Data is a strategic asset

There was consensus that data needs to be clean, harmonised and connected long before AI arrived. Hygiene has become the whole conversation, so companies spend years polishing data. Clean data is the floor of a data strategy, and without it nothing will ever make sense, but it should not be mistaken for the ceiling.

I find it more useful to think about construction layers (that good old dinosaur of a metaphor) where each one only looks good if the one beneath it holds. The first two are about the data itself and the last two are about the people around it, which is where a lot of buildings stall.

The ground: build on land you own

A lot of brands build their customer understanding on rented land. Platforms, retailers and agencies hold the richest signals, and they set the terms. First-party data is the plot you actually own, but ownership alone doesn't make it valuable. Martin Casado and Peter Lauten at a16z argued that data is rarely the moat it is claimed to be, because each extra data point tends to add less than the one before. So the value sits in data nobody else can get and that refreshes every time you trade with a customer, not in the size of the pile.

Consent is the title deed on that land. Cisco's 2026 privacy benchmark found that 99% of organisations see at least one tangible benefit from their privacy investment, yet only 12% describe their AI governance committees as mature. Without a clean deed you can still build, but not very high and not for very long.

The frame: keys that carry the load

Data starts carrying weight when a customer, a product, a market and a week mean the same thing in every system. Without those shared keys, the same shopper in your app and in your store becomes two customers, and until that is fixed, personalisation is mostly guesswork with a prettier interface. The same goes for the commercial chain from media to market share to sell-in and sell-out, which often never meets even when every part of it is perfectly clean. Sell-in can rise while sell-out stays flat, which means you grew stock in the channel and not demand.

There is a reason these systems don't connect. Melvin Conway observed in 1968 that organisations design systems that copy their own communication structures. If media, sales and insights rarely sit at the same table, their data won't either, so the frame is as much an organisational decision as a technical one.

The valuation: give it an owner and a price

Here is the uncomfortable part of calling data an asset: it appears nowhere on the balance sheet. Doug Laney, who coined the term infonomics, puts it plainly: "when you don't measure an asset, you don't manage it". He argues the CFO should be the organisation's chief data valuer, because finance already knows how to think about depreciation and risk.

You don't need an accountant's precision to start, but you do need owners. McKinsey's work on data products suggests most of a company's data value comes from just five to fifteen of them, and that the value grows with every use case that reuses one instead of building its own version. A customer dataset with a named owner, a clear purpose and a rough value is an asset. Otherwise, it's a cost centre with good intentions.

The occupants: decisions, not dashboards

A building is only worth something if people live in it. Brynjolfsson, Hitt and Kim studied 179 large US firms and found that those making decisions based on data had 5 to 6% higher output and productivity than their IT investment alone would predict. The difference was how decisions got made, not how much data sat underneath them. Wavestone's 2026 survey of data and AI leaders points the same way, with 93.2% naming culture and change management as the biggest barrier.

This is where the data team's mandate comes in. A lot of data and insights teams work like the handyman on call: requests come in, dashboards go out, and nobody checks in with the architect on what the building needs next. The team needs its own roadmap with the leadership's blessing, and a global set-up where the global organisation owns the standards and core data products while the teams downstream own the use cases. That sounds simple, and it's glaringly not. It means a leadership team giving up some of its own requests so the data team can build what everyone will need later.

Clean data gets you a foundation. Ask your data team what they would build if nobody sent them a request for a quarter, and then let them.