How churn risk is calculated

What actually drives a person or company's churn risk score in Synthight, and why it's Low, Medium, or High.

Last updated: 3 min read
how to calculate customer churnchurn risk score algorithmpredict customer churn from support ticketscustomer health score factors

Churn risk isn't a single metric like sentiment or resolution rate — it's a composite score built from multiple signals pulled from a person's (or company's) full incident history, recalculated automatically as new tickets come in.

What feeds the score

Broadly, the model weighs a few kinds of evidence:

  • Tone — how positive or negative their conversations are, and whether that's trending up or down over time.
  • Support outcomes — how reliably their tickets get resolved, including whether their most recent ones are still open.
  • Engagement patterns — how much they're contacting support, whether that pace is accelerating, and how recently they last reached out.
  • Severity and spread — how urgent their tickets tend to be, and whether their issues cluster around one topic or span many.

No single signal decides the outcome — a person with high volume but consistently positive sentiment and fast resolutions won't be flagged just for being vocal. It's the combination that matters, which is why two people with the same ticket count can land in completely different risk bands.

Reading the score

Under the hood, every profile gets a precise underlying score that Synthight then simplifies into the Low, Medium, or High label you actually see in the app — so you can act on a clear read without needing to think in percentages.

Low-data confidence

When someone doesn't have much of a ticket history yet, there simply isn't enough evidence to be confident — Synthight pulls the score toward a neutral baseline in that case rather than letting one or two bad tickets swing it straight to "High." The more incidents on record for a person, the more their score reflects their actual pattern rather than a cautious default.

Where you'll see it

Churn risk appears as a filter and column on People and Companies, as a stat card on both pages, and as a headline KPI in people and companies reports. For a company, it's a weighted average of the churn risk of the people at that company, rather than a separately computed formula.

Next steps

Frequently asked questions

What factors actually go into the churn risk score?
Four kinds of evidence: tone (positive/negative sentiment and its trend), support outcomes (how reliably tickets get resolved, including whether recent ones are still open), engagement patterns (contact volume, pace, and recency), and severity/spread (how urgent tickets are and whether issues cluster around one topic or span many).
Why do two people with the same number of tickets get different churn risk levels?
Because no single signal decides the outcome — it's the combination that matters. Someone with high ticket volume but consistently positive sentiment and fast resolutions won't be flagged just for being vocal, while someone with fewer tickets but negative tone and slow, unresolved outcomes can score higher.
Why isn't someone with only one or two bad tickets marked High risk?
Synthight pulls the score toward a neutral baseline when there isn't much ticket history yet, rather than letting one or two bad tickets swing it straight to High. The more incidents on record, the more the score reflects their actual pattern.
Is churn risk calculated differently for companies than for people?
Yes — a company's churn risk isn't computed from its own separate formula. It's a weighted average of the churn risk scores of the people at that company.
Can I see the exact numeric churn risk score, or only Low/Medium/High?
In the app you see the simplified Low, Medium, or High label. Under the hood every profile has a precise underlying score, but Synthight surfaces the simplified label so you can act on a clear read without thinking in percentages.

Was this article helpful?