TL;DR
Enterprise data platforms usually break not because of bad technology, but because organizations build them like projects to be shipped rather than products to be maintained.
TL;DR
Enterprise data platforms usually break not because of bad technology, but because organizations build them like projects to be shipped rather than products to be maintained.
2026-08-24 - Operating Models, Platform Ownership, Enterprise Data
There is a familiar rhythm to how large organizations build data platforms. A multi-month initiative kicks off with considerable energy. Consultants or internal engineering teams build out the storage layers, wire up the pipelines, and set up the reporting tools. Everyone celebrates the launch date.
Six months later, nobody knows who owns a broken pipeline, business users are exporting data back into spreadsheets to do their actual work, and the platform that cost millions to build is treated as a liability rather than an asset. I have watched this play out from inside the room more times than I care to count.
The root cause is rarely technical. It is almost always an operating model failure disguised as an engineering challenge.
The fundamental mistake is treating a data platform as a project with a finish line. A project has a start date, a delivery date, and a point where the team disbands and moves on to the next thing. Once the handoff happens, the architecture is left in the hands of whoever happens to be around when something breaks.
When you build platforms this way, you create an environment dependent on individual heroics. A senior engineer or a dedicated manager spends late nights keeping the system afloat because no permanent structure exists to maintain it. Scale cannot happen on the back of heroics. Systems eventually break when they depend on specific people rather than repeatable processes.
Moving from project delivery to platform ownership means changing how you staff, fund, and govern the system from day one. You do not just build a platform and walk away. You establish ongoing accountability for its uptime, its data quality, and its cost.
An effective operating model requires three distinct layers of ownership:
Without these lines drawn clearly, responsibilities blur. When everything is everyone's job, nothing is anyone's job.
Another symptom of a broken operating model is the rush to automate everything before the underlying controls are in place. Organizations often adopt modern tooling and fast-moving pipelines hoping it will fix messy data practices. It rarely works.
Automation on ungoverned data just means bad data moves through your systems faster. Controls and governance frameworks have to come first. When you establish clear rules for data lineage, definitions, and access upfront, automation becomes a multiplier for efficiency. Without those controls, automation becomes a multiplier for chaos.
When I think about enterprise data platforms, I look at them through the lens of a balance-sheet asset. You would never buy an expensive piece of commercial real estate, abandon maintenance on it the day the builders leave, and hope it retains its value. Data infrastructure deserves the same operational rigor.
If you want a platform that executives can rely on for critical decisions, stop treating its creation as an event. Treat its ongoing operation as the core of your data strategy.