Most data projects don't fail on technology. They fail on things that were visible from day one - no one noticed them, or didn't want to notice them.
Here are five warning signs before your first euros are spent.
- Nobody knows who the business owner is.
„Data belongs to IT.“ If you hear this phrase at a kick-off, the project has a problem. A data project without a specific business owner – someone who defines what the output should mean and is responsible for its use – will end up with a report that no one uses.
Signal: Only IT people are present at the first meeting.
- No one has seen the source data
The team designs the architecture, estimates the workload, and plans milestones—without looking at the actual data from the source systems. Volume, quality, format, duplication—all come as a surprise later.
Signal: Source data analysis is not part of phase 1.
- Scope is growing every week
„And could we add…?“ Every new requirement without formal approval and impact assessment breaks the project apart. After three months, the team has twice the scope, the same budget, and half the time.
Signal: There is no change request process. Every change is taken for granted.
- The platform was chosen rather than the requirements
The company has decided on a particular technology – and only then is it looking for a use case to justify it. Technology is not bad. But when the decision precedes analysis, the project serves the platform, not the business.
Signal: The first meeting is more about technology than the business problem.
- The team doesn't understand the domain
An experienced data engineer who has never worked in banking can build a pipeline. But they can't tell why two records with the same client are correct or incorrect. Domain knowledge is not a bonus - it's a requirement for making good modeling decisions.
Signal: No one on the team comes from the client's industry and no onboarding with business experts is planned.
What about it?
None of these signs alone means a project is bound to fail. But the combination of two or more of them without active resolution significantly increases the risk.
The good news is that they are all identifiable within the first two weeks – and addressable before they become expensive.