Which numbers actually tell you how big the cancellation problem is?
Four of them, and they are small enough to pull by hand.
Authorized hours against delivered hours, per client, inside one authorization period. That is utilization, and it is the only version worth reporting, because a plan is written in hours and delivered in hours.
Cancellation rate, counted in hours rather than in sessions. A cancelled full afternoon and a cancelled short session are not the same loss, and counting sessions hides that.
Callout rate, kept as its own number. A staff callout and a family cancellation both leave a hole, but they come from different causes and they take different paths back, so averaging them together tells you less than either one alone.
And the fourth, which is the point of the whole exercise: of the hours that were cancelled, how many came back delivered inside the same authorization period.
Why does the fourth number matter more than the other three?
Because the first three numbers measure the hole, and only the last one measures whether anything filled it.
You can watch a cancellation rate every week for a year and still not know whether a single lost hour ever came back. The rate rises, the rate falls, and the child’s total delivered hours sit outside that report entirely. I spent a decade as an RBT and I never once saw that number on a whiteboard.
It is also the hardest of the four to produce, because the schedule alone does not answer it. Schedule data records that an hour was cancelled. Whether it came back means following that specific hour forward: was a makeup offered, was it approved, did somebody actually deliver it, and did it land before the authorization period closed.
Infinite Suite OS counts an hour as recovered only at delivery, for exactly this reason. A family cancellation opens a makeup offer that the family approves, and a staff callout posts an open hour that qualified technicians may choose to claim. Nothing is ever auto-assigned, and nothing on the recovery path verifies a license, runs a clearance, or screens eligibility: credential checking lives in a separate billing claims gate. Neither one moves the recovered number until the session has been delivered. Authorization caps and daily unit limits are checked when the makeup is booked, so a makeup cannot be booked past what the authorization allows.
What happens when a rebooking gets counted as a recovery?
Every number underneath it inflates, quietly, and in the direction you were hoping for.
A rebooking is the easiest moment to count. The call went well, the family said yes, the block is on the calendar, and it feels finished. It is not finished. That session can cancel again, and the second cancellation rarely gets subtracted from the credit the first one created.
Then the inflation spreads. The recovery rate looks strong, so the utilization forecast built on it looks strong, so the staffing plan built on that forecast is wrong in the same direction. When delivered hours come in under plan at the end of the authorization period, nothing in the reporting explains why, because the reporting already called those hours saved.
The fix is unglamorous. Count at delivery. Keep the loss on the books until a real session happens, and let the number be uglier than you want it to be for a while. A rescheduled session is a promise. A delivered session is a recovery.
Should you measure against an industry average or your own history?
Your own history, because it is the only source that knows your clinic.
Cancellation behavior moves with things an average cannot see: in-home versus center, your payer mix, school calendars, the illness season in your region, how far families drive, whether one team is short staffed this quarter. Two clinics in the same state can run very different numbers for reasons that are entirely ordinary.
An average also cannot tell you where your losses cluster, and clustering is what makes a loss actionable. Look at whether yours concentrate in particular families, particular weekday slots, or a particular team.
Pull one full authorization period at minimum, and more if you have it, so seasonal weeks are in the sample. Tag every cancelled hour with who cancelled it, family or staff, and mark the ones you genuinely cannot classify as unknown instead of guessing. An honest unknown bucket is worth more than a clean-looking split you made up.
This site publishes the Lost Hours Calculator, where a clinic runs its own numbers rather than borrowed ones, and, separately, the recovery math simulation, labeled a simulation with its assumptions written out. Neither one is a benchmark, and neither should be quoted as your result. There is also an ABA lost hours calculator overview page for the same tool.
What do these numbers still not tell you?
They do not manufacture technicians.
If your callout rate is high because you are short staffed and nobody has a free hour, the measurement is accurate and the hole stays exactly where it was. That is a hiring and retention problem wearing a scheduling problem’s clothes, and no number fixes it.
They also do not decide anyone’s hours. Utilization can show that delivered hours are running under authorized hours, and what that means for a specific child’s plan belongs to the BCBA. Measurement never sets a treatment plan.
And they do not, by themselves, recover anything. Knowing precisely how many cancelled hours died is the beginning of the work, not the end of it. What measuring buys you is the ability to tell whether anything you try next actually moved delivered hours. That is a smaller thing than a fix, and it is the part you can check.