Engagement event
An engagement event is a measure of whether a user keeps showing up: coming back the first week, logging in three times, sticking around. It tells the team the user is still present and still has some connection to the product, but says nothing about whether that user is getting value or would ever pay.
Definition
An engagement event is a behavioral proxy keyed to undifferentiated activity — the fact that a user is present at all, with no regard for what they did while present. It sits at the bottom of the three tiers of behavioral proxies, ordered by intent-based explanatory power.
Activity without differentiation
W0 retained, three sessions in a week, returned on day 7 — these count a user's presence and stop there. The event fires the same way for a user who opened the app, looked around, and left as for one who did the thing they signed up to do. What separates the tiers is what the underlying behavior is keyed to:
- Engagement events — that the user was present.
- Feature-usage events — which in-app feature the user engaged.
- External-impact events — an effect the user produced outside the offering.
Presence carries near-zero explanatory power about intent. Two users both retained into week one may have arrived pursuing entirely different outcomes, and the event gives you no way to tell them apart or to say why either would convert. That gap is what makes the tier weak as a candidate: interventions built on it act on a bare correlation rather than a known mechanism, so lift in the proxy has no reason to carry lift in the commercial outcome event with it.
Engagement as a health metric
The tier does real diagnostic work in its own register. An engagement event tells you the user's connection with the provider is intact and that engagement is still flowing, which is why it's the measurement site for engagement retention. What it can't tell you is what's flowing through that connection — whether the user is getting the value they came for, or whether the provider is getting the commercial outcome it's pursuing.
Designing against presence
Every converter engaged, so engagement events correlate with conversion by construction. In multiplier analysis against a signup cohort, the baseline includes accounts that signed up and never came back, so almost any trace of presence reads above 1x against first paid conversion. The correlation is genuine and identifies nothing to pull on.
The structural failure follows from that: engagement events measure what all converters did, so designing to lift them produces more lowest-common-denominator behavior without producing whatever it was that actually converted those users. Presence is necessary for conversion and gets mistaken for sufficient. Tiering is a predictive heuristic rather than a definitional gate — an engagement-tier candidate isn't disqualified from becoming a value milestone by category — but candidates at this tier rarely survive scale verification, because there's no intent-level account of the correlation to check against when it moves.
Example
A team's activation dashboard leads with W0 retention: the share of signups who return at least once in the week after signing up. The number correlates strongly with conversion, and the team treats it as their activation metric and ships onboarding reminders to move it.
The reminders lift W0 retention. Conversion doesn't follow, because the users pulled back into the product by a reminder email are present without pursuing anything in particular. The metric was reading the floor that all converters cleared, not the behavior that distinguished them from everyone else who cleared it.