According to a Nielsen study, reach and targeting combined contribute with 31% of sales generated by advertising. The understanding of targeting and frequency across media, browsers and devices are highly important advertising elements to enable incremental sales gains.
Advertisers, agencies and media companies invest heavily in technology, which in theory should facilitate the optimal campaign efficiency against advertiser objectives. But as it is often the case, there is a gap between theory and practice.
Despite our collective human efforts combined with the wonders of technology, algorithms and data, too often campaign exposure is not accurately reaching the intended target group. In fact, according to AudienceProject who has taken a look at 13.000 audience-targeted digital campaigns from 2018 across 6 markets, the industry average accuracy is around 45%.
Now, accuracy should be seen as relative to the incident rate of the target group in the online population, and therefore in many cases, 45% could be a satisfying result. E.g. if you are trying to reach 18-24-year-olds, which make up 15% of the online population. In that case, you would have achieved an affinity of 300. Well done!
However, taking the incident rate into perspective, the average affinity (index) in our industry, ranges around 115. The study does not take into account if media buyers were using contextual targeting, media affinities, re-targeting or any other approach to target the desired audience.
But in a world where 90% of all data has been created in the past 2 years (that is worth re-reading!) and we have access to an infinity of tools, what are the reasons for us not being able to do a better job of targeting the right consumers? There is no single answer to this, but here are the top five reasons:
1. We under-measure. It sounds bizarre that digital is under-measured. But when it comes to the audience, it is. Our best estimate is that under 10% of digital campaigns have the audience properly measured (reach, frequency, accuracy, demographics, affinity achieved etc.). The consequence of not measuring the audience on campaigns is that we navigate without a map and a compass. How can you know what to change if you don’t know how you are doing? Just as we measure fraud, viewability, impressions and clicks, audience measurement should be “license to operate” when buying and selling media.
2. Data segments are used incorrectly in combination. Using multiple segments to target a campaign often leads to unexpected results. Let us imagine you are using a female segment where the accuracy of the segment is 80%, and you overlay that with a sports segment. The underlying gender split of the sports segment could very likely be leaning towards men, e.g. 75%. In that scenario, it must be expected that the male part of the female segment is overrepresented in the overlap between the two segments, and therefore you end up having a much lower accuracy against females than 80%, which you might intuitively think.
3. The same dataset behaves differently when applied to specific media. This use case is similar to #2. Let us take the same female segment again and imagine you apply that to the auto section of a news publisher. As the auto section is over-represented by males, the female segment will under-deliver in this context, as the ad server will serve ads to the IDs in the female segment who are actually males.
4. The same dataset behaves differently when applied to specific devices/OS. The same logic applies as to #2 and #3 above, if you are applying e.g. an age segment of 50+ on iOS devices, probably your accuracy will suffer, as the iOS audience is under-representing the 50+ audience compared to the online population distribution.
5. Targeting data is unvalidated. Where the incorrect application of targeting data makes the main explanation for why digital campaigns are not effectively reaching the desired target audience, we also have to take a close look at the quality of the data itself, before it is applied to media. The prediction of UIDs into different segment classifications is always done on the back of a set “confidence score” based on the behavioural log lines connected to the specific ID. It indicates how certain the prediction engine is that the UID does, in fact, match a particular segment. The lower the “confidence score”, the lower the accuracy of a segment, but the bigger the volume of the segment. The accuracy of a segment is always achieved as a trade-off to reach. The physical consequence of this is that data segments come in very different levels of quality. So even if you are buying a female segment, the accuracy of the segment can vary a lot. To be on the safe side, shop validated audiences.
I choose to see 45% hit-rate in target group across our industry as a massive opportunity to increase advertiser efficiency in driving incremental reach for their investments. Regardless if you are an advertiser, agency or media company, the first step is to audit verify the audience efficiency of your own in-house day to day operations. The second step is to secure that you work with the right data vendors and that people handling data understand the limitations just as well as the opportunities. The third step is to consistently measure the audience on campaigns to document improvement and efficiency.