CASE STUDY · ONLINE EDUCATION
Silence turned into a retention signal before the refund window closed
A cohort-based education provider knew learners went quiet before they churned, but nobody saw it until the refund request arrived. We turned attendance and channel silence into a single scored signal routed to the program lead.
How to read these numbers
This client asked not to be named. The engagement shape is real; the figures below are directional estimates pending client sign-off, not audited results. (Directional estimate · client not named)
2 sessions
Earlier detection than manual review
Directional; pending client sign-off
< 24 hrs
From flag to an owned outreach task
Previously ad hoc or never
1 queue
Signals landing where leads already work
No new dashboard introduced
We always said we could feel when a cohort was slipping. Now it is a number that shows up before the refund email does.
Paraphrased from the engagement and shown without a named person while we wait on client sign-off. We will not attach a name or a title to a quote the client has not approved.
Context
- A provider running paid multi-week cohorts across Slack, a video platform, and an LMS.
- Program leads managed several cohorts at once and relied on memory to notice a struggling learner.
The problem
- Attendance, channel activity, and assignment submissions lived in three systems that never spoke to each other.
- By the time disengagement was obvious, the learner had already decided to leave.
What we did
01 · Define
Agreed one metric with the client: learners re-engaged after a silence flag, per cohort.
02 · Design
A disengagement score combining missed sessions, channel silence, and unsubmitted work, with an explicit threshold rather than a black box.
03 · Build
Connectors into the LMS and community, an orchestration layer that scores nightly, and an outreach task written into the program lead's existing queue.
04 · Measure
Tracked flag-to-outreach time and whether flagged learners returned to the next session.
Tooling shown is what was used in this engagement, not a required stack. We build to the platforms and systems you already run.
What we would do differently
We would tune the threshold on historical cohorts before going live. The first two weeks over-flagged, and the team briefly learned to ignore the queue.
NEXT STEP
Recognise the pattern?
Tell us where your signals live and we will tell you whether this shape fits.