Too late
Termly reconciliation means a four to twelve week gap between disengagement and anyone noticing.
In production with paying customers
ClassAura
Engagement intelligence for UK higher education, built natively on Microsoft Teams. It tells you who is drifting, why the system thinks so, and what was done about it last time.
0UK contracts signed
0isolated core tables
0engineering sessions delivered
decision tracelive
signal inattribution backevidence
The problem
Retention decisions get made once a term, from spreadsheets, about students who left weeks ago.
By the time a learner appears on a list, the intervention that would have worked has usually expired. And when someone asks why a particular student was flagged, nobody can reconstruct the answer.
Too manual
Three systems, two spreadsheets and a member of staff copying between them, with every transcription a chance to lose someone.
Unaccountable
A flag with no visible basis cannot be checked by a tutor, questioned by a student, or defended to an inspector.
What is in it
Six modules, one data model.
You can start with cohort intelligence alone and add the rest as the institution is ready.
Cohort intelligence
A ranked view of an entire cohort with the contributing signals shown beside each learner, refreshed continuously rather than at census points.
Signal ingestion
Attendance, participation, submissions and tutorial contact read from the systems you already run, through Microsoft Graph, with no duplicate data entry.
Intervention tracking
What was done about a flag, by whom, and whether it worked. Closes the loop that most analytics products leave open.
Student transparency portal
The learner sees the same signals their tutor sees, and has a defined route to question them.
Staff Hub
Timetables, notices and cover on a phone, distributed privately to your institution.
Reporting and export
Governor-ready and inspection-ready reporting, exportable rather than trapped behind a dashboard login.
Architecture
What the technical evaluation will ask about.
ClassAura is a multi-tenant SaaS platform. Every row of institutional data is scoped to an organisation identifier at the database level, so one customer's query cannot reach another customer's data even if application logic fails.
| Front end | React 18, TypeScript, Vite |
| API | FastAPI, SQLAlchemy, Pydantic v2 |
| Database | PostgreSQL, row-level tenant scoping |
| Identity | Microsoft Entra ID single sign-on |
| Signals | Microsoft Graph — Teams and Zoom |
| Hosting | Azure, UK region by default |
| Isolation | organisation_id across 13 core tables |
| Audit | Write-time logging on every access and change |
Getting started
Four steps from first call to live cohort data.
Most institutions are looking at real data within six weeks of the first conversation.
Technical and governance walkthrough
We show you exactly what data is read, where it sits, who can see it and how long it is kept, with your data protection officer in the room.
Tenant provisioning
Your isolated environment is created and Entra ID single sign-on is connected to your existing identity provider.
Signal connection
Graph permissions are granted at the scope you approve, and historic signals backfill so the first cohort view is not empty.
Pilot cohort and training
One programme goes live first with its tutors trained, so the model and the workflow are both tested before the institution commits.
See it against your own cohort.
The walkthrough uses your structure and your terminology, not a demo dataset. Your data protection officer is welcome on the call.