On the surface, it looks like the digital publishing industry should be extremely happy. After all, they’re the beneficiaries of an overwhelmingly successful automatic — or programmatic — marketplace for buying and selling ads; a marketplace which reached $22B in 2016. But these same publishers, brands and agencies are suffering from an unintended result — it’s a classic case of the tragedy of the commons and a major downside for publishers and users: a full one-third of digital display ads that make all this content possible are “bad,” meaning, they are too big, too slow or poorly constructed and can cause desktop or mobile sites to slow to a crawl or jump around as ads load. The result? Consumers get frustrated waiting for sites to load (inspiring many of them to install ad blocking software). Brands whose messages are delivered in badly constructed ad units can suffer, and publishers can see page views, revenues and consumer loyalty plummet.
When you think of “ad quality” issues, you may first think of bad creative like those “belly-fat” ads. As a publisher, you may think: “malvertising” or those annoying mobile app redirect ads that are so difficult to quash. Much has been reported on these types of “bad ads.” In fact, earlier this year, Google announced that they took down 1.7 billion ads in 2016 because they violated the company’s creative content policies, contained malware or were self-clicking mobile “redirect” ads. Google’s efforts and numbers illustrate how serious and rampant these types of bad ads have become.
And my company, Ad Lightning, released a report recently (available here) which found that nearly 30% of all display ads created “severe” issues that lead to site performance problems. So as bad as the belly-fat or “malvertising” issues are, the cumulative effect that lots of seemingly innocuous ads are probably having on the user experience and publisher ROI is even worse for publishers, brands and audiences.
The Effect of Bad Ads on Publisher ROI
The vast majority of bad ads that occur across popular websites are saddled with a different set of issues: they’re often more difficult to detect, address and quantify in production. Bad ads can severely increase page load times and create other ugly experiences for site visitors. They appear on the surface to be normal ad units and they’re often ads for trusted brands. But, the ways in which they’re designed, formatted and delivered negatively affect publisher brands, audiences and revenues. They can be boiled down to five types of ads that, for publishers, can create real ROI problems:
- Oversized Ads: Ads that exceed industry standard file size limits and take too long to load
- Over-Requested Ads: Ad requests made to networks too many times
- Processor-Intensive Ads: Ads that use too much CPU processing time and power
- SSL Non-Compliance: Ads that are not HTTPS encrypted.
- Intrusive Ad Formats: Ads delivered in unsupported formats
New Ad Quality Study Reveals Scope of Issues
At the end of 2016, my company, Ad Lightning, set out to gain an understanding of the magnitude of bad ads across the industry. Among the key findings that surfaced: 28% of all of the display ads had at least one of these performance issues rated “severe” and many more exceeded recommended industry guidelines. We also confirmed something we had long suspected: that ad-supported web pages generally take twice as long to load as the same pages without ads. Throw in some significant ad quality issues and the ad-supported pages can take seconds longer (and thus, create negative experiences that can drive audiences away). And we saw a widening gulf between accepted industry standards and what’s really going on.
The Gap is Wide Between Industry Guidelines and Reality
Recent studies have found that each second of page delay erodes user engagement and satisfaction significantly. The IAB has established its LEAN ad quality guidelines to help prevent that erosion. However, across the board, the Ad Lightning study revealed sizable gaps between recommended industry practices and actual industry practices. Here’s a glimpse of some of the issues we tracked and the differences between the industry guidelines and reality.
Over-sized Ads: IAB guideline: The maximum file size for display ads is 300KB. The Ad Lightning Ad Quality Study: 41% of the ads tracked exceed that size limit. Nearly 10% of the ads were larger than 5MB.
Over-Requested Ads and Excessive Tracking Scripts: IAB guideline: a max of 15 network calls for any given ad unit. The Ad Quality Study: The average was 56 network requests per ad.
Processor-Intensive Ads: Ads that take more than 300 milliseconds of CPU processing time are generally considered unacceptable. The Ad Quality Study: About one third (32%) of the ads were overly processor-intensive (took longer than 300MS). Roughly 12% of ads took more than 500MS.
SSL Non-Compliance: IAB recommendation: The digital ad industry should adopt HTTPS (HTTP/2) encryption across the board. The Ad Quality Study: On average, for each ad, about 51% of network calls were not SSL compliant.
We have a long way to go.
The presentation of display advertising plays a critical role in the way users perceive and experience ad-supported websites. When publishers sell inventory programmatically, the process demands that they cede control of those ad units (and the way site visitors experience the ads) to an array of programmatic partners. While it may be the publisher’s responsibility to track and manage these issues when they occur, the responsibility to reduce the high rate of occurrences falls on just about everyone in the digital advertising ecosystem. Everyone has a stake, too. The cumulative impacts of these issues reach well beyond publishers. When ad blockers go up and page views go down, the addressable audience wanes. And that affects the reach and success of even the most pristine ads and campaigns.
Fortunately, the Coalition for Better Ads is working with the IAB to set new standards and establish formal rules for measuring and enforcing them. We strongly support their efforts on behalf of the industry.