Big Data

How data quality affects the success or failure of a political campaign

Paul Westcott, vice president of data at L2, explains why data quality is everything when it comes to reaching voters and winning elections.

This is the final interview in a series of four videos with Paul Westcott, vice president of data at L2. The other videos may be found here:

Politics has always been a data game. Going back to the days of direct mail, Carl Rove mastered these strategies. As we emerged into the 21st century of course, digital transformation hit politics hard. In 2008, social media was the defining technology. In 2012 and 2016, data was the defining technology, but there has been a narrative that data failed in 2016. There's also a counter narrative that data is powering AI, bots, and other tools that will inevitably be used in the 2018 midterms and be optimized beyond.

TechRepublic met with Paul Westcott, vice president of data at L2, to discuss what lessons we can draw from the success and failures of 2016 and how big data is being used now, without the narrative that data failed. Below is a transcription of their interview.

Westcott: We heard a lot of that after the campaign and specifically from what happened with the Clinton campaign. People said they relied on analytics. They relied on these predictive models and at least that's what was being fed out publicly. Really at the end of the day, their models looked good. Their models were actually put together in a really solid way and the team there was among the best in the country. What failed was the underlying, a poor data set, that they had been provided.

I won't mention who, but these large data sets that get provided to campaigns are in some cases, especially when the team has such an analytics type focus or sometimes an afterthought, whereas pulsters look at it and say, "This is the core of what we're reaching out to when we're trying to get our 600 person sample or 1,000 person sample."

So the core data matters. The underlying data matters. That's really what failed because in the L2 national voter file, we have 185 million individuals. Some of the people we either compete against or the political parties or even the secretaries of state, meaning the core data providers, have 200 million, 220 million voters in their nationwide files. Not that Secretaries of State maintain that, but individually they add up to 220 plus million voters. So you say, "What's the difference there?" It's all about processing, cleaning, deduping, getting rid of garbage data. Data quality matters. So that underlying core data which failed in 2016 for a number of campaigns, that needs to be reevaluated and how they process that. While analytics and machine learning is all great, if you're building it on a bed of sand, it's not going to stand up.

Also see

Image: iStockSvetaZi

About Dan Patterson

Dan is a Senior Writer for TechRepublic. He covers cybersecurity and the intersection of technology, politics and government.

Editor's Picks

Free Newsletters, In your Inbox