It’s no secret that Nesta’s current programme of work to pilot offices of data analytics in London (with the GLA and 14 boroughs) and the North East (with seven local authorities, the Digital Catapult and Sunderland Software City) is inspired by the Mayor’s Office of Data Analytics in New York City (MODA).
Yet it’s been apparent from the very start that this will not be a simple case of copy and paste. New York is not like any other city in the US, let alone any in the UK. I could provide a whole list of relevant differences, but one stands above the rest: the fragmentation of UK local government.
Whereas in New York many issues are coordinated by one city-wide organisation, in London most public services are delivered by the boroughs. As a result, issues tend to be identified, addressed and reported on by 32 separate teams in as many different ways (except for specific cases of cross-borough cooperation or shared services). Likewise, in the North East, the seven local authorities individually look after the issues that affect their areas.
That creates a challenge for applying MODA’s 10 step methodology for addressing a public service problem with data. The first two steps are understanding 1) how a team delivers its particular service, and 2) how they use data.
Instead of examining how just one organisation does these things (a complicated enough task), tackling issues in UK cities requires running the same process with as many teams as there are local authorities. All may operate in different ways, turning it into a daunting project.
So what’s the solution?
In the presence of too much information, focus is required. Our approach is therefore to start small, then scale fast. In London, where we’re exploring the feasibility of tackling unlicensed HMOs, instead of diving straight into asking every single borough how they inspect HMOs and use data to identify potential unlicensed properties (a process that could lead to dozens of different answers), we’re starting by looking at how just a few operate. This week will see us visiting Westminster and Lambeth for that purpose.
Those early conversations will help the ASI – our data science team – get up to speed on current practice. Who typically does the inspections? How many inspections are conducted per week? Are inspections proactive as well as reactive? And so on.
We’ll also be asking those initial boroughs to provide a list of all the datasets they believe correlate with a particular property being an HMO. In addition, we’ll request that they describe each dataset, since different local authorities may call them by different names. We can then circulate this list to all the other boroughs to ask if they have similar records. In this way, we can build up a picture of how much data diversity there is across London, and where there are gaps.
With all this information in hand, next week the GLA’s Andrew Collinge will host all 14 pilot boroughs at City Hall to review the datasets and to jointly discuss how HMOs are handled across London. We believe this approach provides the best of both worlds: quickly getting to grips with the details of the service by talking to a few, but then validating the findings with the whole community to ensure that initial assumptions are correct and that nothing has been missed.
The intention is that these discussions will enable the creation and testing of a minimal viable data model (MVDM). That MVDM will make it simple for each borough to understand exactly which datasets are required for it to work in their area. It can then be extended, growing in sophistication as it scales to more local authorities.
Already it seems likely there will be further hurdles to overcome.
Some local authorities may simply not collect the datasets the model requires. (There’s an interesting question around whether the data that public services collect in order to report against their KPIs are actually those needed for reforming a service). Yet that in itself is a useful insight. It’s surely empowering for public sector leaders to know that if they started collecting data in a certain way, then they would be able to benefit from all the advantages that a city-wide predictive algorithm can bring.
The second major challenge comes with the intervention itself. Our working assumption is that asking 14 different teams to completely change the way they operate is a non-starter. We therefore plan to stick to a rule of thumb that informed MODA’s early work: the data intervention must not require front line workers to change what they do. If the HMO project works, we imagine inspectors will still be sent to visit properties. The only difference is that their list will have been pre-prioritised to focus on the most statistically likely offenders first.
No doubt many other challenges will arise; we’ll document and share them as we go. The aim is that by the end of these pilots we’ll have a much better idea of what a UK Office of Data Analytics methodology looks like, allowing many more cities to address their own priorities with data.
Follow Eddie on Twitter
Image credit: Pixabay CC0 license