A new whitepaper from ComScore claims that fraud and low viewability don’t just cost advertisers money – they’ve had a major impact on campaign measurement, which has led advertisers’ media planning models to consistently undervalue digital advertising.
Much ink has been spilled about the advertising investment gap in digital and mobile, where advertisers have been reluctant to push the majority of their spend. The report’s authors argue that a big factor in lagging digital investment is that until recently, campaign analysis tools have not been good enough to capture indicators like non-human traffic and viewability. This poor-quality inventory has dragged down measurements of the effectiveness of digital advertising.
The report quotes Aaron Fetters, director of digital insights at the Kellogg Company. “When you consider that each impression’s impact has generally been diluted by fraudulent, non-viewable, and out-of-target ads, the question is how much better can these results get when we have cleaner inputs?” he says. “We’ve seen that a 20% increase in viewability can produce a 60% effectiveness gain in our mix models, suggesting there’s still plenty of ROI upside in digital.”
ComScore offers a case study on CPG ConAgra Foods to support the claim that once low-quality inventory is identified and excluded through optimization, the remaining ads are significantly more effective. ComScore measured ConAgra’s ROI from digital advertising both before and after it implemented viewability optimization to screen own non-viewable inventory. The result was a 74% increase in sales volume per impression and a 39% increase in sales volume per advertising dollar spent.
The report stresses that viewability and fraud are interrelated issues, and can often reinforce one another. In another case study, ComScore compared two campaign measurement providers that worked on the same campaign. Provider B, which looked at only viewability, found the overall campaign scored an 80% viewability rating. But provider A, which first identified fraudulent inventory and excluded it as non-viewable, found that the overall campaign’s viewability rating was only 55% even though the viewability rate among the verified human-served ads was still 80%. (This may well be one reason that there is such a large disparity between viewability providers’ measurements.)
Using provider B’s measurements, an ROI model would be forced to blame the ineffectiveness of the campaign on the digital channel, rather than the 4 million fraudulent impressions identified by provider A.
Fraud disproportionately affects a few campaigns
The good news is most advertisers appear to be implementing effective anti-fraud measures. A study ComScore performed in November in the U.S. found that among measured campaigns, only 21% of them accounted for lion’s share (75%) of non-human traffic, leaving the vast majority of studied campaigns seeing less than 5% non-human traffic.
On the publisher side, results were also promising: 50% of publishers had less than 1% non-human traffic, and 80% had less than 3% non-human traffic.
ComScore reiterated findings that open exchanges and networks pose the greatest risk of fraud, because inventory sold on open exchanges is not as thoroughly vetted as publisher-direct inventory. Although the paper didn’t include a benchmark, a study by Integral Ad Science found that 14.5% of inventory on U.S. exchanges and networks is fraudulent, compared to 3.3% direct bought. However, it’s possible that exchanges and networks see the same long-tail effect that publishers and campaigns do: a few large-scale exchanges account for the majority of fraud, while others are relatively safe.
