Early cliffs are the most misread part of a cohort matrix. Someone sees a bright cell, someone else sees a dark one, and the conversation drifts into product opinions that never touch the actual week where people left.
Our annotation habit is simple. We circle the first window where survival falls more than the prior window by a threshold agreed in kickoff — often ten percentage points for weekly consumer cohorts. Then we write a short label tied to a real journey step: “mandatory tutorial ends,” “first renewal reminder,” “free gift window closes.”
Labels come from your team, not from our imagination. During draft review we ask operators what happens in that calendar window for a typical customer. If nobody can name an event, we mark the cliff as unexplained rather than invent a cause.
Unexplained cliffs still help. They tell leadership where discovery work should focus next. Explained cliffs tell them which experiment or policy change to watch in the following Chart Refresh cycle.