What would happen if London could source, analyse and act upon its public sector data at a city scale?

This question formed the basis of a year-long pilot of a London Office of Data Analytics (LODA), a collaboration between the GLA, Nesta, 12 London boroughs and ASI - a data science firm. Together, we sought to develop a machine-learning model to find unlicensed houses in multiple occupation (HMOs), properties that are often linked to dangerous and exploitative living conditions and remain largely unknown to local councils.

In this report, we outline the pilot’s origins, methods and what we have learned to date. The report presents a series of key recommendations, which will be used to inform future LODA projects, including the continued development of the HMO model. Public sector bodies embarking on other collaborative data initiatives should find our lessons and recommendations equally useful.

Finally, the report also outlines a roadmap for a permanent LODA within the Intelligence Unit at the GLA, which has now been established by the Mayor of London.

Key findings

• Technology: local authorities that do not have the ability to join up and match records held in different IT systems within their own organisation will find it extremely challenging to collaborate with other organisations and their data. Public sector organisations should prioritise future IT investment in data matching tools that enable them to link records across systems and facilitate analysis. Conducting an organisational data maturity assessment at the start of a project can also bring to light important gaps and challenges relating to technology and data management early on.
• Data: a lack of standardisation in frequently used field names (e.g. lines of an address) in different IT systems makes joining up data for analysis much more difficult than it needs to be. For place-based data, public sector organisations should commit to using Unique Property Reference Numbers (UPRNs). Significant gains can be made to improve the usability of data at relatively little cost, as only an organisational commitment to recording data consistently is required.
• People: projects like LODA are more about getting people to collaborate in new ways across organisations than they are about doing new things with data and technology. Data projects can never just be delegated to data science teams; they must be organisation-wide efforts. In particular, frontline staff must be involved in the process. In-house data analysts should also be given the opportunity to work with service managers to tackle public service challenges. Strong leadership is necessary to support these major process and cultural changes.
• Process: data analytics projects that are technically challenging, have changing data requirements, involve many partners and are culturally new require a more adaptive approach to project management to deal with higher levels of complexity and uncertainty. Such initiatives should adopt an agile approach to project management, which builds in regular opportunities for analysis, testing and reflection.
• Legal: different organisations have varying levels of risk appetite for sharing the same type of data, even in cases where non-personal and non-sensitive data is involved. While this is difficult to overcome, consistent legal advice and common interpretation of the same data legislation could provide assurance and speed up the process of data sharing. A future London Office of Data Analytics should include a legal function that can provide consistent guidance on data sharing.

Authors

Nevena Dragicevic, Eddie Copeland, Nesta, and Andrew Collinge, Paul Hodgson, Wil Tonkiss and Alan Lewis, Greater London Authority