Attendance leads in schools are busy people. In most trusts, they start each day responding to what happened yesterday or that morning: chasing registers, making calls to parents. It is reactive by necessity. The data arrives after the fact, and the conversations and interventions follow.
But tools available to the sector are changing. Advanced analytics - once the preserve of industries with deep technical resource - are increasingly within reach of schools and trusts with solid data foundations. Attendance, behaviour, attainment, SEND: the sector's most persistent challenges are, at their core, information problems. The question is whether, as a sector, we have the right infrastructure to gather data intelligence into timely, useful insight.
At Inspiration Trust, Director of Data and Performance Mark VandenBergh and colleagues decided to find out.
Getting the foundations right
Before any prediction data model can run, data from multiple systems and sources needs to be ingested, standardised, and made ready for analysis. This is the unglamorous work that determines whether valuable insight is possible at all.
Mark and his colleagues built their attendance prediction work on the Open Education AI data architecture, which provides the education sector with a free to access, open-source data architecture designed to get data from multiple sources flowing and transformed into a consistent, usable form.
Open Education AI is also a community working together, so they didn't have to begin from scratch. Another leading trust had developed an early concept for attendance prediction, and Mark extended this into a production-grade system across the trust's18 schools and 11,000 pupils. What began as a minimum viable product by the end of 2024 was live in schools by September 2025.
The data model generates next-day absence predictions for every student, every morning, drawing on attendance history, seasonal patterns, personal milestones, and environmental factors. In the first month, it processed nearly a quarter of a million predictions at 93.3% accuracy, generating over four thousand genuine intervention opportunities.
But the accuracy headline figure was not the one that mattered most.
Rethinking what a wrong prediction means
What does it mean when a student is predicted absent and then walks through the door? The instinct is to call it an error. The team at Inspiration came to see it differently. A student predicted absent who then attends is often a student where a successful intervention has taken place. The prediction was the prompt. The relationship did the rest.
The aim is not to guess the future perfectly, but to change it. Every prediction that turns out to be wrong is a story about something going right.
Pastoral teams stopped scoring the model against registers and started asking what the prediction had made possible. That shift from accuracy to agency is the more important outcome.
The predictive model produces a list. What schools do with that list is where the real work begins. The Inspiration Trust is now exploring how to bring intervention data back into the same data architecture - tracking what was done in response to each prediction, not just whether the student attended. Over time, that builds a genuine evidence base about what works, where, and why. That kind of learning cannot happen at school level alone. It requires the scale of a trust, and ideally a community of trusts working from shared foundations.
The sector has long known what its hardest problems are. What is changing is our ability to bring evidence to bear on them systematically, ethically, and at scale.
We are still learning. But we are learning faster because we are working together as a sector.
- Amy McJennett is Director of Open Education AI.
- Amy and Inspiration Trust's Mark VandenBergh will be sharing more of their experiences at the CST Data and Digital Transformation Conference on 18 June 2026.
We welcome perspectives from a diverse range of guest contributors. The opinions expressed in blogs are the views of the author(s), and should not be read as CST guidance or CST’s position.
