The engagement sieve: every churn signal, and where it went. Automation takes the many accounts that only need a nudge; people get the ones that need a conversation. Click a stage or a month to list its accounts, and an account to open its timeline.
The engagement sieve: churn signals, stage by stage
Each bar is a share of all churn signals; the figure on the right is its share of the stage above. "Called" counts a churn call, or a QBR in the 30 days after the signal.
Renewals at churn risk in the next 180 days
CRM renewal dates and today's flags only: what actually happened is on the "Did it work?" tab.
Leadership always asked me two things: what's red and what are we doing about it, and is the team in front of customers. This page answers both.
Sorted by value, which guides the work and is never anyone's credit. The segment bar applies; every number opens its accounts.
Churn: biggest ARR at risk first
Growth: biggest upside first
Team: in front of customers, against the median
Meetings held in the last 30 days and the share of each owner's accounts met in that time, against the median of owners of the same kind (CSMs with CSMs, pool members with pool members). Counts, not hours: the data has no meeting length. Pools are queues, not people, so they aren't listed.
Every task the engine sent to a person, and what became of it: Done means the right meeting happened, Resolved that the flag went away. Pick your name under Manager to act on your team's tasks: give one to someone else, or approve a hand-back. A decision is saved here and applied on the next run, when the task goes to its new owner in Slack.
Tasks
Oldest first. CS ops items (handoff gaps, no CSM) close only when the CRM is fixed, so they can't be reassigned here. A hand-back is approved by the manager it went to.
The backtest: today's churn flags against what happened at each renewal in the six months after the data. The outcomes are generated, with surprises no flag can see, so this shows the method, not proof.
Flagged vs. what happened at renewal
Lost = churned or downsized. Click a cell to list its accounts.
Warning time for the losses we caught
Days from the first signal to the renewal. Less than days is too late to act.
Did a churn call help? (Act accounts, Scale and Enterprise)
Act accounts only, so like is compared with like. Read the counts: with groups this small, chance can hide a real effect.
ARR at risk vs. ARR really lost at renewal
At risk: the data's rule ([money.at_risk] in settings.toml: the whole ARR at Act, the unused part at Watch and Shelfware). Lost: ARR minus what the account renewed at. Under 100% the rule over-calls the risk, or CS saved some of it; over 100% it's too relaxed. The renewal values are generated (a downsize shrinks to about what the account uses), so this shows the method, not proof the rule is right. Owners' forecasts join this table once their accounts renew.
One account over 90 days: what it did, what CS did, and what the engine flagged.
Summary: what was done, what's next
Usage and CS touches
Utilisation is the 30-day average share of contracted volume, so it starts on 19 August, 30 days into the data. The dashed lines are the flags' lines: under 40% counts towards churn, 80% and up towards expansion.
New Self-service and Scale accounts. Welcomed accounts got the getting-started email the day after signing; the control group is the onboarding holdout and got none. The effect in this data is planted, so this shows the method; it doesn't prove that real emails work.
Days from signing to SDK install, accounts per day
Outcome
Activated: SDK in and 2+ active users within 14 days of signing, counted only for accounts signed 14+ days ago.
Click a bar to list its accounts
Every account at churn or expansion Act with a CSM or a pool. The clock starts when its 30-day utilisation crossed the flag's line and stops at the first churn, growth or activation call. Check-ins and QBRs don't count. A call up to 14 days before the signal counts as 0 days.
Days from signal to first call, one dot per account
By owner
Most CSMs have one or two Act accounts. Read the counts: one account is an anecdote, not a pattern.
Meetings per week
All meetings held, QBRs and check-ins included. One cell is one person's week; the key under the grid gives the scale.
Click a cell to list its meetings
Capacity: who can pick up more
Each owner's whole book, whatever the Region and Tier filters, so a manager can see who has room for a task from another region. Open tasks are the ones on the Tasks tab still waiting. Below the median of the same kind of owner is marked "room".
Share of live accounts under 40% utilisation, by week
Each cell is the share of the row's accounts on that week's last day, using 30-day utilisation; hover for counts, click to list them. Accounts appear once they have 30 days of data.
Click a cell to list its accounts
Shelfware: live 60+ days, never above 40%
Flat and low, so it never shows as "falling". Enterprise and Scale get an adoption review; Self-service gets the use-case inspiration email.
Covered from another region
The owner sits in one region, the customer in another. Worth a check on language, time zone and local events.