Our workFind the learner
Learning Curve Group · Data & AI
Find the learner
before you lose them.
A machine learning system that turns early warning signs into timely, human support.
The opportunity
Spotting the risk is only useful if someone can act.
Learning Curve Group supports around 40,000 active learners. Some were enrolling but never beginning; others were disengaging before tutors knew they needed help.
The question wasn’t simply whether data could predict dropout. It was whether that prediction could reach the right person, early enough to change what happened next.
The challenge
Move from reacting late to supporting early.
Tutors could intervene once a learner had disengaged, but by then the relationship was harder to recover. Chasing every learner wasn’t the answer either; it would pull tutors away from teaching.
Waterstons needed to make the data useful without making the experience feel automated. That meant predicting risk, choosing the right moment and reaching out through a channel learners already used.
What mattered
Focus human attention where it could make the greatest difference.
The model doesn’t replace the tutor. It helps the tutor arrive sooner.
The idea that held the whole solution together
The story in motion
From signal
to support.
See how the idea moves through the learner journey. This short visual story follows the system from a nightly data check to a timely conversation.
01 / 03
What we did
One connected journey.
Four chances to help.
- 14days
Spot learners who enrol but do not get started.
- 25%
Read the early patterns in engagement and progress.
- 50%
Reassess risk as learner behaviours change.
- 75%
Help learners through the final stretch.
Data made actionable.
Waterstons reshaped Learning Curve Group’s warehouse data, tested around 20 variables and built separate machine learning models for each point in the learner journey.
A nightly process scores every active learner. Those most at risk can receive personalised WhatsApp outreach, while controls prevent repeat messaging and A/B testing shows what is genuinely working.
≈4%
The early result
A small percentage.
A lot of lives.
The initial pilot reported a circa 4% decrease in learner non-completion. Behind that figure are more learners staying on course, tutors using their time where it matters, and a funding model made more resilient.
Earlier support
Risk is identified before disengagement becomes irreversible.
Focused time
Tutors can prioritise the interventions most likely to help.
Evidence built in
A/B testing and audit logs make progress visible and improvable.
Advice that holds up
Built with the people
who would carry it forward.
The final handover gave Learning Curve Group’s development team the working model, scoring service, test coverage and practical knowledge to own what came next.
What could your data make possible?