Work Hours
Monday to Friday: 8AM - 5PM
CST
The Unified Stack Is a Good Idea. Until It Isn’t.
Let us be direct: a unified data stack is a genuinely attractive proposition.
A single platform means a single support relationship, a single licensing conversation, a single set of documentation, and a consistent operational model across every layer of your data infrastructure. For an organisation starting from scratch, or one that has spent years managing poorly integrated point solutions, the appeal is rational and the benefits are real. Faster time to value. Lower cognitive overhead. Clearer accountability when something breaks.
The case for unified stacks is not wrong. It is incomplete.
What the Case Leaves Out
A unified stack is an excellent answer to today’s problem. It is a less reliable answer to the problem you will have in three to five years.
The history of enterprise data platforms is, in large part, a history of acquisitions, pricing restructures, end-of-life announcements, and strategic pivots that left customers mid-migration with a support contract and nowhere obvious to go. This is not a fringe experience. It is the normal lifecycle of enterprise software. The organisations currently being told to move from open-source ETL tools to commercial cloud platforms are not outliers. They are the latest iteration of a pattern that has been repeating since the first enterprise data warehouse was sold.
The unified stack solves the integration problem by making one vendor responsible for it. That is convenient until the vendor decides the economics no longer work, the product is discontinued, the company is acquired, or the pricing model changes in ways that no longer reflect the value being delivered.
At that point, the integration problem returns — and it is now accompanied by a migration problem.
The Question Worth Asking
The right question before adopting a unified stack is not “does this platform solve my current problem?” If it is one of the solutions in the Gartner® Magic Quadrant, it almost certainly does. The right question is: “what is my position if this platform is no longer available on acceptable terms in four years?”
If the answer is “we would have significant re-engineering work, meaningful data migration effort, and limited leverage in the renewal conversation,” that is a risk worth understanding and pricing into the decision. It is not necessarily a reason to walk away. But it should be explicit, not assumed away.
What Open Standards Actually Provide
The open-source stack is not a rejection of the unified stack idea. It is an attempt to achieve the same outcome — coherent, integrated, maintainable data infrastructure — without concentrating the risk in a single vendor relationship.
Apache Iceberg, for example, is a table format that any major query engine can read. Data stored in Iceberg on S3-compatible object storage is accessible to Snowflake, Databricks, Trino, DuckDB, and others. An organisation using this approach is not locked out of any platform. It simply retains the right to choose which platform queries its data, and to change that choice without migrating the data itself.
This is not a theoretical advantage. It is the difference between negotiating a renewal from a position of choice and negotiating it from a position of dependency.
The integration work is still real. Assembling, operating, and maintaining a coherent open-source stack requires engineering capability and ongoing discipline. That is not a task to underestimate. But it is a task that can be performed by a services partner without a software licence attached to the engagement — which means the incentive is to build the right thing, not the thing that maximises recurring revenue.
A Practical Position
Unified stacks are worth considering. They are particularly well-suited to greenfield environments, organisations without dedicated data engineering capability, and situations where time to value is the dominant constraint.
They are less well-suited to organisations that have been through a forced migration before, that operate in regulatory environments with strong data sovereignty requirements, or that are large enough that vendor pricing becomes a meaningful variable in their cost structure.
The honest answer is that the right architecture depends on the organisation’s actual constraints — not on which survey result a vendor’s partner community is currently citing.
What does not change is this: data stored in open formats, on infrastructure you control, with engineering capability that is not contingent on a single vendor’s continued existence, ages better than the alternative.
That has been true for a long time. It will remain true regardless of what the next unified platform is called.
Eumetix is a data engineering consultancy. We build and operate modern open-source data platforms for organisations that take their data seriously.