Sustaining churn
Sustaining churn is the slow, steady loss of users that follows the steep early drop-off. The cohort that remains has filtered down to a more stable group, and the people who leave from here tend to do so because of changing needs, budget cuts, or gradually drifting out of the use case, rather than for the failure-to-get-going reasons that drive most early losses.
Definition
The flatter portion of a signup cohort's drop-off curve, after the freefall period of starting churn has filtered the cohort down to a more stable population.
Where it sits on the survival curve
Plot retained users or revenue from a signup cohort over time. The curve typically falls steeply in the early days or weeks, then bends and continues falling more gradually. The steep portion is starting churn; the gradual portion is sustaining churn. Both are losses from the same cohort, but they're governed by different mechanisms and respond to different interventions.
Blended metrics like monthly churn or net churn collapse the two regimes into a single rate. The blend hides that most actionable loss in self-serve SaaS is concentrated earlier, and it muddles diagnosis when the rate moves — a change in the headline number could reflect starting-churn dynamics, sustaining-churn dynamics, or both, and the blended view can't tell you which.
The cohort has formed its evidence base
In starting churn, users are still establishing whether the offering will produce what they came for. By the time sustaining churn is operating, that establishment has happened — the users who remain have decided, provisionally, that the offering does enough of what they want to keep using it. Their evidence base is larger, individual events weigh less, and their prediction about whether continuing engagement will pay off is stable enough to absorb interactions that a starting-churn user would have left over.
The same offering serves a structurally different population in this period. Same users, but operating with much more context about what to expect.
Losses run on accumulated prediction-update
Sustaining-churn losses don't typically follow from a single bad experience. They follow from small intent-satisfaction failures — hopscotch trust moments that produce "hmm, that was not so great" rather than clean traversal — which each fall below the threshold a stable user would leave over, but compound across months into long-arc trust erosion.
By the time a user churns in this regime, the original failure may be months old. Losses surface long after the events that caused them, which is part of why sustaining churn is harder to diagnose with the same apparatus that works in starting churn — the contemporaneous reading between behavior and outcome is much weaker when the causal chain stretches across months.
Example
A B2B SaaS cohort of 1,000 signups. By day 30, 650 have dropped — most never converted, others paid for one month and left. From month 2 onward, the remaining 350 lose roughly 1-2% per month: about 80 gone by month 6, another 50 by month 18.
The day-0-to-day-30 freefall is starting churn. The shallower losses from month 2 forward are sustaining churn. On a survival curve, the inflection between them is visible as a knee — the curve bends from near-vertical to nearly-flat. A blended monthly churn rate computed across the full 18 months would average these regimes together and hide that roughly two-thirds of total loss happened in the first 30 days.