In February our CTO, Mark Jablonowski, was invited to speak on a panel at RampUp2017 in San Francisco. As a part of his part of his attendance, Mark was asked to blog about how political marketing uses data, and how that compares to corporate marketing.

Political advertising is pretty fascinating stuff, and if it’s possible, even more complicated than corporate digital marketing.

The reason for that is pretty simple: We’re looking to get to at least 50% market share by the time the polls close on Election Day. And, since voting registration and vote histories are a matter of public record, we approach most of our advertising campaigns with a pretty finite list of people who could possibly vote for our clients. That’s not always the case (Hi, President Trump!), but for the most part, our job is to find the specific people out of the total universe of likely voters that our clients need to either persuade to support their candidate and ensure they actually get out and vote.

We’re basically brand marketers who target their advertising like direct marketers.

When it comes to data, we also have some unique challenges. Again, generally, our clients will have a pretty good idea of what subsets of voters they need to motivate, and what message each group of voters responds to best. So we have to match their first-party data to online ids deterministically to make sure we get the right messages to just the right people, at enough scale to make sure our clients win.

Third-party data is crucial to what we do, but it’s a little upstream from where it would typically be used by corporate marketers. I’m sure you’ve heard of “micro-targeting” in politics, and that’s where third-party data comes in handy. Our clients will have tested their messages in polls, and used those results — along with third-party data — to help score the rest of the registered or likely voters to find out exactly which individuals we should be speaking to.

Political advertisers are also unique in that we have a lot of interesting challenges in keeping that voter file updated. One of the more recent innovations in political data is the ability to create one big national dataset of voters and potential voters. Until the last decade or so, marketers in each state tended to have their own state-based file, so a lot of voters who move around may lose data, or appear to be infrequent voters. As a result, our targeting would miss a lot of important people. Now, we ensure that our data partners are using complete national files, with long voter histories, and have appended high-quality additional data points from third-party providers to add richness to our predictive models.

If we had a target universe as big as most commercial marketers, we’d probably run out of budget before we ran out of qualified targets. However, as partisanship and polarization has grown, there aren’t always a large number of persuadable voters available in any given state or district. And when you combine that with our deterministic targeting, you can run into scale issues pretty quickly. To help ameliorate that problem, we work with multiple onboarding partners to capture as much additional incremental reach as possible. Again, if we were just selling CPGs, a 30% match rate would be plenty — but try to tell a candidate you’re only going to advertise to 20-30% of their potential persuadable audience, and they’re not going to be thrilled.

In any case, data quality is an extremely important issue in political marketing so we take a lot of care in selecting our data partners. Inaccurate advertising data for a corporate marketer can mean a drop in sales and a bad quarter. A candidate with bad data loses — and the campaign is over. For good.