Micro Targeting vs. Macro Targeting
We all know the power of micro-targeting in campaigns. Harnessing the power of voter files with hundreds of millions of voters and hundreds of bits of information allows us to reach more voters with more effective messaging at a lower cost. And yet, we also all saw the limitations of micro-targeting in spectacular fashion in the 2016 presidential cycle. It’s clear at this point that straight micro-targeting operations are missing a big part of what is going on in a political campaign. It seems political data’s best-practices need some revision.
It is notoriously difficult to capture low-propensity to vote voters and, while there are methods to adjust for this, micro-targeting approaches are challenged to account for circumstances without an historic analog. Also, micro-targeting has difficulty evaluating the effectiveness of ‘macro-targeting’ platforms, like broadcast TV and radio. When there is such a big difference in earned media coverage (such as in the 2016 presidential cycle), tracking earned media is essential.
Some straightforward changes might include paying more attention to earned media, low-propensity to vote voters and the power of automated creative generation on such platforms as Facebook. In addition, there’s also a very real danger of an over-correction away from micro-targeting techniques, so as we head into the 2018 and 2020 political cycles, it is probably worth taking a step back and considering just how to integrate these elements.
At the same time the advent of micro-targeting on social platforms like Facebook have shown impressive results. These efforts will continue. Indeed, the new thing from the 2016 cycle wasn’t the use of micro-targeting on these platforms, but generating, testing and scaling micro-creatives in real-time.
But how to apply what works in micro-targeting to macro-targeting platforms? How do we translate the mountains of spend that go into broadcast TV to the vote gain on election day? The solution we came up with at Scripps is agent-based modeling, and after testing this approach over 14 races from the 2016 and 2017 cycles, we’ve finally launched it publicly as MarketPredict.
One thing we knew we had to account for was the major differences in data scale, quality and format across the political and media data landscapes. Agent-based modeling allows a great deal of data flexibility, and we’ve managed to adapt and integrate data from a myriad of sources from polling to media spend to web search and more. This data flexibility means we see the whole political campaign landscape, strengthening the reliability of our recommendations.
Again the 2016 election cycle was one of the most volatile we’ve ever seen. To adapt, we needed a platform that could quickly determine what matters when things change. Agent-based modeling offers such a solution and we could effectively simulate the impact of major events for our clients in near-real time. This was mostly about showing the impact of major events (i.e. hurricanes or terrorist strikes), but we also worked with campaigns to test out different creative messaging mixes.
Another priority we needed in our platform was a way to tell a story. Scripps is a news company and we thrive on our storytelling. MarketPredict allows us to tell the story of every election and race we enter. This gives us an unprecedented ability to determine and evaluate the effectiveness of campaign narrative, rather than just the impact messaging will have on a given individual or group of voters. We can say how important social and earned media are relative to paid media, what paid media is effective, what impact a creative spot will have, even how many voters will change their voting intentions from an ad. This isn’t like the macro tools of yesteryear, showing how many voters are hit with an ad. MarketPredict can say how many voters seeing an ad will actually change their behavior, all in near-real time.
What will 2018 and 2020 bring? One can only assume more micro-targeting and more ads on Facebook, but smart campaigns will be looking at data and agent-based modeling tools like MarketPredict to augment their data toolkits. The best campaigns will use both effectively to gain an electoral edge.