The most uncomfortable number in your customer base

Churn dashboards are good at one thing: detecting decline. An account used the product a lot, now uses it a little, the alert fires and someone picks up the phone. The problem is the customer who never had any usage to decline — and who, for that very reason, is the most dangerous account in your base.

Two accounts, one wrong signal

Consider two paying customers today:

• Account A: 40 uses a week. Two weeks ago it started dropping — last week, 9 uses. A 78% decline. Your dashboard raises an alert.
• Account B: pays $169/month. Three logins in total. Active for seven months.

Intuitively, you bet A will cancel first. But A has history, context, someone to call. B never built any value — calling them means making a brand-new sale to someone who has already said "no" seven times in a row.

Why B drops off the radar

The typical usage-decline formula is:

usage(last 7 days) ÷ usage(previous 7 days)

For Account B, that is 0 ÷ 0. The number simply does not exist in the system.

It is not a bug, it is design. Measuring decline assumes a baseline. Accounts that never activated have no baseline, so they disappear by construction.

The calculation that hurts

List your active paying accounts. For each one, find the date of the first meaningful use — not a login, actual use. Add up the MRR of every account without that date.

In SQL, assuming a subscriptions table and a product-events table:

select count(*)   as accounts,
       sum(s.mrr) as mrr_without_real_usage
from subscriptions s
where s.status = 'active'
  and not exists (
    select 1
    from events e
    where e.account_id = s.account_id
      -- what counts as real usage in your product
      and e.type in ('report_created', 'integration_connected')
  );

That is the most uncomfortable number in your base — and the one almost nobody has ever calculated.

Worse: when these accounts cancel, they do not leave saying "it stopped making sense". They leave saying "I paid for seven months and never used it". That is not churn. It is reputation quietly eroding.

Further reading

I wrote the long version of this argument on the Keep blog — Keep is a product of mine that detects exactly this pattern by crossing Stripe data with real usage: Pays and never used.

The broader pattern — numbers nobody calculates, controls nobody tests — is in 45 minutes, $460 million: when technology governance is missing.