Case Study: AI for OTT Subscriber Churn
In one instance that I’ll use, we found a negative correlation between the frequency of cancellation and churn risk, so the group that canceled the renewed the most was less likely to turn than those cancel and renewed fewer times. So maybe there’s a limit to that behavior. In some cases, we’re digging into it right now. But again, the subscriber that comes and goes, it looks like it might be related to what we would see in the cable universe where the HBO subscriber would come and go based on a show coming and going from the service. Right to save themselves a few bucks. The other story that I love to share is it’s about unsupervised time. So we found a tiny fraction of subscribers, something like one 1/10 of a that were on a free trial promotion with a particular service. The problem is nobody noticed that the promotion wasn’t deactivated, but I did. And it flagged it and saved the customer about a million of potential revenue. So you could have detected that. But the idea and now our client can automate the promotion dates to avoid losing more revenue in the future. So the virtual.