How would you figure out if it makes sense for FB to run Whatsapp and FB Messanger as two separate apps or merge them into one?

This is a common challenge data scientists face in large companies having different, but somewhat similar, products (Messenger-Whatsapp, Airbnb-Hotel Tonight, etc.). Should the different products be merged into just one product/app or live separately? From a data science perspective, how to estimate whether the main metric (engagement, revenue, etc.) will be higher with two separate apps or just one?

We can safely guess that many users use both apps. Whatsapp has almost 3 billion users and Messenger 1 billion or so. It would be hard to imagine those numbers are achievable with no overlap in user base.
Assuming we have full data about user behavior for both apps, the main question we want to answer is how users interact with them. What are the differences in usage? Why are they switching from one to the other one? Are they taking different actions when chatting on Messenger vs Whatsapp? Are they chatting with the same people or different ones? Are chats longer on one app? Essentially, do the different apps serve different purposes or it is fairly irrelevant for people whether they use one or the other one?

The more overlap there is in user behavior across the apps, the safer would be to merge them. If users are using the apps in different ways, merging them would be very complicated and risky. For instance, currently Whatsapp makes it easier to take and send pictures or to make phone calls. Messenger makes it easier to post a story. Which feature would take priority in a hypothetical final app?

Eventually, you want to treat this just like any other data science problem. Most of data science is based on the idea of trying small and incremental changes via A/B testing. Create a clearly defined hypothesis, test it, look at the results, iterate. There is no reason why here the approach should be different. A major sudden change like merging the apps would imply so many changes, effectively making it impossible to test any small hypothesis independently.

Instead, it would be much more sound and safer to go step by step, one small test at the time, trying to make the two app UI/UX become more similar. If, as the apps start converging, you can show that engagement goes up, keep iterating. If, as you test feature by feature trying to make the apps more similar, engagement drops, then figure out why and either try different tests or, in the extreme case, give up on the idea and just develop the two apps separately.

Example: currently, posting a story on Messenger is one tap. It is two taps on Whatsapp. Use current data to identify the impact of that feature on, say, engagement. Then decide if you want to test making it one tap on Whatsapp too or 2 taps on Messenger. In any case, test a UI change such that it becomes the same number of taps in both apps. If engagement wins after the test, keep iterating moving to a new feature that’s implemented differently in the two apps. And so on.

For a tool that can answer any custom product case study question, check out the DataScienceGPT app.