The account that cancels in March usually warned you in November, just not anywhere your CRM looks. It filed four tickets about the same broken export, its replies went from "thanks!" to "any update?", and it asked twice for SSO. A churn risk score rates how likely each account is to cancel or shrink at renewal, and one built from your support desk and product usage catches that pattern while there is still a renewal to save.
This guide builds one in Oneprofile. You import your Zendesk tickets, product usage, and Stripe billing next to your CRM accounts, and AI reads every ticket for tone and feature asks. It then scores each account for churn risk and expansion likelihood, the two halves of a customer health score, and writes the reason in a sentence.
Both land on the CRM record the account owner already opens. It is the same import, read, and export loop that keeps the rest of that record current in CRM contact enrichment.
To build it on your own renewals as you read, connect your CRM and support desk in a free workspace first.
What goes into a churn risk score
Not every signal deserves a place. The ones below share two traits: each shows up before the cancellation, and none can be bought from a data company, because they only exist inside your own stack.
Signal | Where it lives | What it usually means |
|---|---|---|
Sentiment souring across tickets | Support desk | Patience is running out, even at flat volume |
The same problem reported again | Support desk | A bug they no longer believe you will fix |
Repeated asks for a missing feature | Support desk | They are looking at a product that has it |
Weekly active users or seats falling | Product analytics or database | The team stopped relying on you |
Subscription past due or cancelled | Billing | Budget pressure, or an exit already underway |
Main contact inactive or gone | CRM and product usage | Nobody left who chose you |
Ticket count is missing on purpose. A customer who files a lot of tickets usually still wants the product to work. The accounts worth worrying about are the ones whose tickets got shorter, or stopped, while logins fell.
A tangent, but NPS earns its place on this list less than people think. A detractor score arrives after the decision has been made, from whoever bothered to answer, and it rarely says what to fix.
Pull support tickets and product usage together
Everything below runs on direct connections, and each source gets a table of its own. Start with the smallest useful set of accounts:
Bring your accounts in. Connect your CRM, whether that is HubSpot, Salesforce, Attio, or something else, import your companies, and filter to the accounts renewing in the next two quarters. Save that filter as a run condition, so the scoring step only runs on renewals. Starting there keeps the first run small and makes every score matter.
Add ticket history from your support desk. Connect Zendesk with your subdomain, an email, and an API token, then import its tickets. Subject, description, status, priority, tags, and dates come with every ticket. On the tickets, add Link to another table and point each ticket at its account, matching on a value both tables hold, such as the Zendesk organization ID if your CRM stores it. Up to three pairs are tried in order, so a second pair can catch what the first misses.
Add product usage and billing. Import a usage table with one row per account from PostHog or your Postgres database, connected with a read-only database user. Keep weekly active users now and 90 days ago, last login, and seats in use on it. Import your Stripe subscriptions too. On the accounts, link each one to its usage and its subscription, and bring back those numbers, the subscription status, and the current period end, which is the renewal date.
The usual route to the same place is a warehouse: land tickets and events there, model them in dbt, then run an LLM pipeline over the result. If a data team already owns that stack, it works. Most small companies probably have no such team, and the customer-success suites that do this for you are priced for a CS department rather than a founder watching thirty renewals.
Every connection in this guide fits the Team plan, which starts at $20 a month with no cap on integrations. On Free, the 3-integration limit covers your CRM, Zendesk, and one usage source.
Weigh ticket sentiment and unmet asks in the churn risk score
Per-ticket sentiment is what most help desks already show, and on its own it says little about churn. Every ticket is a little negative. Somebody wrote in because something broke.
What carries the warning is how many of an account's tickets ran out of patience. A customer who opened with a two-paragraph repro in May and now writes "still broken" is a different account from one that has been curt since day one.
Read each ticket for sentiment and unmet asks. Add a Use AI column on the tickets and have it write three fields for every ticket:
tone: calm, frustrated, or angry.feature_asked: the feature the ticket asks for, if any, such as SSO via Okta.expansion_ask: whether it asks for more seats, higher API limits, or a second workspace.
