Your Data Problem Is Not What You Think

At some point in the life of most startups, someone in a leadership meeting says some version of the same thing: “We have all this data and we’re not using it.” The response is usually to buy a tool. A BI platform, a data warehouse, maybe a dashboard that shows the numbers everyone’s been asking about.

Six months later, the dashboards exist. The numbers still don’t quite make sense.

The problem was never the tool.

Something I hear constantly from founders is some version of: “we have a lot of data.” And they’re right: their systems are generating data all the time. But generated data is not the same as collected, prepared, and usable data. There is usually a significant gap between the events firing in your product and the point where any of that means something to your business. Closing that gap is the actual problem.

Your production database was not built for questions

When your engineers built your product, they made the right call: they designed the database to serve your users. Fast writes, consistent reads, minimal latency. That is exactly what a production system should do.

But that same database is a terrible place to ask business questions. “How many customers churned last quarter after upgrading their plan?” is not a query your production database was designed for. Running it there is slow, risky, and produces results that are hard to trust because the data is structured around operational transactions, not analytical thinking.

This is the core problem, and it has nothing to do with your tools. Your production system and your analytics layer are doing fundamentally different jobs. When they’re tangled together, both suffer.

The separation that changes everything

The companies that get data right draw a clear line between the two. Production systems serve your product and your users. The analytics layer, built separately, serves your business and your decisions. Data flows from one to the other, transformed into something that reflects how your business actually works.

Once that layer exists and is trustworthy, everything else becomes possible. Your dashboards pull from a single, consistent source. Your growth team can answer acquisition questions without waiting for an engineer. Your data products (recommendations, scoring models, customer health signals) have a reliable foundation to build on. None of that works without the separation.

Without it, every data product your team tries to build is resting on shaky ground. Analysts spend their time arguing about numbers instead of finding insights. Engineers get pulled into reporting work that distracts from the product. And the business keeps making decisions on intuition because the data is technically there but practically unusable.

What this means if you’re a founder or CEO

You don’t need to understand data engineering to make good decisions about data. But it helps to understand that the question is not “which tool should we buy?” It’s “do we have a trustworthy analytics layer, and is it decoupled from our production systems?”

If the answer is no, no amount of tooling will fix it. If the answer is yes, even relatively simple tools will take you a long way.

Most of the data problems I see at startups and scaleups trace back to this. The investment that compounds over time is not in dashboards or platforms. It is in getting the foundation right: clean, well-modelled data that your whole business can trust and build on.

If you’re trying to work out where to start, I’m happy to have that conversation. No jargon, no sales pitch. Just a clear-eyed look at where you are and what would actually move the needle.