Companies are failing to recognise that the data they generate should be economically defined as capital, which is having an impact on their competitive strategies. What does this mean? The textbook definition of capital is that it is a produced good, as opposed to a natural resource, and is a necessary input to create other goods or services.
It’s what economists call an economic factor of production and if you don’t have it, you can’t make the things you may be interested in making. This is the view of Oracle’s senior data strategist, Paul Sonderegger, whom also argues that because organisations aren’t thinking about data in the right economic context, they’re flailing to unlock their internal “hidden data economies”.
The conversation I had with Sonderegger was very interesting, as his thoughts on company data nicely form a framework for companies thinking about their data strategies. And this is an area that we know enterprises find particularly challenging. He explained:
When you think about algorithmic services, for example, if you don't have the dataset that that algorithm requires, it is an engine with no fuel, that service literally cannot exist. And so, as a result, data is in fact a kind of capital, even though accounting rules do not allow it to be accounted for as such.
One of the things that is really curious about data as an asset is that the vast majority of it never goes to market. The vast majority of data gets created and used inside the same company. this would be a little bit like an oil company burning all of the oil that extracts out of the ground instead of selling it for you. And that’s a little strange.
As such, Sonderegger argues that because the vast amount of data creation in the digital economy happens inside companies, these companies need to recognise that they have an internal data economy that is “hidden in plain sight”.
What does this mean in practice?
If we accept that data is defined as capital, this has implications for how we think about the supply and demand of data, Sonderegger argues. He said that when application developers create apps, they’re not thinking very much about what an analyst would need from the app’s data in order to run their analytics effectively or efficiently. They’re just creating a data asset the way they want it, which leads to a diversified supply base.
Then, on the demand side, you have analysts and data scientists who are trying to discover the data assets available to them and are trying to get them into the shape that they need. Sonderegger explained:
And there are no market signals inside the firm that indicate which datasets are the most scarce, the most valuable - the datasets that are going to be the most effective in changing some outcome, some decision or action.
Individual departments aren't going to start buying and selling data assets to one another. So what do you do? So if you have a hidden data economy, and the transaction costs of getting data from its point of creation to its multiple points of views are too high, but you're not actually going to have true market competition to drive that down, what do you do?
Clearly in line with Oracle’s product portfolio, Sonderegger argues that this is why autonomous data management is necessary. If a company has an autonomous data platform that handles all of these data management tasks automatically, on its own, and makes the data easier to repurpose, this would drive down the transaction costs of getting data from its point of origin to its multiple points of use. Transaction costs in this context are the time that analysts and data scientists spend looking for the data and making it useable. Sonderegger added:
So the key, we think, is a data exchange to make it easier to discover these data assets - not just that they exist, but what they're like, what they contain.
Sonderegger adds that there are three critical attributes to data in the enterprise that companies thinking about their data strategies should be aware of. These are:
Data is non-rivalrous. This means that multiple people can use the same data set simultaneously (unlike physical capital assets).
Data is also non-fungible. This means that one data set is not equivalent to another, which brings data scarcity into content. The implication for this is that companies are actually in competition for datasets and this should direct your competitive strategy.
Data is an experience good. In other words, you don’t what it’s worth until after you have and after you’ve used it.
So, knowing all of this, what should an enterprise wanting to get to grips with its data do? What approach should it take? Sonderegger suggests:
It’s just all about following the money, and helping the data to flow to where the money either is spent, or where it is generated. And so companies should first look at, where are your big costs generated? And where do you generate the most revenue? And then look into those processes, and the decision points in those processes, looking for opportunities to inject AI there.
The big challenge really is about imagination. The big challenge is really seeing the data that isn't there yet. That means looking at interactions with customers, looking at interactions with partners, and imagining how they might be not just more efficient, but might be redesigned entirely.