My CSR says the call was not bookable but the system flagged it. Who is right?
Often neither, because bookable is a policy definition rather than an observable fact. Before you can adjudicate a single call you need a written rule covering service area, services offered, decision-maker status and warranty work. Then sample disputed calls, have a human reviewer and the system score them independently, and use the disagreements to fix the definition. Persistent disagreement is a definition problem, not a model problem.
Bookable is a decision your company makes, not a property of the call
Two well-run companies can listen to the same recording and reach opposite conclusions, because one services that zip code and the other does not, or one takes landlord-authorized rental work and the other refuses it. Neither is misclassifying. They have different rules.
So the first artifact in any dispute is the rule itself, written in a sentence a new hire could apply without asking. If nobody can produce that sentence, the argument is unresolvable and the metric is not yet real.
The exclusions worth naming explicitly
- Out of service area. Define it by the boundary you actually dispatch to, not the radius set in the ad account.
- Service not offered. Separate a service you never perform from one you decline seasonally or under load.
- Not the decision maker. Tenants, adult children and property managers without spending authority.
- Existing job follow-up. A customer calling about work already scheduled is operational traffic, not new demand.
- Warranty or another company's work. Frequently mislabeled as lost opportunity when it never was one.
A calibration loop that actually converges
Take a sample of calls the system marked bookable and the team disputes. Have a supervisor score them blind against the written rule. Compare. Where the human and the system disagree, decide which one applied the rule correctly, then either correct the rule or correct the configuration. Repeat monthly at first, then quarterly once the gap narrows.
The output of this loop is not a perfect model. It is an agreed definition plus a known error rate, which is what makes the number usable in a performance conversation. We run this loop as part of rubric-based evaluation, where the standards belong to the company and the system applies them consistently.
Keep the metric out of the wrong conversation
Bookable-call counts are a coaching and staffing instrument. They show where handling is failing and where demand is being wasted. They are not an automated verdict on a person, and evaluation output should support a manager's judgment rather than substitute for it. That framing is also what keeps the data credible with the team, which is a prerequisite for anyone using it at all.
When the definition is stable, the same rule can be applied across locations and across time periods, and only then does a trend line mean anything. That link between clean definitions and trustworthy reporting is the basis of revenue intelligence.
Topics: bookable calls · definitions · calibration · quality assurance
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.