Gilding the ‘golden records’: finding a single source of truth in government

To build services around users’ needs, departments need a common frame of reference and ways to reconcile conflicting information. At a recent GGF webinar, experts from governments, the OECD and the private sector explored how civil service bodies can make progress on the thorny but unavoidable challenge of master data management
In Lithuania it started with a map.
When Gintautas Mežetis, director of the country’s Committee on the Development of the Information Society, began thinking about a data strategy with his colleagues, they drew out the existing inputs to government information systems.
“That was a very big map!” Mežetis told a recent Global Government Forum webinar. “We saw a mash of information, connecting to one system and then creating additional connections to others, where they were not needed.”

The map dramatised a problem common to governments across the world. States hold vast amounts of data about citizens and economies which, if marshalled well, could allow for the more efficient delivery of public services. Think of an automatically-generated tax return, for example, which a citizen would only have to approve rather than fill in.
Standing in the way is the bewildering variety of existing data architectures in governments. Information is often stored in silos, held in incompatible formats, and rife with errors and duplicates.
Mežetis and government practitioners from around the world gathered virtually to discuss one way to solve this problem and unlock the value of government held data: master data management. This depends on the use of ‘golden records’: single sources of truth stored in pre-agreed formats, operating as master lists of key datasets and providing an essential foundation for cross-departmental services and coordinated action.
A change in the DNA of government
The argument for using data to deliver better services is obvious. Recent years have seen unprecedented levels of personalisation in private sector service delivery, from music suggestions on Spotify to Instagram follower recommendations.
But these are digital-first companies, whose very existence is built on the management of vast quantities of highly-structured data that can be made available to the rest of the organisation on demand. And governments often start from an opposite premise: data is hoarded within specific departments, in a jumble of different formats, while privacy is sacrosanct.
Some of these problems are rooted in legislation: often, laws will state not on what needs to be done, but how it should be done – sometimes even specifying how the necessary information should be collected. Requiring a department to collect data in a unique format, rather than in the same manner as other departments, is a “big no no”, Mežetis argued. It is one way in which complexity has been added to state information systems over the decades, and it remains a problem today.
Centralise or devolve?
The panellists identified two ways in which civil servants can move towards a system of golden records, ensuring that at least one point in government holds up-to-date, accessible information in an agreed format.

First, a centralised government authority can be empowered to define how a specific piece of data – a citizen name or street address, for example – is to be structured. The rest of the government then has to adopt this format, while the central authority holds the ‘master’ data. APIs can be used to enable other departments to access these records – ensuring that it’s updated when required, and feeding through in turn to update other relevant datasets across government.
Gregory Kuhlmey, digital identity business development manager for the event’s knowledge partner, technology firm IDEMIA, noted that India’s government has taken this approach. A central state body built an identity system in which all citizens can enroll. It then allowed other government bodies and private sector institutions, such as banks, to interact with individuals based on this identity.
In the second method – favoured by Tomas Sanchez Lopez, chief data architect at the UK’s Office for National Statistics – individual data artefacts are assigned to the departments where they fit most naturally. A vehicle licensing authority, for example, might hold the golden record for driving license numbers and define how they should be stored, while the justice department might hold records of traffic violations.

“IT departments need to own the data that they collect, because they use it themselves for their own operational purposes. They know better than anybody else how to manage it,” Lopez said.
In either scenario, state bodies have to work with data in a format defined outside their own department. And any body that is the keeper of a golden record must accept incoming requests from other departments to query that information. Both features imply treading on existing domains and power centres within government.
Ben Welby, a policy analyst for digital government and open data at the OECD, warned: “That is a hard thing to do, because you are talking about changing the DNA of government.”
Evangelising ‘data as a product’
How can states go about achieving this change? The consensus among participants was to use carrots, rather than sticks. Rather than be compelled, different departments should be persuaded to adopt common formats on the basis that it will benefit their own service delivery.
In Israel Rachel Ran, data policy manager at the Ministry of Digital Affairs, has been leading what she describes as an evangelisation campaign – talking to “everyone and anyone, at every committee and every conference,” to persuade government and civil society actors of the value of adopting common standards.

The result is an iterative approach to data standardisation, with individual departments adopting new definitions and formats as and when that becomes possible. Ran’s ministry is first focusing on high-value data artefacts, which have the greatest potential to unlock action across government.
Ran uses the term “data as a product” to sum up the mentality her ministry is seeking to spread across the Israeli state. “We need to get our partners and the different stakeholders to view data not as a by-product of a new system or a new service or new digital infrastructure, but as a product on its own,” she told participants.
Welby from the OECD suggested another way to persuade departments to adopt common standards: creating clear institutional incentives. For example, government funding pots can be disbursed contingent on the adoption of common data standards; or business cases for new government projects can be evaluated on how they will contribute to a government data strategy.
Privacy by design
Another major obstacle for golden records will be gaining citizen consent for personal information to be shared.
One option, of course, is to compel citizens to agree to data sharing in order to access public services. But webinar participants warned against such a course of action. Instead, just as departments should be persuaded to adopt common standards, citizens too should see a clear benefit from opting in.

“If citizens don’t trust a [data sharing] platform, they will not embrace it and it will be a failure,” warned Kuhlmey from IDEMIA.
In a stylised example, a citizen applying for a new driving licence could opt for a form to be prefilled with information held by other government departments. Choosing this option would allow them to immediately submit a completed form. But they would be free to opt to enter information manually if they prefer, or to edit the pre-filled information.
“I would be glad if I would not need to even push a button to file my tax return,” Mežetis from Lithuania said.
Ran said that in Israel the plan is broadly to follow the transparency and privacy rules entrenched in Estonia’s pioneering X-Road system – giving citizens the ability to view how and when their data has been accessed, and by whom. The Ministry of Digital Affairs has coined the term “privacy by design” to succinctly communicate the idea to the public.
Another mechanism for gaining consent could be to keep data-sharing to a minimum. In some cases, government departments may not need to access a piece of data itself. Instead they can simply be told whether the data conforms to certain conditions. For example, when processing an application for a driving license, a department could request another department’s system to reveal whether the applicant is aged over 18. A binary yes or no response could be returned, instead of the age itself.
Lopez from the UK’s Office for National Statistics called this “a principle of proportionality”, with state bodies requesting exactly the information needed to complete a given process – but nothing more.
Room for difference
Despite their undoubted potential to make service delivery more efficient, there will be times when golden records are not the appropriate solution for a given data artefact.
Different departments may, for example, hold different numbers for the rate of unemployment, as their metrics will fit around their roles and interests: one measure might include those who are no longer seeking work, for example, while another excludes them. Ran from Israel’s Ministry of Digital Affairs pointed out that it is important for central data organisations to resist the temptation to impose a common format in such cases. Instead, they should investigate why the numbers are different, and then decide if there is a justification.
When it comes to unemployment in Israel, for example, “one agency defines who is unemployed, and another agency defines who is entitled to benefits. And it’s not the same definition.” Imposing a common standard in such a case could cut benefits from justified recipients.
The example sums up why golden records need to be pursued with care and attention in government. Where the state is involved, the potential impact on users is never small.