Renewal is decided in
month one.

Nobody churns
at renewal.

What a new account has to do to be genuinely activated, defined per segment and instrumented so stalling is visible within days.

The sequence, the milestones and the intervention points, built so a stalled account triggers a person rather than a reminder email.

Usage trajectory, engagement breadth, support sentiment and commercial signals combined into a score that predicts renewal rather than describing the present.

What happens at each health band, starting a quarter before the date, with owners and required actions rather than a task in a calendar.

Which usage patterns indicate readiness for more, so expansion is triggered by evidence instead of by a quarterly sweep.

Net revenue retention by cohort with expansion, contraction and churn shown separately, because a single number hides the story.

Several accounts left with no warning, and afterwards the signals were obvious in the data nobody was watching.

Growth looks fine gross and flat net. Acquisition is funding a leak rather than compounding.

Every CSM runs it differently, so activation depends on who was assigned rather than on a defined process.

Upsells happen when a customer asks. Nobody knows which accounts are ready for more.

Every renewal arrives as a price conversation, which means value was never established or measured.

Investors want net revenue retention by cohort and the current reporting cannot produce it reliably.

Every account watched,not the twenty on a list.

Every account, continuously

Usage trajectory and engagement breadth read across the whole base rather than the accounts a CSM had time to look at this month.

Tickets and calls read at scale

Support conversations and call transcripts classified for frustration and risk signals, so tone is a tracked input rather than an anecdote.

A quarter of notice instead of thirty days

Decline patterns matched against accounts that previously churned, so the flag arrives while intervention is still possible.

A flag is a prompt for a person, not a verdict.

Health scores start conversations. What is actually wrong with an account, and what to do about it, is established by a human who calls them.

When health scoring
is premature

Below roughly thirty churned accounts there is no reliable pattern to learn from. Talk to the ones who left instead.

Scores with no CSM or account owner to act on them change nothing. The motion matters more than the model.

Health scoring needs behavioural data. Without product events the model scores billing history and little else.

If accounts leave because the product does not do what was promised, scoring measures a known issue rather than fixing it.

Expansion less contraction and churn, by cohort, with the three components shown separately rather than netted into one figure.

Days from signature to the activation milestone, by segment. The single strongest predictor of whether an account renews.

Whether flagged accounts churned more than unflagged ones. If not, the model gets recalibrated rather than defended.

What happened to accounts a CSM engaged after a flag, against those nobody reached. This is how the programme proves itself.

Whether healthy accounts actually expand, and whether the expansion playbook is being followed.

Accounts lost with a flag that nobody acted on. Uncomfortable to report and the most useful number here.

What retention work
depends on