How to choose a personalized video platform.

Put the same scenario to every vendor: one real journey (a renewal, an onboarding), your actual volumes, your language list, and one mid-year product change. Ask each for a completed price table, the data-flow diagram, and a live demo of the same record rendered twice. Running the same scenario everywhere makes the differences visible in a way feature checklists never do.

Start from the six on this page, expanded to your own systems and volumes. The ones we see skipped most often: whether the system can trigger from events rather than campaign sends, whether it can run so personal data never leaves your environment, what one product change costs at full volume, whether the same record produces the same video twice, and what evidence with a control group exists that it moves renewals, activation or call numbers.

Not necessarily. Different levels of privacy protection are possible, including setups where personal data never reaches the platform vendor's servers. Ask each vendor which levels they support, and have your privacy team pick the level your data requires. At Lont, all current healthcare customers run the strictest setup.

It depends on the architecture, which is why we recommend having the same journey priced at several volumes instead of taking one quote. Per-recipient generation gets more expensive with every recipient and bills again on every change. Assembly from approved scenes concentrates the cost in creating the scenes and keeps the cost per recipient low, so high volume and frequent change favor it heavily. At low volume the difference rarely matters.

Inside the channels your message already uses: email, web, portal, app. The video should open from a per-recipient link and play on one click, with no login and no app install, because every extra step between the message and the play button loses viewers. Count the steps during the demo.

Rendering, yes: ffmpeg and similar tooling can produce video from data, and for a simple, stable, single-language flow at modest volume an in-house build can be reasonable. What teams underestimate is the rest: event triggering, per-recipient delivery, measurement, keeping languages in sync, and keeping every live video correct when products change. Price it as a system you will operate for years.

Run the first journey as an experiment: a randomized holdout that keeps getting the existing communication, one metric agreed upfront (churn, activation, call rate), and a measurement period long enough for that metric to move. That design is what made the Allianz result credible: 45,685 customers randomized, churn measured for 100 days, a 10.9% relative reduction. Any vendor should help you design this; be wary of one who resists a control group.

A generator produces one asset from a prompt: fast, creative, and the same video for every viewer. Personalized-video platforms produce a different video per recipient from your data and rules, at whatever volume your systems drive. The two increasingly work together, with generation creating the scenes and rules deciding who sees what. The full comparison is in our glossary.