Reimagining Data Migration: Tackling a large-scale ERP consolidation with user-centred innovation

Over 60 organisations nationwide needed to export, upload, and map data across 5 ERP domains, from at least 1000 distinct source systems. This would enable the data to be imported into a reduced set of target systems, to be run by several new, centralised entities. Additional complexity stemmed from variations in staff technical ability, availability, and limited government specialists to assist manually. We partnered with Section6 (Software development + DevOps) to deliver this innovative approach to data migration and ERP transformation.

Data migration tools typically present mapping tasks in the order ‘source to target’. With this layout, users are encouraged to focus on their source data, to try and find a home for all of it. While this may be important in the long run, it may not actually be necessary for Day 1 go live. Encouraging users to do more than needed can jeopardise meeting deadlines.

For the new base systems to become operational on Day 1, organisations only needed to map a subset of their source data to the mandatory target attributes. Post-launch, additional source data could be mapped to optional target attributes with the flexibility of Business as Usual (BAU) time in the new entity.

With our Data Migration tool, we innovatively presented mapping tasks as 'target to source’. This, coupled with being able to filter to show mandatory target attributes only, guided the users to perform their mapping tasks in a prioritised way.

‍

There was no off-the-shelf mapping tool globally that could fit the programme’s needs, especially none with flexibility that allowed for the data standards to evolve along the way.

Existing tools were too rigid and would have required too many hours performing manual tasks and work arounds - the business case favoured building a custom tool.

We expedited the usual timeline of a custom build by working with real users early to understand what really mattered. Our early design prototypes and testing allowed us to fail fast and learn. Total time from early discovery through to design and MVP build (where the mapping feature was available) was just 5 months. All planned major features were live by the 1 year mark.