I’ve previously written about plans by the GLA’s Andrew Collinge and Nesta to run a pilot for a London Office of Data Analytics, inspired by the Mayor’s Office of Data Analytics (MODA) in New York City.
Last week saw significant progress when 15 London boroughs came together for a workshop with the GLA to select a public service challenge that could be tackled with data. We’ll shortly publish details of the six shortlisted challenges discussed (which covered areas from health to waste management, and from housing to social care), and the conclusions we reached. The boroughs were asked to score each one according to the extent that it would be likely to: 1) save significant money, 2) have good data available, and 3) lead to actionable insights.
Deliberations were made much easier thanks to the presence of Mike Flowers – Chief Analytics Officer of Enigma (a leading US data analytics company), and the creator of the MODA model. Mike advised the boroughs to consider three additional factors when making their assessments.
Given the extreme pressure on public finances, local authorities are understandably tempted to tackle their most expensive and important problems first. After all, what good would an office of data analytics be if it didn’t address areas such as adult and child social care – the two biggest line items of local authority expenditure?
Mike’s advice: walk before you run. One significant challenge with social care issues is that they entail using a lot of personal data. Ensuring that all the right protections have been put in place, laws correctly adhered to and consents received can take a huge amount of time. At best, that delays the start of any data initiative. At worst, there’s a risk of inadvertently stumbling into another care.data, spooking people, and setting the whole data analytics agenda back by years.
Mike’s answer? Start with the low hanging fruit of potentially less financially urgent but still important issues that are less sensitive. Sort out dangerous buildings. Improve waste management. Focus on problems that involve data on things, places and events instead of people. These can be tackled much faster and provide quick wins that help build momentum and gain public and political support for further work. And if they can be done well, the more challenging issues can be dealt with down the line.
A second reason to avoid the complex social problems that public services address is that they may be the product of dozens of different factors. As a result, it’s often not clear what type of intervention would make any difference; the public sector may only have control over some of the factors; and the interventions may take a long time to have any effect. It could therefore be years before it’s clear whether a model has worked.
Childhood obesity is a good example. There are numerous factors contributing to poor childhood health – not even the research is 100% clear on which are the most important. Councils can control whether parks are open and well maintained, but they have little sway over what families eat or how much activity they do. And any interventions that they could put in place would be unlikely to show concrete results for many years.
Instead, Mike urged the boroughs to start with issues where there is an obvious cause and effect, and where data can help them do something differently tomorrow that will produce results in just weeks or months, not years.
He gave the example of how New Orleans’ Office of Performance and Accountability (the city’s version of MODA, led by Oliver Wise) used data to optimise the location of ambulances on standby. Previously, their resting locations were chosen based on dispatchers’ habits or their gut feel for where they could most easily get to emergencies. Based on data analytics work, they have been able to predict the locations of most likely need and position the ambulances closer in order to reduce response times. It works well because the factors are known and limited (time to emergency = distance x speed). The response is entirely within the control of the public service (ambulances can choose where to park). And the results are immediately identifiable (response times are reduced compared with the old model).
This case study also highlights another point that will be vital for any UK region wishing to trial data interventions: it involves almost no change in the activity of front line workers. In many US cities, there is often a single organisation looking after each issue (e.g. New York’s Department for Buildings). In London, by contrast, most issues are identified, addressed and reported 33 separate times (i.e by each borough). It is impractical to consider running change management programmes with all those teams. By simply optimising what front line workers already do, there’s a much stronger chance of data having an impact.
Once the boroughs had identified what they’d like to be able to do more effectively, Mike asked them to think about the “Minimum Viable Information Product” they would need. For any given issue it may be possible to identify numerous datasets that are relevant. But volume is not enough: can anything useful be done with them?
Would better interventions require a map that highlighted the density of particular problems, opportunities or demand? Or would it be an algorithm that predicted the location and timing of new instances of a particularly problem? Or perhaps a new widget that correlated multiple disparate datasets? If it’s not possible to even conceive what information product you’d need, an abundance of data is unlikely to be a enough.
Mike closed his presentation with one final message: data analytics is no cure-all. There are some problems where the honest answer is that more money is needed, not more data. The whole data agenda is done a disservice if we pretend it’s the answer to everything. Public sector bodies must therefore pick their battles wisely.
Thinking about each problem through the lens of Mike’s three criteria is a good place to start.
Find me on Twitter.
Image Credit: David Altabev