The salad customer's real rival is a burger. Your order data will never tell you.
Seed a cold-start recommender from your own order history and it makes the obvious bet: the salad customer wants another salad. Off-platform consumer signal says otherwise. In one home metro, the customers of a national salad chain cross-shop a fast-food burger chain more than they cross-shop the nearest direct salad rival — 5.8% vs 5.2%, the ordering literally flips. In another city, a burrito chain's customers cross-shop the burger chain at 16% while the nearest same-cuisine rival sits below 1% — a 17x gap, with no direct competitor anywhere in the top 14. The same pattern holds in a third cuisine, same city: a national coffee chain's customers cross-shop a burger chain more than any direct coffee rival (15.7% vs 10.3%). Three categories. Two cities. The indirect rival beats the direct one every time. First-party data cannot surface this. It only records the orders that already happened on your app, never the burger the salad customer chose somewhere else. That blind spot is exactly where cold-start goes wrong — and it is exactly what off-platform consumer cross-visitation fixes.
Similar cuisine → similar customer
A cold-start recommender built on first-party history buckets customers by cuisine: salad customers get salad suggestions, burrito customers get burrito options. The assumption is natural. It is also wrong. Your app only records what already happened on it — never the off-platform choice that reveals the real consideration set.
The real rival is cross-cuisine — and ranked first
For a national salad chain in its home metro, the #1 consumer cross-shop rival is an indirect fast-food brand — sitting above the nearest direct salad rival (5.8% vs 5.2%). For a burrito chain in another city, every one of the top 14 cross-shop rivals is indirect; the nearest same-cuisine rival ranks 15th, at 0.96% against a 16% indirect leader. A national coffee chain in that same city: the burger chain again ranks first, ahead of the nearest coffee rival (15.7% vs 10.3%). The correct first-tap recommendation requires knowing the whole market, not just your funnel.
- Cold-start recommendations — built on the real consideration set, not cuisine buckets
- Merchant acquisition — sign the merchants your customers already choose off-platform
- New zone launch — map demand before a single order is placed
- Newly-signed merchant targeting — find the right first customers before order history exists