• by Gerard Broussard , Joel Rubinson , June 29, 2020

Ad targeting can be thought of as a tradeoff: As you increase the reach and frequency of advertising delivery to a targeted segment, you fund this hyper-targeting by shifting some media weight away from consumers outside the segment.

So the third-party segments you are activating better be worth it. But sadly, that is often not the case.

Why do third-party targetable segments sometimes fail to deliver a higher response to advertising?

When targeting fails to produce results, we have found two main causes:

1. Lack of Segment Accuracy. The consumers in a segment do not actually possess the purported attribute.

2. Lack of purchase forecasting science. The segment represents consumers whose future purchase likelihood of your brand is below the norm. 

What can a marketer or agency do to avoid these pitfalls? A lot.

Accuracy. We found that there is highly variable accuracy across even “name brand” data aggregators.

For a particular ethnic segment, we evaluated five providers and found that the concentration of the desired characteristic varied from an index of 250 down to 75. One provider was below the incidence of this ethnic group in the general population.

This difference in accuracy can be translated into hard CPMs per intended target consumer -- you might find your effective target CPMs could range from $10-$40, depending on the third-party party vendor choice.

Recommendation: Determine which segment is most accurate at the broadest scale before committing marketing funds, using available survey research protocols.

Purchase Forecasting Science. Targeting should be accompanied by a science of prediction. You should be forecasting, based on intention status and higher brand interest, that spending more money against a certain segment will produce a higher return.

The second criteria may surprise some who might think “if they already like my brand, why target them?”

Here is evidence that two factors for CPG brands studied -- a history of heaviness of brand buying and recency (being probabilistically close to an upcoming category purchase) -- were drivers of ROAS up to 16 times the average for the campaign. 

Recommendations: Measure if the consumers in the targets you have selected have above-average interest in your brand as part of the survey used for segment validity testing protocols, before committing to your choice of partner. 

What NOT to do -- conquesting gone wrong. An auto marketer, working closely with a broadcast network, wanted to maximize efficient reach against a segment of consumers who drive a particular competing brand of auto. A media plan was created by matching the consumer segment to a leading smart TV panel.

We predicted that the lift would be practically nonexistent because interest in the brand was below average. And once the campaign was run, that is exactly what happened.

What could have been done differently? If the marketer had also targeted consumers with at least moderate purchase interest in the marketer’s brand, better media choices would have been made, which we estimate could have resulted in at least 25% greater lift. 

When a segment is constructed to win, everyone wins. When it is unintentionally constructed to be low-performing, everyone loses.

The marketer wastes money, the media partner is told that advertising with them is not working, and the whole ad-tech stack is left explaining the failure.

Targeting tragedies can be avoided every time, simply by applying the principles of accuracy testing and forecasting science.