What is data actually worth?

Data that creates no value is a cost

It's easy to talk about data as valuable in itself: "data is the new gold" and similar phrases. The reality is less poetic: if you have data that doesn't create value, it's just a cost. If that data is also poor quality, it's a cost and a risk. Data isn't valuable because it exists. It's valuable because it's used.

That sounds simple, but most organizations sit on enormous amounts of data without a clear plan for turning it into anything. This article is about exactly that: the concrete paths to turning data into money, and what it actually takes to pull it off.

What you don't know you know

A useful way to think about data and information comes from philosopher Rafael Echeverría (1995), who divides knowledge into four fields:

What we know we know: the obvious data, already in use.
What we know we don't know: known knowledge gaps, things we can actively go looking for.
What we don't know we know: data that already exists in the organization, but that no one has connected to a question that would make it valuable. This is usually where the easiest win sits.
What we don't know we don't know: blind spots, only visible in hindsight.

That last field is also where Nassim Nicholas Taleb's concept of the black swan belongs: the rare, extremely improbable events that still end up having enormous impact precisely because no one saw them coming, and that in hindsight always get explained as if they should have been obvious. That's a big enough topic for an article of its own, but worth mentioning here: no amount of data eliminates the fourth box, it can only shrink it a little.

The point of the matrix isn't that it's scientifically precise, but that it forces a question most people skip: before collecting more data, what do we already have that we're not using?

From data to action

Data by itself does nothing. The path from data to actual value is usually described in three steps:

Data → Insights → Action, where each step corresponds to a different type of analysis:

Descriptive (what happened?): requires the least data quality, the most human interpretation.
Predictive (what will happen?): requires higher data quality, less human involvement.
Prescriptive (what should we do?): requires the highest data quality, the least human involvement. The system suggests or even executes the action directly.

The further toward action you move, the less tolerance there is for poor data quality. A dashboard with a bit of junk data is annoying. An automated decision system with a bit of junk data is dangerous.

Three ways to actually make money from data

Barbara Wixom and her co-authors of the book Data Is Everybody's Business describe three fundamentally different ways to turn data into money, and it's worth knowing all three, since they require entirely different internal capabilities:

Improve: use data to make what you already do better. More efficient processes, better decisions, lower costs. This is the low-hanging fruit, the one that requires the least new investment and gives the fastest results.

Wrap: package data or insights around an existing product or service so it becomes more valuable to the customer. A good example is when a physical product gets a digital layer on top. Not a new product, but the same product made better with data.

Sell: sell data or data-driven insights directly, as a product of its own. This is the most demanding path, both legally and technically, but also the one that can pay off the most if done right.

In practice, the vast majority of returns from data initiatives come from improve, a smaller but meaningful share from wrap, and the least from sell, worth remembering before jumping straight into "selling data" as a first move. Most organizations would gain more from first getting genuinely good at improve.

The five capabilities that decide whether you can pull it off

Choosing a path (improve, wrap, or sell) isn't enough if the organization lacks the underlying capabilities. MIT CISR's research identifies five capabilities required to monetize data at all:

Data Asset: do you have data people can actually find, trust, and use?
Data Platform: can data be processed and delivered reliably and quickly?
Data Science: do you have the scientific methods and algorithms that detect what people can't see on their own?
Customer Understanding: do you understand your customers deeply enough to know which needs the data should actually meet?
Acceptable Data Use: do you have the governance in place, both compliance-wise and ethically, to actually be allowed to use the data?

Each capability can be measured across maturity levels, from foundational to advanced, and it's rarely the technical capability (data platform, data science) that's the bottleneck. It's usually data asset and acceptable data use, the less glamorous capabilities, that decide whether the rest can even be built on top.

Cost or capital

So, back to the question in the headline: what is data actually worth? The answer depends entirely on whether you have the capabilities to turn it into something, and whether you've chosen the right path (improve, wrap, or sell) given where you actually stand today. Data that just sits there is a cost. Data connected to an ability to act on it is capital.

That leads to a bigger question, one that deserves its own answer: if data is genuinely capital, should it show up in a company's valuation, the same way other assets do? That's exactly what the next article gets into.