The randomized trials behind the numbers.

Both studies were randomized controlled trials, the same method used to evaluate medicine. Customers were split at random: some received Allianz's existing, standard communication; others received the same messages rendered as personalized video. Everything else was held constant, so the difference in outcomes is attributable to the medium, not the offer.

They ran inside live renewal and onboarding journeys, so the numbers reflect what tens of thousands of customers actually did.

Between May and August 2025, Allianz split 45,685 renewal customers at random: 90% received personalized video, while a 10% control group kept the standard email. Over the 100 days that follow a price increase, the window in which customers decide whether to leave, churn fell from 12.8% in the control group to 11.4% among those who received video: a 10.9% relative reduction in a single product flow.

Each video reminded the customer what their coverage meant for them, in a format easier to notice and follow than the standard email. It was assembled from the policy itself: the customer's name, their car's brand and model, co-branded labels, policy specifics and no-claims recognition. 75% opened the email, and 41% of them clicked through. Allianz's CX lead put the result down to the way the personalized GIF, video and call-to-action worked together.

From August to December 2025, Allianz ran a second randomized trial across 1,268 newly onboarded BMW and Mini customers. This one measured comprehension and sentiment.

Net Promoter Score reached 36 among customers who recalled the video, against 13 for standard onboarding. 74% opened the email and 52% clicked through.

The onboarding star rating is a different kind of measurement. It was not part of the randomized split: Allianz compared the rating before the video was introduced with the rating after, averaged over several months, and it rose from 4.5 to 4.7 stars across ease, clarity, speed and coverage.

In 2025, Marketing Science published a randomized field experiment by Kapoor and Kumar: personalized video ads lifted engagement by six to nine percentage points over both generic video and personalized images. That is a different industry, tested with the same design, and the result points the same way.

The same renewal moment shows up in J.D. Power's 2025 U.S. Home Insurance Study: customers who fully understood a premium increase and were offered options scored 184 points higher on satisfaction (on a 1,000-point scale) than customers who got neither, and 33 points higher than customers whose premium never rose at all. Understanding the increase mattered more than avoiding one. Explaining the increase is what the renewal video did at Allianz.

Lont controls rendering and playback, so every outcome ties back to a specific template, rule and audience. That is what let Allianz run a controlled measurement. The same inputs and rules always produce the same video, AI is applied only where the team chooses to apply it, and nothing changes unless the team changes it. That control is also what lets a regulated insurer run it in production.