WTF Is Activation and Why Should Engineers Care?

This is the best read on product activation I've ever seen. It lays out that most of the time we're just guessing what the best activation metric might be. Once you have a guess, run a regression analysis and see how well historical data matches your hypothesis.

From the article:

Our process starts by gathering a big list of events we think might "hook" users into our product. We get these by looking at usage of that specific product, and by asking customer-facing teams what they think.

 
Next, we create groups of 3-5 events to test together. Typically, we aim for 5-10 different groups, some of which compare how often an event is completed – e.g. watched five replays vs. watched one replay.

Once you have these groups, calculate how many activated (usually within 14 days) and how many were still active three months later:

In the above hypothetical example, three companies activated but only two were still using the product after three months, and were therefore retained. This means we have a retention rate of 66% for activated companies:

• If our average retention was lower than 66%, this would be a good activation metric. 🥳

• If our average retention was higher than 66%, this would be a bad activation metric. 😭

We repeat this process for each potential activation metric. The one with the best retention rate (and a large enough sample size) wins.