The Challenge

The organisation ran a regulatory submissions platform that had grown over several years to support eCTD publishing, document management, review and planning across several regions. Underneath, every kind of information had its own separate storage — one place for submissions, another for products, another for substances, another for the links between them. Each came with its own retrieval logic, written and maintained separately.

That worked while the platform was young. But as it expanded to support new kinds of information — grouped submissions, compliance domains, submission plans and AI-driven features — keeping a separate arrangement for each one created compounding problems.

The Approach

Rather than continuing to extend the old arrangement, the team built a single, flexible foundation. One shared store holds every kind of information, with configuration describing what each kind looks like. New kinds of information can now be introduced by changing configuration rather than by rebuilding storage.

The move followed a methodical path: each of the fifteen legacy routes was analysed individually, mapped onto the new foundation, and switched over while everything it fed carried on working exactly as before. A single, consistent way of reading and writing replaced all fifteen. The translation between what the software calls a field and where that field is actually held was brought into one place, rather than repeated in fifteen.

Critically, there was no big-bang changeover. Old and new ran side by side throughout, so each part of the platform could move independently while the system as a whole stayed live.

What Was Delivered

The work also resolved several long-standing data quality problems. Cases where two different fields had been quietly sharing the same storage were found and corrected. Inconsistencies between how a value was saved and how it was read back were standardised. And requests that had asked for “everything” were replaced with requests that name exactly what they need, so that adding new information could no longer produce unexpected behaviour elsewhere.

Key Outcomes

Why It Worked

• One route at a time: moving a single route at a time allowed continuous checking without risking the whole system. Each move was independently testable and could be undone
• One place for translation: bringing field translation into a single place eliminated an entire category of inconsistency
• Running old and new together: both operated simultaneously during the transition, so live operations continued without interruption
• Asking only for what is needed: replacing “give me everything” requests with named ones prevented surprises as the shared store grew
• Keeping things ready: holding each customer’s field translations ready meant the new, more flexible foundation did not come at the cost of speed

Looking Ahead

The new foundation now underpins all new development. Compliance domain management, AI classification results, search indexes and submission plans all store and retrieve information the same way. The organisation’s roadmap can be delivered without further rebuilding — new capabilities arrive as configuration rather than as engineering projects.

For the regulatory operations teams themselves, the change was largely invisible; their day-to-day work continued uninterrupted. The impact showed up in the pace of everything that came afterwards: capabilities that used to require months of foundational work now ship as ordinary features.