What’s the ideal model for an Office of Data Analytics?

by Eddie Copeland

For the past several years, I’ve argued that the UK’s cities and regions should establish offices of data analytics (ODAs). Inspired by US cities like New York and New Orleans, the idea was to set up dedicated teams of data scientists who could help cities source, analyse and act upon data from multiple local government and public sector partners to improve decision making and reform services.

Having that kind of function could help cities implement tried and tested ways of doing more and better with less. This would include designing shared services, intelligently coordinating the actions of different teams, and – best of all – predicting and preventing problems from happening in the first place.

I’ve also argued that ODAs are key to making a success of city devolution. How, I wondered aloud, would the new generation of metro mayors be able to fulfil their remit of reforming local public services, boosting local economic growth and strengthening local democracy if they didn’t have a mechanism to get the data on all those things from across their whole region?

It has therefore been pleasing to see that many UK cities and regions have reached the same conclusion. Manchester has its GM-Connect programme. Worcestershire has created an Office of Data Analytics (WODA). Mayor Andy Street recently announced the creation of an ODA for the West Midlands. London has been piloting a LODA. For the prize of best ODA-inspired acronym, we must keenly await the creation of an ODA for Yorkshire.

I used to believe that creating an Office of Data Analytics essentially entailed having a crack team of data experts, based in a city hall, who could make all this happen. Having worked on several ODA pilots at Nesta, I now realise that was naive on at least three fronts.

First, with demand (and therefore salaries) rising fast for data scientists, many cities will struggle to hire such a team.

Second, an ODA was never going to work just with a group of data scientists. To achieve data-enabled public service reform you need legal expertise to create information sharing agreements; technical support to integrate IT systems; communications to ensure data is used in a transparent way that warrants public trust; a political liaison function to ensure each project has the backing of leaders; and project management resources to pull the whole thing together.

Third, as well as being practically difficult for a mayoral office or combined authority to provide these resources, it may not be desirable for all of them to be in-house. The UK is blessed to have many outstanding organisations who can help on all these fronts, so why not use them?

In short, Nesta’s view is that it is possible to create a better, more open and more inclusive ODA model that draws strengths from many different local groups and individuals.

Proposal for a New ODA Model

The function of a modern ODA should be to oversee and project manage an end-to-end process made up of six steps.

New Nesta ODA model

Let’s briefly explore each one.

1 – Generate Ideas

Ideas for new data projects could be gathered from many different people. They  could come from mayoral manifesto pledges. They could be solicited from local government and public sector organisations. They could come from the public – potentially crowdsourced via digital democracy platforms like Your Priorities.

The ODA could offer a couple of windows each year for ideas to be proposed in these different ways. This open process would help demonstrate that the ODA was designed to benefit everyone across the city or region, not just city hall.

2 – Assess Feasibility

The ideas generated in step 1 need to be vetted for feasibility. Data science is, after all, just one set of tools in a wider toolbox. On this front, the NOLAlytics team in the New Orleans’ Office of Performance and Accountability has helpfully defined six opportunities that data analytics can enable. These include:

  • Finding the needle in a haystack

  • Prioritising work for impact

  • Creating early warning tools

  • Making better, quicker decisions

  • Optimising resource allocation

  • Experimenting for what works

Having checked that a given idea meets one of these six types, Nesta’s ODA approach has been to use nine principles (outlined below) to test the feasibility of tackling a specific challenge with data. We’ve found this requires clear thinking under three headings: 1) the problem we wish to address; 2) the new action we want to enable; and 3) the data relating to that issue.

I’ve previously written about the thinking behind each principle. Critically, ODAs would ensure this step was always conducted with the full involvement of the staff (and especially the frontline workers) who deliver the service.

Nesta’s nine principles of reforming public services with data

9 Principles of ODA Model

3 – Design Project

Once a suitable problem and desired outcome have been clearly defined, the ODA would select one of several project methodologies, which might include:

  • Running a Pilot: The ODA designs and manages all facets of the project in collaboration with partners (see step 4).

  • Challenge Prize: The ODA offers a cash reward to the organisation or individual who provides the best data-enabled solution to a defined problem.

  • Hackathon: The ODA organises one or more hackathons for smaller scale problems where data is abundant, or as an initial phase of the Pilot or Challenge Prize methods.

As well as selecting and designing the best methodology, the ODA would define success criteria at this stage to ensure the intended impact of each project was transparent and measurable.

4 – Identify Partners

The ODA would assess its own internal capacity and identify any gaps in covering the data science, technology, project management, legal, and communications functions of each project.

Using its network across the region, the ODA would seek partner organisations to fill those gaps. Partners could include local authorities, public sector bodies, tech firms, data science SMEs, academic institutions, think tanks, civic hacker groups, and so on.

These partners could be secured on a voluntary basis (there’s plenty of organisations willing to work on projects that benefit their community), or be hired via a simplified procurement framework, put in place by the ODA.  

5 – Execute Project

The data project would then be carried out by the ODA and its partners.

If running a traditional pilot, this would likely involve:

  • Onboarding of partners and agreeing responsibilities

  • Conducting data maturity assessments (for any partner supplying data)

  • Creating Information Sharing Protocols

  • Exploring, acquiring and processing data

  • Analysing data

  • Testing and evaluating

  • Reporting results

For a challenge prize, a methodology similar to that used by Nesta and the Open Data Institute in the Open Data Challenge Series could be applied.

And for a hackathon… well, there are plenty of organisations who know how to run those well.

6 – Codify and Learn

The final phase would be for the ODA to assess the impact of each project and share the lessons learned. This would be likely to include:

  • Assessing the impact of the project against the success criteria developed in step 3. This could potentially include rigorous evaluation methods such as a Randomised Control Trial.

  • Communicating the lessons learned in regular blogs throughout the project, in a final project report, and through workshops and events.

  • Releasing codes, templates, tools, and any other materials that could support others in applying the same work or designing something new.

It’s hard to over emphasise the importance of this phase. Done well, it would help systematically lower the barriers to trialing new data projects, and help build a library of proven data-enabled interventions that could be adopted by other cities and regions.

The essential and the nice to have

With this model, an ODA would have two essential functions:

  1. Doing: conducting several projects per year based on the six stage process described above.

  2. Templatising: creating reusable codes of ethics, data standards, legal documents, process guides and open source tools.

If resources were available, two additional functions would be desirable:

  1. Convening: acting as a hub for the city’s data science and policy community; supporting and nurturing the ecosystem of organisations using data to drive public service reform.

  2. Training: running courses catered for the specific needs of public sector leaders and data science practitioners to increase both groups’ understanding of the other’s needs.

In summary

Effective data projects are not just about hiring a crack team of data experts. Reforming public services requires a much broader network of talents – not all of which can easily be found inside the public sector.

The UK’s cities rightly celebrate the richness and diversity of their tech sectors, their universities, their civic hacker groups, and their local businesses. They assert their commitment to engaging citizens in the decisions affecting their communities. The ideal ODA model would ensure all these groups and individuals have a chance to participate.

I still believe that having a dedicated ODA team with a presence at the heart of local government is vital to ensure that the smarter use of data becomes thoroughly embedded in how our cities and regions are run. It’s for each locality to decide the details of what that core team looks like. But there’s no reason why those ODAs shouldn’t augment their own capacity by making the most of the talents on their doorstep.

Find me on Twitter

You may also like