Customer Hub · Customer 360
Churn calls that come with a scorecard.
Every risk score names the drivers that produced it, and every prediction is written down and graded against what actually happened — so you can argue with it, act on it, and check whether it was right.
One account, end to end: your CRM meets the graph, risk gets named drivers, and the call is written down — then graded against what actually happened. Hover any station for what's happening inside.
Everything in this walkthrough ships today — the explainable churn read, the eight customer dimensions, the public-vs-your-lens split, the prediction ledger, and the accuracy panel.
Step 1 · The account book
Your private record meets the public graph.
Two worlds fuse per account — and the boundary between them is structural, not a policy promise.
Yours — tenant-private
The relationship as your systems know it. It builds your lens and no one else's.
CRM accounts & contacts · renewal dates · health status · NPS · contract value (ARR) · activity recency · win/loss history
Theirs — from the public graph
The same company as the market knows it — the half your CRM was never going to contain.
firmographics & hiring · detected tech stack · employee-reviewed tools · reviews & sentiment · filings & funding · fiscal calendar
Live — the same signal engine
The engine that watches your competitors, pointed at the companies you serve.
leadership changes · layoffs / WARN · funding · incidents · initiative signals · plus the interviews and surveys you run
Step 2 · The churn read
A risk score you can argue with.
Risk is 0–100, and every point of it traces to a driver. That matters more than the number: a score you can't decompose is a score you can't act on.
| Driver | Why it moves the number | Read from |
|---|---|---|
| Renewal proximity | A renewal 12 days out with no activity is a different account than one 12 months out. | renewal / next_renewal date |
| CRM health status | Your team's own read, weighted — amber and red move the number, and say so. | health_status |
| NPS movement | Not the absolute score: the direction. 31 → 18 is the signal. | nps_score |
| Activity gaps | Silence is data. Days since anyone from your side touched the account. | activity recency |
| Displacement | A competitor detected in their stack or their employees' reviews — the driver your CRM can't see. | stack + review evidence |
Step 3 · The Customer Score
Eight dimensions, and two lenses.
The same discipline as the Competitive Score, pointed at your book of business. Snapshotted daily, so the trajectory — not the raw number — is what ranks your attention inbox.
| Dimension | What it reads | Example components |
|---|---|---|
| Relationship | How deep you are in the buying committee | contact breadth · leadership coverage |
| Adoption | Their stack, and what employees actually use | tech-stack breadth · reviewed tools |
| Value | What the account is worth — to you and in scale | your ARR · revenue · employees |
| Voice | What their people say in public reviews | review satisfaction |
| Risk | The composite safety read | CRM health · operational · reputation |
| Expansion | Room and reason to grow | growth signals · initiative overlap |
| Momentum | Is anything actually happening? | signal velocity |
| Fit | Match to the profile you win | account-fit / ICP components |
Public baseline vs your lens. Anyone can compute the public read. Connect your CRM and win/loss record and your lens diverges — relationship, value, and risk re-weight to your reality. Until you connect data, the two are identical by design.
Step 4 · Predict, then prove
Our AI shows its grades.
A prediction you can quietly forget is an opinion. Every churn call here is written to a ledger the moment it's made, resolved against reality, and scored — with the misses counted.
The call is recorded
Account, probability, drivers, and the date it was issued — stamped before anyone knows the outcome, so it can't be edited into looking right.
Reality closes it
A terminal health status, a renewal that lapsed with no activity, or a relationship removed. The outcome is read from your systems, not asserted by ours.
The accuracy panel
A Brier score, the number of calls resolved, and calibration — when the model says 70%, does it happen 70% of the time? On your accounts, not a benchmark.
Step 5 · Plays
From a risk number to Tuesday's to-do.
Prescribed plays
Retention and expansion plays generated from the drivers — renewal rescue, champion re-engage, whitespace pitch — each citing its evidence.
Attention inbox
Your book ranked by who needs a human this week, and why, with the churn column on the compare grid.
QBR brief
The account story assembled for the meeting — wins, risks, asks — exportable.
Deal-aware battlecard handoff
Displacement detected? A battlecard for that competitor, in that account, with the deal context preloaded.
Engagement Studio
When the data says “ask the human”, launch an AI interview or survey to the account's own users.
Alerts & digests
Watch rules route account events to Slack or email; the weekly digest keeps the team honest.
Bring one account you're worried about.
We'll show you what the graph already knows about it — the committee, the competitor inside, and the play — in a live session on your book of business.