The Drift List: Catch Students Who Quit Before They Cancel

A parent in Lima almost never calls to quit. They just stop showing up.

Parents don't cancel. They fade.

In August, the Ohio Capital Journal reported that Ohio school districts charge as much as $500 per sport, per season, with no state limit on what they can charge — and that some districts now split the fee into installments so families can absorb it. The Aspen Institute's Project Play survey put average family spending on a child's primary sport at about $1,016 in 2024, a 46% jump in five years.

Your tuition lives in that same household budget. When it tightens, almost nobody sends a cancellation email. They miss a Tuesday. Then the next Tuesday. Then they're "taking a break." By the time the card declines, the student has been gone six weeks and the conversation is already over.

The signal that would have caught it was sitting in your building the whole time — in the check-in log, the test roster, the class group chat. It just wasn't connected to anything.

Build the drift list first

Not a dashboard. One list on your phone every morning: students whose pattern changed, ranked by how far they've drifted.

Off-the-shelf studio software tracks two things well, payments and door swipes. It doesn't know that your Tuesday/Thursday 6pm kids' class is a different commitment than Saturday open mat. It doesn't know that a white belt missing two weeks in month three is a much worse sign than a brown belt missing two weeks in year five, or that a dancer going quiet eight weeks before recital usually means costume money got tight.

A custom CRM knows, because you told it. The data model gets built around what a school actually has:

• Student and guardian as separate records — the student trains, the guardian pays and reads your texts
• Enrollment, with program, rank or level, expected weekly frequency, and start date
• Attendance tied to a specific class slot, not just a building entry
• Milestones: next test, recital, tournament, renewal date
• Trial inquiries, with source, class attended, and a clock

With that underneath, the AI part is small and deliberately boring. It reads each student's own history, flags the deviation from their normal, and drafts one specific message: "Maya has missed three straight Thursdays but is still coming Tuesdays, and her stripe test is in two weeks." That's a text you send from the parking lot in fifteen seconds. It is not a blast to 200 people, and parents can tell the difference immediately.

The same model handles the front of the pipeline. You teach from four to eight, which is exactly when parents sit down to research classes. A trial inquiry becomes a record with a stage and a deadline, and follow-up gets timed around your teaching schedule instead of your inbox — so "came Saturday, hasn't decided" is in front of you at nine that night, not buried by Wednesday.

What starting looks like

One design constraint belongs in the spec from day one: you are storing minors' names, birthdates, photos, and sometimes medical notes. The FTC's amended children's privacy rule reached its full compliance deadline on April 22, 2026, and it added limits on how long covered operators may keep children's data at all. Whether or not your school is a covered operator, decide up front what your system keeps and what it deletes on a schedule. Fifteen minutes now, or an expensive cleanup later.

For context on timing: Census Bureau data shows roughly 18% of U.S. firms were using AI at the end of 2025, and Federal Reserve researchers found uptake among the very smallest firms running stronger than firm size alone would predict. Small shops are already moving on this.

Our Custom CRM work produces three things. A CRM spec written in your language — belts, levels, sessions, families, not "opportunities" and "accounts." A data model showing every entity and how they connect. And a build plan that says what gets built first, what waits, and what it costs, before anyone writes code.

Start with the drift list. It pays for the rest of the build.