How do I estimate what my unbooked calls are actually costing?
Multiply four terms: unbooked bookable calls, the share of those you could realistically recover, the rate at which recovered bookings turn into completed jobs, and average job value for that call type. The third and fourth terms are where estimates go wrong. Use the average ticket for the specific call category rather than the company-wide figure, and present a range rather than a single number.
The formula, and where each term comes from
Lost revenue equals unbooked bookable calls, times a realistic recovery share, times the completion rate for bookings, times average job value for that category. Each term has a different source and a different level of confidence.
- Unbooked bookable calls. From call analysis, after junk filtering, deduplication and the delayed-booking window have all been applied. Skip any of those and this term is inflated.
- Recovery share. The portion you could actually convert with better handling or coverage. This is a judgment, and it is never all of them.
- Completion rate. Booked jobs that get performed. Cancellations and no-shows are real and they belong in the model.
- Average job value. From the field service system, filtered to the same category of work the unbooked calls represented.
The two errors that produce absurd numbers
The first is double counting: including calls that already booked later through a different path. Applying the delayed-booking window before you count is what prevents this, and skipping it can inflate the estimate substantially.
The second is using company-wide average ticket. Unbooked calls skew toward small jobs, price-sensitive callers and edge-of-territory work, while the company average is lifted by replacements and installs that rarely come from a lost inbound call. Using the wrong average is how a lost-revenue estimate ends up larger than the entire market.
Present a range and name the assumptions
Run the calculation at a conservative, a moderate and an optimistic recovery share, and publish all three with the assumptions written next to them. An owner who can see which assumption drives the number will engage with the model instead of dismissing it.
This is also the format that survives scrutiny. A single confident figure invites an argument about the figure; a range with stated inputs invites an argument about the inputs, which is a far more productive conversation and usually where the real decision sits.
Use it to size the fix, not to assign blame
The purpose of the estimate is to decide whether a change is worth making. If the moderate case comfortably exceeds the cost of extending coverage hours, adding a person, or fixing an IVR path, that is a decision. If it does not, that is also a decision, and a cheaper one to reach than by trial.
Because the estimate depends on clean definitions of bookable and on a real join to job outcomes, it tends to be the last metric a company can compute rather than the first. When it is available, it is one of the more useful outputs of connecting call data to revenue reporting across the systems a service business already runs.
Topics: lost revenue · estimation · booking rate · modeling
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.