A lot of the work that we’ve done at the ODI over the last few years has involved helping organisations to recognise their data assets.
Many organisations will have their IT equipment and maybe even their desks and chairs asset tagged. They know who is using them, where they are, and have some kind of plan to make sure that they only invest in maintaining the assets they really need. But few will be treating data in the same way.
That’s a change that is only just beginning. Part of the shift is in understanding how those assets can be used to solve problems. Or help them, their partners and customers to make more informed decisions.
Often that means sharing or opening that data so that others can use it. Making sure that data is at the right point of the data spectrum helps to unlock its value.
A sticking point for many organisations is that they begin to question why they should share or open those data assets, and whether others should contribute to their maintenance. There are many commons questions around the value of sharing, respecting privacy, logistics, etc.
I think a useful framing for this type of discussion might be to distinguish between data assets and data products.
A data asset is what an organisation is managing internally. It may be shared with a limited audience.
A data product is what you share with or open to a wider audience. Its created from one or more data assets. A data product may not contain all of the same data as the data assets it’s based on. Personal data might need to be removed or anonymised for example. This means a data product might sit at a different point in the data spectrum. It can be more open. I’m using data product here to refer to specific types of datasets, not “applications that have been made using data”
An asset is something you manage and invest in. A product is intended to address some specific needs. It may need some support or documentation to make sure it’s useful. It may also need to evolve based on changing needs.
In some cases a data asset could also be a data product. The complete dataset might be published in its entirety. In my experience this is often rarely the case though. There’s usually additional information, e.g governance and version history, that might not be useful to reusers.
In others cases data assets are collaboratively maintained, often in the open. Wikidata and OpenStreetMap are global data assets that are maintained in this way. There are many organisations that are using those assets to create more tailored data products that help to meet specific needs. Over time I expect more data assets will be managed in collaborative ways.
Obviously not every open data release needs to be a fully supported “product”. To meet transparency goals we often just need to get data published as soon as possible, with a minimum of friction for both publishers and users.
But when we are using data as tool to create other types of impact, more work is sometimes needed. There are often a number of social, legal and technical issues to consider in making data accessible in a sustainable way.
By injecting some product thinking into how we share and open data it might be helpful in addressing the types of problems that can contribute to data releases not having the desired impact: Why are we opening this data? Who will use it? How can we help them be more effective? Does releasing the data provide ways in which the data asset might be more collaboratively maintained?
When governments are publishing data that should be part of a national data infrastructure, more value will be unlocked if more of the underlying data assets are available for anyone to access, use and share. Releasing a “data product” that is too closely targeted might limit its utility. So I also think this “data asset” vs “data product” distinction can help us to challenge the types data that are being released. Are we getting access to the most valuable data assets or useful subsets of them. Or are we just being given a data product that has much more limited applications, regardless of how well it is being published?