By Jane Steen, Head of Engagement at ETL.
If you’re accountable for the data in a large organisation, the problem usually isn’t hard to spot. The data sits across systems that were never built to work together, the wider business can’t use it properly, and integrating it is plainly the answer. However, any hesitation is rarely about whether to do it, it’s about the risk of the project itself.
The thing worth remembering is that these projects seldom fail on the data. They fail on how they’re run. So here are a few things we’ve learned that separate the integrations that work from the ones that don’t:
1) Audit the data before you plan the project. You can’t scope or cost an integration realistically until you know how messy the source data actually is. The projects that overrun are usually the ones that committed to a timeline before anyone had profiled what they were dealing with. A short discovery phase is the cheapest risk reduction you’ll find – it’s something we insist upon with any new project where the data is looking murky.
2) Settle the definitions before the technology. The same term – a “customer,” a “completed order,” an “active account” – often means something subtly different in each system. Reconcile that at the start, or it surfaces late, when it’s expensive to unpick. It’s dull, but it’s a job that often saves the project.
3) Ring-fence your team’s time, explicitly. These projects rarely stall on technology. They stall when the internal people needed for context and sign-off are buried in their day jobs. Decide up front who’s involved and how much and protect their time, or hand the work to someone else who can carry it.
4) Favour phases over a big bang. A staged delivery that proves value on one system or dataset first tells you whether the approach works before the whole budget is on the line. It’s slower, but it’s far less likely to end in a year’s effort that’s partly wasted.
While we manage data integration projects, this is all just as true if you tackle it in-house. But if you do weigh up outside help, the questions worth putting to any supplier are the plain ones: can they show you named clients who trusted them with this, where do they stand on security and accreditation, and how would you get your data and processes back if it ever came to that? We’re happy to answer all three (click here to book a call).
