches on once you cross specific data thresholds. Enter your numbers below and see exactly which predictive metrics you qualify for and how far you are from the rest.
To unlock GA4 predictive metrics, your property must have, in the last 28 days, at least 1,000 returning users who triggered the relevant condition and at least 1,000 who did not. Purchase probability and predicted revenue require a purchase or in_app_purchase event; churn probability requires returning-user activity history. The model must also stay trained, so eligibility is re-checked continuously. This tool tells you which of the three you qualify for and the precise example gap to close.
GA4 Predictive Audience Eligibility Checker inputs and result
| Predictive metric | Status | What it needs |
|---|
How to use this calculator
- Pick your collected eventsChoose whether you collect a purchase / in_app_purchase event, returning-user activity for churn, both, or none. This sets which predictive metrics are even possible.
- Enter your 28-day countsAdd the number of returning users who triggered the condition (positive) and who did not (negative) in the last 28 days. Pull these from a GA4 exploration.
- Read the eligibility tableSee which of purchase probability, churn probability and predicted revenue you qualify for, and what each one still needs.
- Close the gapIf you are short, the tool shows exactly how many positive and negative examples you lack. Grow qualifying volume or fix the missing event.
- Re-check and activateRe-run as volume grows; once eligible, build predictive audiences and export them to Google Ads. Export your status with the buttons below.
RGM Expert Says
Clients ask for predictive audiences long before their data can support them, and the disappointment is avoidable. The thresholds are not arbitrary: GA4 needs roughly a thousand examples on each side of the line — users who converted and users who did not — within a 28-day window so the model has a balanced training set. A property with plenty of purchasers but almost no measured non-purchasers fails just as surely as a low-traffic one. We run this check first so the conversation is about closing a specific gap, not waiting and hoping.
The most common blocker we see is not traffic, it is the missing or malformed purchase event. If revenue fires under a custom name, GA4’s prediction engine never sees it and purchase probability stays dark forever. So the first remediation step is almost always tagging, not acquisition — which is why we pair this checker with the event-name validator on the same engagement. Get the event right, then the volume thresholds become a marketing problem you can solve.
Once a property qualifies, the payoff is concrete: export ‘predicted top spenders’ or ‘likely 7-day churners’ to Google Ads and let bidding chase value instead of raw conversions. We treat eligibility as a milestone in the measurement roadmap, not a switch — because if volume dips below the threshold the predictions go stale and quietly stop refreshing. Monitoring eligibility is part of keeping the audiences alive.
How it works
The checker compares your inputs against Google’s published minimums. A metric is eligible only when the required event is collected and both the positive and negative example counts clear 1,000 within the trailing 28 days. Predicted revenue inherits purchase-probability eligibility.
- Positive examples — returning users who triggered the predictive condition (purchased, or were active) in the last 28 days.
- Negative examples — returning users who did not trigger the condition in the same window; the model needs both sides.
- 1,000 threshold — Google’s documented minimum for each side before a model trains.
- 28-day window — eligibility is evaluated on a trailing 28 days, so it can turn on and off as volume moves.
- Required event —
purchase/in_app_purchasefor purchase and revenue; activity history for churn.
Thresholds per Google’s GA4 predictive metrics documentation. Google states the model also needs to meet a minimum prediction quality and that eligibility is re-checked over time — this tool reflects the documented volume and event minimums, not Google’s internal model-quality test.
Why eligibility is the gate to free predictive targeting
Predictive audiences let you bid on who is about to act, not just who already did. ‘Likely 7-day purchasers’ and ‘predicted top spenders’ export straight to Google Ads, where value-based bidding can chase future revenue. That is a meaningful edge, and it costs nothing beyond meeting the data bar — which is exactly why knowing your eligibility status is worth more than guessing.
The thresholds exist to protect prediction quality. A model trained on too few examples, or on a one-sided dataset, would produce noise dressed up as insight. By requiring a balanced 1,000-and-1,000 set in a recent window, Google ensures the predictions reflect current behavior rather than a stale or skewed history. Understanding that logic tells you how to fix a failing check: grow the thinner side of the ledger, not just total traffic.
Eligibility is dynamic, and that catches teams out. A property can qualify during a seasonal peak and lose eligibility in a quiet month as the trailing window thins. If your predictive audiences suddenly stop populating, re-run this check before assuming a bug — the likely cause is volume falling under the threshold. Treat the 1,000-and-1,000 bar as an ongoing operating minimum, not a one-time unlock.
Go deeper on measurement
Keep learning
Common questions
What are the requirements for GA4 predictive metrics?
purchase or in_app_purchase event; churn probability requires returning-user activity history.Why are my predictive audiences not available?
purchase event, or not having 1,000 positive and 1,000 negative returning-user examples in the trailing 28 days. This tool shows which threshold you miss and by how much.What is the difference between purchase and churn probability?
Does predicted revenue have its own requirement?
Can a property lose predictive eligibility?
Do I need a purchase event for churn prediction?
purchase / in_app_purchase event for purchase probability and predicted revenue.