Two things moving together doesn't mean one caused the other.
Data shows users who use feature X retain better. Do they retain because of X, or are they the kind of user who'd retain anyway and also happens to like X? Acting on correlation as if it were causation is one of the most common product mistakes.
| Correlation | Causation |
|---|---|
| Cheap to observe | Expensive to prove |
| Found in any dashboard | Requires experiment |
| Can mislead | Can guide real decisions |
| A starting point | The destination |
"Power users use feature X, so if we push X on new users, they'll become power users." Often false — the causation runs the other way.