Here is the most uncomfortable thing I've learned leading a data platform: you can build a technically excellent platform — clean pipelines, trustworthy models, fast, governed, genuinely good engineering — and watch it change nothing, because the organisation around it doesn't have the culture to use it. I've seen beautiful platforms sit underused while decisions got made the same gut-feel way they always had, the data right there, ignored. It's a deflating lesson for someone who loves the engineering, and it's the most important one I've absorbed: the platform is the easy part. The culture that decides whether anyone actually uses it is the hard part, and the one that matters.

The seductive trap of the platform-first mindset

The reason this lesson is hard to learn is that the platform-first mindset is so seductive, especially to technical people like me. Building the platform is a concrete, tractable, satisfying problem. It has clear requirements, a clear notion of "done," and it rewards the skills I'm proud of. Culture is vague, slow, human, and frustratingly resistant to being engineered. So the temptation is enormous to pour your energy into the part you can build and quietly hope the using-it part sorts itself out.

It does not sort itself out. A platform is a capability, and a capability that no one has the culture, the trust, or the habit to use is just expensive potential energy that never converts to anything. I've watched organisations invest heavily in the technology of data and almost nothing in the culture of data, and then wonder why the return never materialised. The answer is that they built the road and never taught anyone to drive, or gave them any reason to want to.

What a data culture actually is

So if culture is the thing that matters, what is it, concretely, beyond a fuzzy buzzword? In my experience it's a handful of very human things, none of which a platform provides:

  • Trust in the data. People will only use data to make decisions if they believe it — and belief is earned through reliability, transparency, and enough good experiences to build confidence. A platform can make data trustworthy; it cannot make people trust it. That's a relationship, built over time, and it's fragile — one prominent wrong number can set it back months.
  • The habit of asking the data. A data culture is one where, faced with a decision, people's instinct is to go and look at what the data says, rather than reaching first for gut and opinion. That instinct is a learned habit, and building it means making the asking easy, rewarding, and normal — cultural work, not technical work.
  • Comfort, not fear, around data. In a lot of organisations, data quietly makes people anxious — anxious they'll be exposed by it, anxious they don't understand it, anxious it'll be used against them. A healthy data culture lowers that fear, so people engage with data as a helpful tool rather than a threat. That's about psychological safety and how data is used on people, which no platform touches.
  • Shared literacy and language. People across the organisation being able to read data reasonably, and share enough of a common language about it, so the conversation between the data team and everyone else actually works. This is the communication layer, and it's cultural to the core.

Look at that list and notice: not one item is a technology. Every one is a human, relational, communicative thing. Which is exactly why the platform-first mindset fails — it's optimising the one dimension that wasn't the bottleneck.

Building culture is a communication job

If culture is the hard part and it's fundamentally human, then building it is — once again, and I've stopped being surprised by this — a communication job. The data leader who changes an organisation isn't the one with the best platform; it's the one who can build trust, teach the habit, lower the fear, and grow the shared literacy. All communication. All relationship. All the slow, human work that the engineering-loving part of me finds harder and less immediately rewarding than building the platform, and that I've had to learn to value as more important.

In practice, this has reshaped how I spend my time as a leader. Less "is the platform perfect" and more "does anyone trust it, use it, understand it — and if not, what relationship or fear or literacy gap is in the way." It means treating the adoption and the trust as first-class deliverables, not as things that'll follow automatically from good engineering. It means communicating relentlessly — explaining, translating, reassuring, teaching — because that's what actually moves the culture, and the culture is what actually moves the organisation.

A data platform is potential energy. Culture is what converts it into anything real. Build the finest platform in the world without the culture to use it, and you've built a beautiful, expensive monument to a decision nobody changed.

Where this leaves me as a leader

I still love building the platform — that's not going away, and a bad platform certainly undermines any culture, so the engineering genuinely matters. But I've reordered my sense of what the job actually is. Building the platform is necessary and it is not sufficient, and the sufficient part — the part that determines whether all the engineering amounts to anything — is the cultural, human, communicative work of getting an organisation to actually trust, use, and think with its data.

That's harder than the engineering, slower than the engineering, and less satisfying to my technical instincts than the engineering. It's also the whole point. So the advice I'd give any data leader, and that I keep giving myself: by all means build a great platform — and then spend at least as much of your energy on the culture that decides whether it was worth building. Because a platform without a culture is a road no one drives, and the road was never the hard part. Teaching people to want to drive it, and trust where it goes, always was.