VariFlight sold insurance against a takeoff time it predicts itself

Flight-delay insurance normally had to be bought a day ahead and paid out based on arrival time, which made it clumsy and hard to price. VariFlight's data changed the underwriting basis.

The new policy sold in the airport, not a day early, and even allowed a traveller to buy it after the flight was already known to be delayed. It paid out on a different trigger: the app's predicted departure time. If the real departure passed the predicted one by even a second, the traveller collected.

That prediction is fed by airport status, the inbound aircraft, weather and air-traffic-control data, and it is a live, changing number. VariFlight's CEO says the prediction is accurate to within five minutes 95% of the time, so the insurer is underwriting a small, well-quantified tail rather than a vague one. Payouts peaked at about 1,000 yuan.

Sell insurance on your prediction, not the event

Underwriting on predicted departure, not arrival, makes the trigger immediate and unambiguous.

A 95%-within-five-minutes forecast lets the insurer price the remaining 5% risk precisely.

Selling at the airport (even after a delay is known) captures demand the day-ahead product missed.

The app's real-time model turns raw air-traffic and weather data into an instantly usable product.

When you can predict an outcome this accurately, the prediction becomes the product. Pivot a vague risk into a sharp, cheap-to-price bet and the ordinary buyer can afford to take it.

The product became an often-cited example of big data changing insurance design — moving from static risk pricing to a live, multi-dimensional forecast. VariFlight kept investing in prediction accuracy, partnering with airports to raise on-time performance, and the prediction-based underwriting approach spread to other instant purchase, data-priced insurance products in China.