A business adding customers quickly can show improving aggregate numbers while every individual group of customers behaves worse than the one before it. Cohort analysis is what exposes that.
Aggregates blend two different things
A total churn rate mixes long-standing customers with those who joined last month, and those two populations behave very differently from one another.
Because early-life churn is normally much higher, a business acquiring rapidly has a large share of young customers and its aggregate rate is dominated by them.
The same figure can therefore rise because retention worsened or because acquisition accelerated. The number alone cannot distinguish between two opposite situations.
Grouping by start date fixes the comparison
A cohort is everyone who joined in the same period. Following each cohort over its own lifetime compares like with like, at the same age, regardless of size.
Placing cohorts side by side then shows whether customers acquired recently are behaving better or worse than those acquired a year earlier at the equivalent point.
That comparison is the actual question behind most retention work, and it is unanswerable from any aggregate view of the customer base.
Curves flatten or they do not
Retention curves fall steeply at first and then either level off or continue declining. Where they flatten, a stable population exists that will keep generating revenue.
Where the curve continues downward without settling, every customer eventually leaves, and the business must keep acquiring at increasing volume simply to stay level.
The shape matters more than the starting point. A cohort with modest early retention that flattens is worth more over time than one that starts higher and keeps falling.
Cohorts explain why growth masks decline
New customers arriving in volume replace departing ones in the headline figures, so revenue can grow steadily while the underlying retention deteriorates each quarter.
The deterioration becomes visible only when acquisition slows, at which point the business faces both a growth problem and a retention problem simultaneously.
Cohort views give warning of this well in advance, because a worsening pattern appears in each new group long before it affects the totals.
The grouping should follow the question
Cohorts do not have to be defined by joining date. Grouping by acquisition channel, first product bought or plan chosen answers different questions about where retention differs.
Comparing those groups often shows that an apparently cheap acquisition channel produces customers who leave quickly, making it more expensive than it first appeared.
Which is why cohort analysis tends to change acquisition decisions rather than retention ones, despite being framed as a retention technique.