Circana Turns Purchase Data Into Predictive Audience Signals for Activation
The Development
Circana, LLC launched predictive commerce signals inside Circana Complete Audiences, using AI and its commerce intelligence to flag households likely to enter a category, increase spend, switch brands, or disengage. Advertisers can activate and measure these audiences against sales, category growth, and retention goals. The signals draw on 129 million modeled U.S. households, nearly 2.6 million UPCs, and 500 billion deterministic signals. In a test across 13 brands, applying the signals improved performance in every campaign, with gains from 11% to 36% and a median uplift of 22%. Agency Tinuiti, which cited its Adaptive Commerce approach, endorsed the launch. Circana did not publish test methodology, control design, or the campaign metric measured.
Our Take
The strategic point is not the AI label. It is that purchase-grounded data owners are moving up the stack from measurement into targeting, and they bring the one thing walled gardens cannot: transaction truth outside their own ecosystems. Predicting switchers and lapsers is where CPG and retail media budgets actually leak, so this lands on real money. Treat the uplift figure as a hypothesis. Thirteen brands, vendor-selected, no disclosed controls, and an undefined performance metric is a case study, not proof. The right response is neither dismissal nor adoption on faith. It is a structured test that makes Circana's claim earn its place against your current audience stack.
What Changed
Advertisers can now buy audiences defined by forecast purchase behavior (category entry, spend expansion, brand switching, disengagement) rather than historical purchase segments. Prediction, activation, and sales measurement sit in one workflow built on deterministic, household-level transaction data rather than modeled interest or intent proxies.
Marketing Impact
Media and shopper marketing teams gain a way to spend against predicted demand shifts, such as pre-empting competitor-bound switchers or reactivating likely lapsers. Measurement teams get activation and sales readout under one data foundation, shortening the loop between audience decision and business result. Brand planning shifts from category-history targeting toward forward-looking segments.
Competitive Implication
Transaction-data owners and agencies with commerce-first practices gain, because predictive audiences are only as good as the underlying purchase graph. Vulnerable are audience vendors built on modeled intent or third-party interest segments, and retail media networks whose data is confined to a single retailer's shoppers and cannot see cross-retailer switching.
Strategic Outlook
Expect NielsenIQ, Numerator, and retail media networks to answer with their own predictive audience products within two to three quarters, and expect uplift claims to escalate before standards emerge. By mid-2027, buyers will demand holdout-based incrementality as a condition of purchase, separating signal vendors with real lift from those relying on case-study averages.
The Exploit
Action Item
CPG and retail media leads should run a geo- or holdout-controlled test of Circana's switcher and lapser signals against their current best audience in Q1 2027, requiring incremental sales lift, not ROAS, as the success criterion.