Most data platforms do not fail dramatically. They degrade slowly. A workaround here, a schema change there, a new team with different requirements that gets bolted onto an architecture designed for something else.
By the time the symptoms are visible, slow pipelines, mounting exceptions, a data team spending more time maintaining than building, the original design is long gone. What remains is an accumulation of decisions made under time pressure, most of them reasonable at the time.
The right question is not "should we rebuild?" It is "do we understand what we have?" This checklist helps answer that.
The earliest sign that a platform has outgrown its design is usually not technical. A data team that dreads certain tickets, an engineer who knows which parts of the system to avoid or a product manager whose requests consistently take longer than expected.
These signals precede the technical symptoms. They are worth paying attention to before they become expensive.
Before any architecture review, get answers to these questions:
If the answers are uncomfortable, that is useful information.
The questions above tend to surface one of three situations. The first is manageable: the platform is sound but documentation has not kept pace. The fix is a sprint focused on documentation rather than code.
The second is harder: the architecture has accumulated technical debt that documentation cannot address. Core pipelines were not designed for current volumes. Key transformations are undocumented and fragile. The fix requires a migration plan.
The third is the most common: a mix of both. Some parts of the platform are solid. Others are held together by institutional knowledge and workarounds. The task is to separate them clearly.
Before the next sprint begins, take stock of where the platform sits relative to what the business needs from it. That assessment does not have to be exhaustive. Even a rough map of what is solid, what is fragile, and what is unknown changes the quality of the planning conversation.
The teams that avoid expensive data platform overhauls are not the ones that never have problems. They are the ones that catch problems early, at the point where they are still manageable. If you are not sure where to start, talk to our data team.