Whether at the ballot box or the supermarket aisle, campaign success requires proper management and integration of data
Over the last few months, dozens of British parliamentarians have been quietly sacked. For the moment, they still have MP after their name and they still have their salary payments coming through. However, party bosses – with varying degrees of regret or barely suppressed pleasure – have donned the mantle of The Apprentice’s Sir Alan Sugar and said: “You’re fired”.
The firing has been triggered not by policy differences but by a hard-headed decision by party bosses to cut loose some MPs in order to concentrate limited party campaigning resources on their colleagues with better chances of winning.
Big data is fuelling these decisions. It comes from many sources, mixing together market research (especially constituency opinion polls) with on the ground performance metrics (such the number of voters canvassed in the last month) and feedback (such as the number of supporters identified).
Data normally sits in organisations in a multitude of silos, splintered into smaller pieces across databases, spreadsheets, online services and email inboxes
As polling day nears, this use of data to fire MPs will move into a real-time phase. With time from telephone canvassing operations along with money for geo-targeted online adverts moved around rapidly the campaign can respond to the latest data showing who else should be abandoned – and maybe even – in the case of a lucky few – who has earned a reprieve.
Data is not only at the heart of strategic targeting decisions but also will be at the centre of the campaign delivery. Data helps identify which individual voters to reach, through which channel and with which message within each of those constituencies that make the selection for resources.
The reality of ground campaigning can be a little more prosaic and old-fashioned than the technological sheen makes it appear. The recent case of Conservative MP David Burrowes shows this. He stumbled with his canvassing team into calling on voters in the wrong constituency – even canvassing the wife of a neighbouring MP by mistake. The grand plans of campaign organisers about what to say to who on the doorstep are often undone by nervous volunteers picking their own words and topics (as documented expertly in the book Ground Wars: Personalized Communications in Political Campaigns).
The biggest obstacle, however, is missing data. Data-powered targeting of voters requires data about voters and a trio of obstacles get in the way.
First, Britain’s relatively tight data protection laws (especially when compared to the US) make acquiring data about voters from other sources harder and more expensive.
Second, only around 4 in 10 UK voters are in the public phone book. This impedes data gathering via phone calls, and the hit rate of doorstep canvassing at finding someone in and willing to talk is also low, getting down to 1 in 10 in urban areas on Sunday mornings.
Third, voters move – in urban areas particularly – that can mean up to a third of voters moving each year. This makes keeping track of people hard and resulting in accumulated data often being lost when someone moves and remains untraced.
As a result, all parties suffer from large gaps in their datasets which is where modelling (often called voter scoring) comes in. It extrapolates from voters for whom they do have data to make educated guesses via statistical analysis about those for whom they do not have data.
Inadequate data is better than none at all
The result is some campaigning that is amusingly misdirected, such as the Liberal Democrat MP who regularly received expensive direct mail from the Conservative Party. Perfection is not required, to be useful is good enough, and the NHS use such techniques in life and death cases as a means to decide how to target preventative health work. As computer science pioneer Charles Babbage put it: “Errors using inadequate data are much less than those using no data at all.”
The best campaigns, moreover, minimise such mistakes by integrating their different data into one, coherent system – so the local knowledge about where the opposition MP lives informs the central direct mail operation too.
However, the best campaigns are rare – and there is a lesson here too for those harnessing data outside politics. Data is at its best when big and integrated, yet it normally sits in organisations in a multitude of silos, splintered into smaller pieces across databases, spreadsheets, online services and email inboxes.
Integrating data, extrapolating to fill in the gaps and then deploying it both to make strategic decisions on prioritisation and to enrich real-time campaigning is the route to both political and non-political success.