Ground truth where it counts. Modeled extension across the fleet.

In 2020, Coinstar operated tens of thousands of kiosks embedded inside grocery stores nationwide. The reach was undeniable, but deploying sensors on every unit was neither practical nor necessary.

Still, without credible audience metrics, advertisers remained hesitant to commit budgets at scale. The question was: how do you bring trusted audience measurement to a network too large to fully sensorise?

To address this, Coinstar's adPlanet media network selected Quividi's Sampling & Extrapolation approach to balance verification and scalability.

Ground-truth audience measurement was deployed on a representative sample of kiosks. An audience model anchored in observed behaviour then extended those observations across the full network.

Crucially, the model was designed to remain explainable and defensible: not abstract projections, but extensions of verified data.

Quividi's high-fidelity impression multipliers let adPlanet transact its inventory programmatically, and the sampling approach meant deployment took weeks and saved on capital costs.

It was the first ground-truth audience model designed for a retail media network of more than 10,000 screens.