Note · 12 May 2026

Opens are not engagement, and the score should say so

An open is evidence that a message arrived somewhere a thumb could reach. It is not evidence that the product did a job. In App Engagement Scoring we keep opens in the dictionary, but we refuse to let them outrank a completed workflow unless the product is literally a periodical.

Teams resist this because opens are plentiful. Scores like plentiful. Plentiful numbers soothe a weekly review. They also teach growth to ship louder copy instead of a shorter task. When we demote opens in the Signal Weighting Lab, the score usually falls. That fall is not a crisis; it is the model noticing that shouting is cheap.

A workable pattern: give the open a weight of one, give a follow-through that happens inside a short window a weight several times larger, and give a follow-through that happens after the user had to hunt a weight even larger. The hunt is a product smell. Scoring it as loyalty is how you end up celebrating confusion.

If your legal or marketing colleagues need an “audience reached” figure, keep it. Put it in a distribution report. Do not fold it into the engagement score and then act surprised when the product team optimises for pings.

This is the essay we send before week two. It is not a ban on messaging. It is a request that the score remember what the user came to finish.