Finding 1
Most observed family relationships extended beyond one lesson
Across 1,784 qualifying business-family relationships, 1,581 contained more than one scheduled lesson. The median relationship contained eight lessons, the 75th percentile contained 14 and the 90th percentile contained 26.
That is a scheduled-activity measure rather than a verified attendance measure. It nevertheless shows that the operational unit in swim instruction is often a sequence of lessons—not a single isolated appointment.
Finding 2
Plans showed the longest scheduled continuity
Among relationships beginning with a plan, 85.9% of age-eligible families had scheduled activity during days 31–90, 79.5% during days 91–180 and 53.8% during days 181–365.
Relationships beginning with an individual lesson measured 61.8%, 32.8% and 7.0% in the same windows. This is a large observational difference, but plans inherently create future lessons and attract families willing to make a longer commitment.
Finding 3
Packages occupied the middle ground
Package relationships produced 76.4% scheduled activity during days 31–90, 41.6% during days 91–180 and 30.8% during days 181–365. Their continuity was stronger than individual lessons and shorter than plans in the longer observation windows.
That shape is consistent with the product structure: a package reduces repeated purchase decisions while still ending after a finite bank of lessons.
Finding 4
Product choice and customer intent cannot be separated here
Families choosing plans may begin with higher intent, more urgent learning goals or greater schedule stability. Providers may also reserve plans for specific customers or seasons.
A causal test would randomize which eligible families receive a plan or package offer, then compare completed lessons, cancellations and repurchases—not merely scheduled records.
Scheduled activity by first lesson product
| First product | Relationships | Days 31–90 | Days 91–180 | Days 181–365 |
|---|---|---|---|---|
| Individual lesson | 1,245 | 61.8% (373/604) | 32.8% (118/360) | 7.0% (16/230) |
| Subscription plan | 355 | 85.9% (195/227) | 79.5% (116/146) | 53.8% (35/65) |
| Package | 119 | 76.4% (68/89) | 41.6% (32/77) | 30.8% (20/65) |
| Semi-private | 59 | 83.3% (20/24) | 80.0% (8/10) | Insufficient cohort |
Every percentage uses only relationships old enough to have completed the full observation window. Cohort sizes therefore change by window.
Relationship depth
| Measure | Result |
|---|---|
| Relationships with multiple scheduled lessons | 88.6% |
| Median lessons per relationship | 8 |
| 75th percentile | 14 |
| 90th percentile | 26 |
| Median first-to-last span among repeat relationships | 49 days |
Methodology
How the benchmark was produced
Swum grouped qualifying, non-canceled historical lesson documents by business and family, ordered them by scheduled start and tested for later scheduled activity in fixed windows.
- The observation window ended before August 1, 2026.
- Booked and successful lesson records were included; blocks, canceled, failed, pending and checkout-pending records were excluded.
- A relationship is a family paired with one provider business. The same family may appear once for another provider.
- Every window includes only relationships whose first lesson was old enough to observe the entire period.
- Scheduled activity is not proof of attendance, payment collection or progress.
Operator response
What to do with this finding
- Measure continuation separately for drop-ins, packages and recurring plans.
- Offer a finite package when a family is not ready for an open-ended plan.
- Track completed attendance so scheduled continuity can become a true participation measure.
- Test plan and package offers with eligible comparison groups before claiming incremental lift.
Limitations
What this data cannot establish
- Plans can schedule future lessons automatically, which mechanically increases scheduled continuity.
- The analysis is observational and does not establish that a product caused a family to remain.
- Semi-private long-window cohorts were too small for a stable one-year comparison.
- Group-program attendance is stored in separate instance records and is not part of this cohort.