WhatsApp Messaging Growth: Design and Prioritize Experiments to Lift Message Volume

You are a product manager for WhatsApp, and the goal is to raise the number of messages people send and receive on it. Which A/B tests would you run? For each one, give the hypothesis, the experiment setup and the metrics that would show whether it worked, then say which test you would run first. ```hint Break the number apart Message volume depends on how many people send, how many conversations each has and how long those conversations run. Aim each test at one of these drivers, and consider who else is affected when a treated user sends more. ``` ### Clarifying Questions - Do messages from businesses count toward the goal, or only messages between people? - Is message volume the goal itself or a proxy for engagement, so that a shift from texting to calls or reactions counts as neutral? - Which users or markets should the tests target first: light users, new users, heavy users or specific countries? - Message content is end-to-end encrypted. Should targeting and measurement rely only on metadata such as counts and timestamps? ### What a Strong Answer Covers - A precise definition of the metric and a breakdown of message volume into the drivers each test targets. - Several distinct tests, each with a hypothesis, target segment, control and treatment, and a primary metric. - Experiment design suited to messaging: spillover between senders and recipients, randomization unit, duration, novelty and variance. - Guardrails that protect user experience and trust, such as notification opt-outs, blocks, spam reports and privacy. - A prioritization of the tests and a justified first pick. ### Follow-up Questions - Treated users send more messages, but total platform volume stays flat. What could explain it? - A test lifts messages but also raises the rate at which users mute chats or turn off notifications. How do you decide? - How would you test a feature that needs the content of messages when that content is encrypted?