Validate the prediction.
Compare saved forecasts with mature outcomes. Check data coverage and model stability across cohorts.
Explore ROAS ForecastTwo customers can look identical on day one. Value Signals reads their early behaviour, predicts what each could be worth and sends that value to your ad platform as an agreed event.
Customer A and Customer B both make a $25 first purchase. Over seven days, A is active on 2 days with 4 sessions and B on 5 days with 12 sessions. Predicted 180-day values: $120 and $360. Each prediction becomes an agreed value event. Synthetic example; no data is sent.
Four steps from a customer’s first week to a bidding input. Your team agrees each definition before anything is sent.
Sessions, active days and purchases in the first days after acquisition.
From your data
Atlas estimates what the customer could be worth over the agreed horizon.
Atlas
The prediction maps to the outcome, event and value format your team approved.
Agreed together
The ad platform receives a value-related signal it can use when optimising.
Your ad accounts
Your team keeps control of campaigns, budgets and the platform setup. Atlas supplies the signal.
Check signal volume, uploaded value and delivery health. Look into gaps before judging the campaign test.
Uploaded value is the sum of predictions sent. Revenue and return are measured separately.
Delivery health: no gaps recorded in this sample week.
Fixed totals for 1,400 sample signals, separate from the customers above.
A signal test has two questions. Is the prediction sound, and does the campaign do better with it?
Compare saved forecasts with mature outcomes. Check data coverage and model stability across cohorts.
Explore ROAS ForecastMeasured in a Google Ads signal A/B test. Your team reviews the agreed campaign outcomes and decides whether to expand.
Outcomes vary by team and setup.Tell us about the customer outcome you want to predict and the platform setup you use.