What an engagement score can include
A scoring method may use replies, completed conversations, portal activity, or other permitted interaction data. The lookback window, weights, update schedule, and treatment of missing data should be documented.
There is no universal engagement-score formula or scale. Compare scores only when their definitions are compatible.
Interactions need context
A reply might contain a dispute, a hardship request, or an instruction to stop contact. A link click may come from an automated security check. Delivery or open signals do not necessarily establish that the intended person read a message.
Hypothetical example: A customer replies twice to challenge an incorrect balance. That activity indicates engagement, but the next step is investigation, not a stronger payment demand.
What the score cannot establish
An engagement score is not proof of identity, consent, willingness or ability to pay, or legal collectibility. A recorded commitment is also different from a confirmed payment.
Low activity may reflect stale contact details, accessibility barriers, or channel preferences rather than deliberate avoidance.
Using it responsibly
Apply account restrictions and required holds before prioritization. Validate whether the score supports the intended decision, check for errors, and monitor outcomes beyond response volume.
Do not present engagement as a calibrated recovery probability without supporting validation.