A grading engine an awarding body can stake its reputation on

The situation

EAL was launching a qualification with a grading model unlike anything it had run before. Final grades had to be calculated from unit scores for every candidate, and one wrong result would land on a real person and on EAL’s reputation. The engine also had to work alongside SAP, which holds all qualification and candidate data, and SAP wasn’t going to stand still.

What we questioned first

• What happens when the systems around it change?Rather than wiring the engine into SAP, we proposed keeping it independent behind its own API with a one-way boundary. SAP can push and pull data, but changes to SAP, or anything else, can’t break the grading.
• What if the data coming in is wrong?A hardened validation layer rejects anything that breaks the business rules before it ever reaches a calculation.
• How will anyone know it’s right?Test-driven development from day one: a full suite of tests proving the engine meets every grading rule, so the team could see it worked.
• Who has to use it?Non-technical staff configure each qualification, so the admin interface went through wireframes and prototypes until it worked without extensive training.

What we built

A secure grading engine with a React admin interface, a technology-agnostic API for SAP, reporting built around the data points administrators use, anomaly alerts, and audit tracking on every interaction with sensitive data.

What changed

Delivered on time and within budget. The engine now grades thousands of candidates, and because it sits behind its own boundary, EAL can change the systems around it without rebuilding the part that matters most.