Nesta’s Offices of Data Analytics (ODA) programme aims to help cities and regions join up, analyse and act upon data sourced from multiple local authorities and public sector bodies to reform public services.
Our hypothesis is that the scale of the financial challenges facing the public sector is such that it’s not sufficient for individual organisations merely to become as efficient as they can. Rather, they need to find ways to collaborate to address issues at a larger scale. The aim of ODAs is to marshal data to that end. Specifically, ODAs can:
Enable smarter decisions on scaling shared services. Public sector data often resembles a jigsaw: everyone has their piece of the puzzle, but no one can see the whole picture. The ODA methodology pieces together disparate datasets so it’s possible to see how problems, opportunities and demand transcend geographic boundaries and intelligently design shared services to respond.
Target resources at cases of greatest need / risk / importance. By overlaying different datasets on a map, or creating an algorithm to prioritise cases, data can help focus efforts where they are needed most. This contrasts with less effective, but common public service practice such as treating cases chronologically.
Predict future instances of a problem to enable prevention or early intervention. By modelling past cases and learning about the factors that correlate with higher risk, data can help spot future problems at an earlier stage when they are cheaper and simpler to resolve.
Of course, data analytics is no cure-all; it cannot address every issue. So how do you work out where it has something to contribute? Our work suggests answering that question requires clear thinking on nine principles under three headings: 1) the problem you wish to address; 2) the new action you want to enable; and 3) the data relating to that issue. For a public sector problem to be tackled with data, you need to:
Here’s what we mean by those 9 principles:
1 - Keep it simple. While public sector organisations are under immense pressure to tackle their most complex and challenging problems first, data analytics is novel to most organisations. Starting with the basics and getting some quick wins is the best route to sustainably build the support, skills and momentum needed for more ambitious data initiatives to succeed further down the line.
2 - Articulate a specific, actionable problem. For data to provide useful insight, broad problems need to be broken down into their smallest component parts. Like a mathematical equation, a data algorithm needs to be targeted at a very specific problem statement. It’s also important to focus on issues for which some corresponding action can clearly be envisaged. Data itself is not the solution: it is merely a route to enable a better intervention. If you can’t already conceive of interventions that would address the problem, no amount of data is going to help!
3 - Have a clear understanding of cause and effect. With some of the most complex social challenges that public sector bodies are tasked to address, there are either too many contributing factors to isolate the effect of one cause, or there’s a poorly understood link between cause and effect. Avoid these problems. You cannot train a data model if it’s not possible to say whether or not a different outcome was the direct result of your data-informed intervention.
4 - Articulate the specific, clearly defined action you want to enable. For data to be useful, you need to focus on achieving actionable insights. In other words, you must be able to clearly answer the question: “What would you do differently if you had better information?” This will run alongside the specific problem defined in principle 2 to create a statement along the following lines: “Our specific problem is X in response to which we want to Y”.
5 - Focus on actions within your control to change: Many public sector issues depend on, or are affected by, the actions of other organisations or indeed citizens themselves, over which you have little or no control. The ODA methodology is about finding data insights that enable you and your partner organisations to do something differently.
6 - Aim for interventions with short-term, measurable results. This requirement is not just designed to get a quick win. If you’re trying to create an algorithm that predicts something, the only way to confirm whether it works, and improve it, is to act on its insights and feed the results back into the model. In short, the feedback loop matters. This necessarily rules out issues where interventions will only deliver results years down the line.
7 - Know the Minimum Viable Information Product (MVIP) you need. Raw data is of little use to most people delivering a service. So what form does it need to take to enable the new action specified in principle 4? Do you need a map that shows the location of instances of a problem? Do you need a prioritised list showing highest risk / need / importance? Do you need a widget that enables unrelated datasets to be combined together?
8 - Ensure sufficient data is available to create the MVIP. Don’t worry about the quality of the data (“quality” means little until you try to use data to do something specific). But for any given problem there at least needs to be data available (i.e. you have it, can access it, or request it) to create the MVIP.
9 - Explore solutions that use non-personal data first. Sharing personal data between different public sector organisations is likely to entail a considerable amount of work designing and writing data sharing agreements, privacy impact assessments and security documents. If personal data is the only way to address your issue, then by all means take this path, but do so with your eyes open as to the work required. However, remember that it’s considerably faster and easier if a problem can be tackled with non-personal data. Think creatively and laterally about alternative data sources that might achieve your MVIP without requiring personal data.
Our pilots in London and the North East to test the ODA model are by their very nature experimental. The principles outlined above will no doubt evolve as we learn more through the process of testing them in the real world. In the meantime, if you have experiences, thoughts or suggestions that could help build upon them and make them better, I’d love to hear from you – please comment below or get in touch via Twitter.