Then add a second link on the tickets with a run condition: tone is frustrated or angry. Tickets the condition skips stay unlinked. On the accounts, add Count rows pointing here once for each link, and every account now carries two numbers: tickets filed, and tickets frustrated. The ratio between them is the signal, far more than either count.
Anyway, back to the SSO request from the opening. Feature requests hide in tickets with subjects like "question about login" or "security review". Nobody tags them consistently, which is why product teams undercount them. With feature_asked on every ticket, two things come out of it:
The account owner gets a reason to call. Clicking the frustrated count on an account lists those tickets, each one a link that opens the ticket with its feature ask. A renewal conversation that opens with the thing they asked for three times goes better than a generic check-in.
Product gets a weighted list. Bring each account's contract value back onto its tickets through the link, then filter the tickets by
feature_asked. That list is probably more honest than any feature-voting board, because it only counts people who cared enough to write in.
A third link, conditioned on expansion_ask, counts the opposite case the same way. That count feeds the expansion score that sits next to the churn one.
Build the churn risk score with AI across every account
Now describe churn risk the way you would brief a new account manager. Add a Use AI column on the accounts. Plain language is the whole configuration:
AI reads every account's ticket counts, usage, and billing, one account at a time, against the same brief. That matters more than it sounds. A score of 70 means little on its own, but sorted beside 150 other renewals scored the same way it tells the owner which three calls to make this week.
Each account pays for the reading it used, up to a maximum the column shows before the run starts. The links and counts cost nothing.
The output looks like this:
Account | Churn risk | Reason |
|---|---|---|
Cleardue | 82 | 5 of 6 tickets since August frustrated; weekly actives down 40% in 90 days |
Juniper Studio | 71 | 3 of 4 tickets frustrated; champion last logged in six weeks ago |
Glass Garden | 35 | 9 tickets, 1 frustrated; seats up from 12 to 18 |
Glass Garden is the account a rules-based health score gets wrong. Points per ticket would flag it. Reading the tickets shows a team that is growing and asking good questions. The same approach scores prospects in ICP scoring with AI, and the reasoning carries over.
Write the churn risk score back to your CRM
A score in a tab nobody opens protects no renewals. Create these properties on the CRM company record once, with the right field type, then map the output to them. The export fills fields that already exist.
CRM property | Type | Example value |
|---|---|---|
| Number | 71 |
| Text | 3 of 4 tickets frustrated; champion last logged in six weeks ago |
| Number | 4 |
| Number | 3 |
| Number | 15 |
Those values are one account, Juniper Studio from the table above. The export matches each account to its CRM company, so a re-run updates the same record instead of creating a second one. The write-back mechanics are covered on the page about enriching CRM records from your own tools.
Then keep it current. Schedule the tickets' refresh weekly from their Source button, and turn on Run automatically on new and changed rows in their Table settings, so every new ticket gets read. On the score and the export, set Refresh to On a schedule and re-run every 168 hours. Sentiment moves faster than usage does, and a churn risk score that ran once in January describes an account that no longer exists.
Edge cases that skew a churn risk score
Tickets with no account match. A requester with no organization in Zendesk matches nothing. The line under the link's pairs counts the misses before you save, so add a fallback pair or fix the organization in Zendesk, and the next refresh picks them up.
More than 20,000 tickets. A link and a count each read a table of up to 20,000 rows. Import with First and a number instead of All rows, and keep the tickets table under that.
Accounts younger than 90 days. Onboarding tickets look like frustration. Tell the account brief to discount the first 90 days, or leave new accounts out with the run condition.
Accounts on two products. Tickets about the product they are leaving can hide inside a healthy total. Zendesk tickets carry a brand, so have the ticket read note which product each one is about.
Scores that flatter. Once a quarter, check churned accounts against the score they held 60 days before cancelling. If most sat under 50, the brief is weighting the wrong signals. Rewrite it and re-score every account in one pass.
Start with next quarter's renewals, Zendesk, and one usage metric. The Free plan includes 200 credits a month for 1 seat, up to 3 integrations, 3 tables, and 100 rows per table, which is enough to score a small renewal cohort and its recent tickets and see whether the reasons match what your account managers already suspect. Score your first renewal cohort in a free workspace this week.

