Signify’s first ever project (before we incorporated the company) was an in-depth look at the snap election called by Teresa May for June 2017 and delivered to our client two weeks before the election. The project perfectly demonstrated the predictive power of machine learning applied to public data – it also told us exactly why the Conservative party performed so badly.
Our client (a Liberal Democrat strategist) wanted to know what voters cared about around the UK. We looked at 5 million relevant posts on social media by UK voters, taken from the previous week. The study took place at a very particular time, just days after the Ariane Grande concert bombing, so ‘Security’ was naturally at the top of mind for both politicians and voters. Just behind that – the NHS and Social Care.
For Signify there were two very exciting aspects of this work. Firstly, by looking at the popularity of candidates we accurately predicted where the Lib Dems would pick up seats from Labour around the North East. Secondly, our identification of key voter issues was borne out in the final days of the campaign and in the result. We told our client that this election was about social care and the ‘dementia tax’. They subsequently told us that this was the top issue brought up on doorsteps in the fortnight after they met us and before polling day. (Unfortunately, they had decided to focus on Brexit.) On the morning after the election political commentators as diverse as Alistair Campbell and Lord Cooper singled out social care policy as the key to the election result.
For the Signify founding team, still working in our day jobs, the election provided a great moment. It showed that our techniques worked as a predictive tool and as a way to highlight the issues that matter to voters – and it proved in a live environment that Signify could zero in on key issues at a local level. We’ve since provided this kind of insight for local candidates in the UK and the USA. But May 2017 was the moment when we knew – this is going to work.