Finding 1
Almost one-third of measured cancellations arrived inside 24 hours
30.6% of cancellation records with usable notice arrived within 24 hours of the scheduled start. 8.5% arrived within two hours.
That concentration supports having a clear late-cancellation workflow, but it does not determine what a fair policy should charge or credit.
Finding 2
Cancellation records are not the same as lost lessons
11% of eligible records were explicitly marked as rescheduled. Other records may have been refunded, credited, restored or replaced through workflows not measured here.
For that reason, this report does not publish a cancellation rate, no-show rate or revenue-loss estimate. Those require a deduplicated booking-level outcome model.
Cancellation record summary
| Measure | Result | Denominator |
|---|---|---|
| Median notice | 98.7 hours | 4,030 |
| Within 24 hours | 30.6% | 4,030 |
| Within two hours | 8.5% | 4,030 |
| Marked rescheduled | 11% | 4,293 |
| Parent-initiated | 34% | 4,293 |
| Provider-initiated | 66% | 4,293 |
Methodology
How the benchmark was produced
Swum ran a read-only aggregate extraction against its production operational records on July 31, 2026. The extraction never emitted document identifiers, names, emails, cancellation notes or other free text.
- Lesson cohort: past lessons only; schedule blocks, future records, failed or pending statuses, unsettled manual cash records and self-booking artifacts were excluded.
- Booking cohort: successful, non-manual parent orders with a valid creation time.
- Cancellation cohort: cancellation records with a valid cancellation time; schedule blocks were excluded.
- Account hygiene: 80 clearly internal, test, fake or generic coach accounts were excluded before analysis.
- Privacy: segmented findings require at least 10 records contributed by at least five provider accounts.
- Reproducibility: the published CSV contains the aggregate values used on these pages. It intentionally excludes row-level data.
Limitations
What this data cannot establish
- Cancellation records are operational events and may not map one-to-one to unique original bookings.
- Only non-negative notice values up to 120 days were included in timing percentiles.
- No free-text cancellation reasons were read, extracted or published.