Measuring what teachers actually know

The brief

Professor David Johnson's team at Oxford had spent two decades researching what teachers know and how that knowledge grows. The research covered around 100,000 teachers across Africa and South Asia and had produced a question bank of more than 400 items. What it did not have was a way to put that research in front of a working teacher.

The problem is stated plainly on OxTrack's own site: in many low-income economies, good teachers do not know how good they are, and teachers with uneven skills rarely get pointed feedback on what to do about it.

Oxford Education Labs, a social impact spinout of the University of Oxford, engaged us in 2024 to build the delivery vehicle. The contract was fixed scope, with wireframes agreed up front.

Built for scale

The first version ran on introductory hosting. That suited a pilot and would not have survived a national rollout. When the client's programme in Nigeria moved from pilot to scale, we rebuilt the platform rather than patch it.

The result is an AWS deployment running a load balancer in front of five application instances, with autoscaling tied to system load. We load tested it at 250 requests per second, which supports around 2,500 teachers each making a request every ten seconds, and well over 16,000 active users at a more typical one request every five minutes. We had the platform penetration tested the same week and no critical issues were found. The system was designed for 20,000 users and to expand past that.

What that scale turned out to mean

Oxford Education Labs' own published account of the Nigeria pilot, written by a World Bank education economist, records it running inside the World Bank-supported Better Education Service Delivery for All – Additional Financing programme, across Abuja FCT and the states of Oyo, Adamawa and Katsina. It registered over 22,720 users, with nearly 5,000 teachers completing their first assessment.

The platform had been sized for 20,000. Deciding that number a year earlier, rather than discovering it under load, is the whole value of the rebuild. We also stayed on standby through the client's demonstration to programme stakeholders.

Built for the classroom, not the office

An assessment takes a teacher over an hour, and the teachers taking it are in classrooms in Nigeria and Kenya, not at a desk on fixed broadband. An app that needed a live connection to record an answer would lose people's work, and losing a teacher's work once is enough to lose the teacher.

So OxTrack does not depend on the network being there. Each answer and each timing record is queued on the device as the teacher works. A background processor watches for connectivity, submits the queue when it returns, retries on failure, and keeps anything still unsent instead of dropping it. On the server, submissions land in a staging area and are processed on a schedule.

The teacher sees none of this. They answer questions, tap Save Progress, and their work is there when they come back.

Saying no to the easy answer

Late in 2025 a research team evaluating the programme asked for full access to the servers holding teacher data.

We declined that request. Broad infrastructure access is the wrong instrument for a research question, and the data involved personal information about teachers in three countries, governed by UK rules on international transfers.

Instead we designed what the researchers actually needed. We asked which fields mattered, whether direct identifiers were required at all, which legal entity would receive the data and in which country it would be processed. The answer was an automated monthly export, scoped to the regions under study, with direct identifiers removed, delivered on a controlled schedule.

The client's response: "You are absolutely right on data security and risk."

That is the part of this project we are most pleased with. Least privilege is easy to agree with in principle and awkward in the moment somebody influential asks for a shortcut.

Where it is now

OxTrack is live on the App Store and Google Play, published by Oxford Education Labs. It is deployed in Nigeria and Kenya, including low-cost private schools in Nairobi. We have supported and extended it continuously since 2024.