How predictions work
The plain rules behind Customer 360's health score and predictions.
Customer 360 uses clear rules, not a trained model. Every number on screen comes with the reasons and the data it used. Here's what goes into each one.
Likelihood and confidence
- Likelihood is how likely something is: High is 60% or more, Medium 35% or more, otherwise Low.
- Confidence is how much ServFlo knows. A furnace with a recorded install date gives a more confident prediction than one with no age at all.
Health score
Every customer starts at 70. Points are added for things like being on their usual rhythm, having a visit booked, an active membership, paying on time or by autopay, a good review and replying to messages. Points are taken off for being overdue for service, a cancelled membership, late or overdue payments, a bad review, declining quotes and not replying.
Service due
The first of: a booked visit, the next membership visit, their own rhythm between visits, the service type's usual repeat interval, a seasonal pattern, or one year after their last service.
Equipment replacement
Adds up the unit's age against its expected lifespan, repairs in the last 24 months and what they cost, R-22 refrigerant, a poor or failed condition, being out of service and notes mentioning trouble (such as leaks, rust or a compressor). Being under warranty lowers it.
Lapse risk
How long it's been since their last service compared with their usual gap. A booked visit or an active membership lowers it. A membership ending without auto-renew, a cancelled membership, declined quotes and a poor review raise it.
Late payment risk
The share of their past invoices paid more than 2 days late, plus any overdue invoice now. Autopay lowers it.
Quote acceptance
How often this customer approves quotes, balanced with your business's overall rate, then lowered the longer a quote stays open and raised for members.