Atrium
ML Model Predicts Course Demand 3 Months Ahead — Saving 2–3 Hours Daily
The situation
The challenge
Atrium monitors course registrations daily to assess whether a course has enough participants or should be postponed. When the need for additional marketing arose, it was typically discovered too close to the course date to react effectively. Atrium wanted an intelligent solution to identify this need much earlier in the process.
What we did
We consolidated data from Google Analytics, CRM and the booking portal and analysed the relationship between course-specific web activity and registration numbers. A Machine Learning model was built to predict course registrations 3 months ahead of time, visualised in a custom dashboard that gives the marketing team a clear daily action list.
Results
The model delivered a custom dashboard with predicted registration levels per course, reduced manual monitoring by 2–3 hours per day, and lifted overall marketing performance by 45% — by focusing effort precisely on courses with the greatest need.
This project was delivered by the team behind Dear Future, formerly operating as IIH Nordic.
The approach
Multi-source data consolidation · ML model: registration prediction · Custom prediction dashboard
Multi-source data consolidation
Gathered data from Google Analytics, CRM and the course booking portal into a unified analysis environment.
ML model: registration prediction
Built a Machine Learning model analysing the relationship between course-specific web activity and historical registration numbers to predict demand 3 months ahead.
Custom prediction dashboard
Delivered a visual dashboard showing predicted registration levels per course — giving the marketing team a clear, daily action list without any manual data work.
The results
Numbers that held up after the project closed.
By focusing marketing spend on courses with predicted low registration — rather than guessing — Atrium lifted overall marketing performance by 45%.
Staff previously spent 2–3 hours every working day manually checking registration levels across all courses. The ML dashboard eliminated this entirely.
Having 3 months notice to act on predicted shortfalls means marketing campaigns can be planned, tested and optimised — not rushed out the day before a course is cancelled.
Is your team doing work like this?
Tell us the problem. We'll tell you if we can help.
We're direct. If we've solved something like this before, we'll say so. If we haven't, we'll say that too. Either way — the first conversation costs nothing.