Mobile Applications

Alteryx platform blends data sets to predict US election result by zip code

I was so fascinated when I was first sent a link to the Alteryx’s election app I popped it straight over to a colleague in the US so she could have a play around.

The app blends static data, such as census information, with SurveyMonkey’s regularly updated election tracking stats. This means individuals and journalists can drill down into anticipated outcomes of the US presidential race by zip code and very specific voter demographics. The results are beautifully presented by Carto and Tableau Public on a map colour coded in red (Trump) and blue (Clinton).

The national predication, at the time of writing, stands at 42.7% for Clinton and 38.9% for Trump which is pretty much in line with other sources. But what is scary – if you’re not a fan of Mr Trump – is just how much of the central land mass is painted red with only a few, albeit densely populated areas, shown in blue.

This app is a great idea and builds on the work Alteryx did in 2012 – a detailed overview on the mostly successes and few failures from back then can be found here. The real trouble is the data isn’t live and lags some way behind SurveyMonkey’s regular releases.

Although SurveyMonkey released its last update last Thursday to factor in any changes since the second presidential debate, Alteryx will not revise it app to include any resulting change till 24th October. This is a long time in politics.

However, this aside, Matt Madden, Director of Product Marketing at Alteryx explains to us over the phone that the big aim of the app is to allow anyone, not just US citizens, to easily see sentiment at a glance. It certainly delivers this while what really makes it unique is it is offering predictions at truly local level. It is mostly used by journalists covering specific regions and of course banter amongst friends, he says.

But how reliable is it? As Madden points out: “The accuracy comes down to the data”. And this in itself is a very loaded point because many political predictions have got it incredibly wrong in recent years.

Interestingly SurveyMonkey was one of the few companies to accurately predict the UK general election in 2015 but like pretty much everyone – even the bookies – it still got it wrong about Brexit. The company’s Head of Election Polling Mark Blumenthal tells us over the phone that the really tricky part in all this is to accurately predict voter turnout. This had few previous indicators for the UK referendum, but is much more stable for any general election. 

In fact, SurveyMonkey – a platform that boasts three million individuals a day – offers quite a unique approach to election data gathering. In this US race it is asking 25 to 30 questions (depending if custom questions are added) about the election to people who have completed other surveys on the site. And the data volumes are far in excess of regular polling.

At the start of the year it was polling 10,000 individuals a week. Now as the pace is ramping up to November 8th it is surveying around 28,000 a week. It is slightly easier to reach better educated Americans, concedes Blumenthal but individuals who take part are “not disproportionately urban,” and are “remarkably diverse”, he tells us.


« Singapore wants to be Asia's fintech capital


Huawei: Broadband needs to adapt for 4K, VR, and IoT »


Do you think your smartphone is making you a workaholic?