Is third-party data targeting more effective than contextual targeting?
A new report from performance marketing agency Roast and ad platform Teads tested whether the costs of user data targeting and of complying with GDPR are worth it.
Can contextual ad targeting work as well as interest- or occupation-based targeting?
That question has become very relevant these days, in light of the consent requirements and other limitations surrounding personal and third-party data.
To help answer that question, two London-based firms — performance-based online agency Roast and ad platform Teads — decided to conduct a test, which is covered in a recently released white paper, “The Enduring Effectiveness of Contextual Targeting” [free, email required]. Roast’s head of mobile/display and paper co-author Lucy Cunningham told me that, to her knowledge, this is the first test of its kind.
The word “enduring” in the paper title relates to the fact that classical modern advertising, such as in the TV- and print-oriented days depicted in the “Mad Men” TV series, was fundamentally contextual. Advertisers bought ads on, say, sports events to reach men and soap operas to reach women.
This contrasts with the data-based approach of current digital marketing, where advertisers commonly show ads to site or app visitors whose cookie-based profiles indicate, say, they are women ages 18-34 on the West Coast. For many ads, the content in which the ad is shown is a way of attracting those kinds of users, but it often doesn’t govern which ad is shown.
But does one way work better than the other? If not, contextual-based ads are compliant with the General Data Protection Regulation (GDPR) because they don’t require personal data. They could be cheaper to manage because the advertiser wouldn’t have to buy third-party data, or ask and track consent permissions, or rely on lightning-fast but highly complex programmatic platforms to recognize the right kind of visitor and immediately serve that ad.
In fact, Pagefair’s GDPR-compliant Perimeter initiative is one effort that relies largely on contextual targeting.
The test’s hypothesis: “A carefully curated whitelist [of websites selected for content, ad density and performance] will perform as strongly as a media buy using a 3rd party data overlay.”
The test, conducted in April, was divided into two groups, each employing an ad offering Roast’s free Voice Search Report. Here’s the creative, shown as a static image:
One group was directed at 40 sites in the Teads network. The sites were selected from Teads’s total network of 400 by their scores on site performance (such as page loading speed, user experience metrics or SEO optimization, all of which indicated a “well-constructed and high-functioning web page”), by lighter ad density and by contextual relevance.
The latter meant that the content was either tech or marketing, on about 80 percent of those 40 sites. The remaining 20 percent offered non-tech or marketing content, like news or lifestyle, but they were included because they scored especially high on performance or density.
The other group was directed at the entire Teads network of 400 sites and targeted the same ad for a voice search white paper at users whose profile attributes, based on third-party data, indicated they might be employed in tech or marketing, or that they might otherwise have an interest. Third-party data included online purchase behavior, demographic, credit behavior, salary, household size, occupation and interest.
Millions of impressions were shown to the two groups over a month, employing the Coalition for Better Ads’ non-intrusive, inRead Scroller display format. It offers a large placement that allows the user to scroll past. Only websites seen by desktop users were included, so as to avoid other variables that mobile or tablet might introduce. Conversion to a sale was also not included as a factor, given the additional variables it offered.
The result, measured in click-through rates:
The Contextual approach returned a 0.236 percent CTR, while the Audience approach (third-party data) was 0.24 percent, a negligible 0.004 percent difference. There was also a marginal 0.005 percent difference in users who viewed the ad for 3 seconds. However, the Contextual approach did show a very slight 0.03 percent increase in Hover rate (that is, users whose cursor hovered over the ad) and a 0.03 percent increase in users who viewed the ad for 5 seconds.
From the report:
3rd party data [often] comes at a premium cost, whereas site curation using this methodology, does not. This test showed that data overlay and careful contextual curation provide comparable results. In a post-GDPR world where there may be a scarcity of data, and potentially a price increase, this is good news for advertisers looking for engagement activity…
As for the lifestyle sites, Cunningham acknowledged that kind of content didn’t work as well as tech, marketing or news for this ad, but the stats for lifestyle sites were not broken out from the overall Contextual figures.
She also noted the many limitations of this test. For instance, it tested only one kind of ad, for one kind of product, and only on desktop websites.
But this effort indicates that, at least in these circumstances, all of the cost and hassle of dealing with data that targets users by interest, demographics, occupation or other characteristics may not be significantly more effective than, in effect, buying viewers of a ball game.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.
New on MarTech