Huge resources have been poured into “modeling” voters based on compiled consumer data. The problem is that none of those predictive models comes close to beating the latest generation of programmatic ad buying algorithms.

These algorithms are used by Facebook, Google and other online behemoths to sell ad inventory in real time, which is another way of saying it’s how the rest of us buy ad inventory in real time. It is a moment-by-moment auction to see which ads will appear in an individual Facebook user’s news feed, for example. As individuals, we don’t all see the same ads, because advertisers value individual prospective customers (or voters) differently and are willing to bid up the price to place an ad in one news feed, while ignoring another.

That’s the “real” Big Data, and for most online ad buyers, it’s making that endlessly re-modelled compiled data out there seem almost quaint by comparison. When we then import voter file data into the Facebook equation, we suddenly find ourselves with a product far exceeding Big Data’s most grandiose promises. We don’t just talk to users who “look just like our targets”, we talk to our targets.

We can schedule ads to reach only likely voters in a Democratic Primary who vote by mail, for example, while targeting Independents who vote on Election Day, swing 45+ Republican married women, etc. with a different series of ads. Even the phrase “likely voter” is now more science than guesswork. Sophisticated propensity index models use a variety of data points in addition to simple vote history in order to rank voters against increasingly accurate scales.

Here’s the part where it gets dangerous. Agencies that understand this dynamic bid up the price to make sure their candidates’ ads make it into the news feeds of their precisely-targeted voters. As a direct consequence, other candidates’ ads get shuffled into the news feeds of voters not in the target. In other words, a failure to understand how programmatic ad buying works means they are, in effect, targeting non-voters with their ad dollars.

When programmatic ad buying software comes to broadcast and cable – and it surely will – the same auction dynamics will vastly improve the efficiency of ad expenditures. We’ll pay for how many people – and who – are watching a show in real time, instead of estimated audiences based on what in today’s world is ancient history, aka the last quarter.