An A/B test splits traffic evenly and reports one winner after it ends. A contextual bandit uses each user's context to choose an option for that user and shifts traffic toward better options while it learns. A/B tests answer "which is best on average?"; bandits answer "which is best for this user?"
More traffic means faster learning, but bandits waste less traffic than A/B tests because they stop sending users to clearly worse options early. Subscription apps also face delayed rewards, since renewals happen weeks after the first purchase, so the setup has to account for that.
You can. A production version needs the bandit itself, a randomized holdout to measure lift, reward handling for delayed subscription revenue, and monitoring for drift. Teams without a dedicated ML group often use a service such as Botsi instead.
The act of ending a subscription, either by the user or automatically due to billing failure. Understanding cancellation timing and reasons is key for designing retention strategies, deflection flows, and win-back campaigns.
A short feedback form presented during the cancellation process to understand why a user is choosing to leave. These insights inform product improvements, pricing strategy, and targeted win-back efforts.
The percentage of subscribers who cancel or do not renew their subscription over a given time period. Churn can be voluntary or involuntary and is a core metric for measuring retention health and forecasting revenue growth.
The ratio of users who click on a specific call-to-action (e.g., email link, push notification, paywall button) versus those who viewed it. CTR is a key engagement metric for evaluating creative performance and funnel effectiveness. Click-through-rate originated in the email and banner ad world, where everything was on a desktop. With mobile, some companies now call this same metric tap-through-rate since people are tapping on their phone vs clicking on desktop, but at the end of the day it's really the same type of measurement.
A method of analyzing user behavior by grouping users based on shared characteristics — most often the date they first installed or subscribed. This allows teams to compare retention, LTV, and monetization trends across cohorts over time.
Botsi picks the offer and paywall each user sees, from options you approve, and measures the lift against a holdout.