The Alleged $7.5 Billion Fraud in Online Advertising
Halibutboy Flatface stashed this in Fail
Internet advertising as it exists now sucks, but surprisingly few seem all that interested in improving it.
Wow, the "impression" model is totally broken:
An "impression" occurs whenever one machine (an ad network) answers a request from another machine (a browser). (For reference, you can see my definition and example of a "request" in a prior Moz essay on log analytics and technical SEO.) Just in case it's not obvious:Human beings and human eyeballs have nothing to do with it. If your advertising data states than a display ad campaign had 500,000 impressions, then that means that the ad network served a browser 500,000 times—and nothing more. Digital marketers may tell their bosses and clients that "impression" is jargon for one person seeing an advertisement one time, but that statement is not accurate.
Just because a server answers a browser request for an advertisement does not mean that the person using the browser will see it. According to Reid Tatoris at MediaPost, there are three things that get in the way:
- Broken Ads—This is a server not loading an ad or loading the wrong one by mistake. Tatoris writes that these mistakes occur roughly 15% of the time.
- Bot Traffic—Whenever hackers write these automated computer programs to visit websites and post spam or create fake accounts, each visit is a pageview that results an an ad impression. According to a December 2013 report in The Atlantic, 60% of Internet traffic consists of bots.
- Alleged Fraud—In Tatoris' words, "People will hide ads behind other ads, spoof their domain to trick ad networks into serving higher-paying ads on their site, and purposefully send bots to a site to drive up impressions." Noam Schwartz described in TechCrunch two additional methods of alleged fraud: compressing ads into a tiny one-by-one pixels that are impossible to see and using malware to send people to websites they never planned to visit and thereby generate ad impressions. AdWeek found in October 2013 that 25% of online ad impressions are allegedly fraudulent.
Tatoris crunches all the numbers:
We start with the notion that only 15% of impressions ever have the possibility to be seen by a real person. Then, factor in that 54% of ads are not viewable (and we already discussed how flawed that metric is), and you're left with only 8% of impressions that have the opportunity to be seen by a real person. Let me clarify: That does not mean that 8% of impressions are seen. That means only 8% have the chance to be seen. That's an unbelievable amount of waste in an industry where metrics are a major selling point.
Essentially: If you have an online display ad budget of $100,000, then only $8,000 of that ad spend has the chance to put advertisements in front of human eyeballs. (And that's not even taking into account the poor clickthrough rates of display ads when people do see them.)
If you are paying $0.10 per impression, then the $10,000 that you will pay for 100,000 impressions will result in only 8,000 human views—meaning that the effective CPI will actually be $1.25.
Essentially, here is what is allegedly happening:
- Clients give money to agencies to purchase online display advertising
- The agencies give the money to the ad networks
- The ad networks give a portion of the money back to the agencies
- The clients' display ads are only 8% viewable
- The 92% non-viewable impressions still earn money for publishers and ad networks
I think we can see who the loser is—everyone is making money except for the clients.
Google, to its credit, has disclosed that 56% of its digital ad impressions are never actually seen—of course, the report was also released with the announcement of a new ad-viewability product.
The author Samuel Scott's recommendations for clients:
- Stop doing cost-per-impression (CPI or CPM) campaigns. Traditional digital advertising strategy recommends that people use CPM campaigns for brand awareness, cost-per-click (CPC) campaigns for traffic, and cost-per-action (CPA) campaigns for sales and conversions. In light of this scandal, I see no good reason to do CPM at all anymore.
- Revise advertising KPIs and metrics in human terms. Earlier in this article, I calculated the following change to a hypothetical CPI value: "If you are paying $0.10 per impression, then the $10,000 that you will pay for 100,000 impressions will result in only 8,000 human views—meaning that the effective CPI rate will actually be $1.25." In addition, half of all clicks in CPC campaigns might also be bots. As a result, a $2 CPC result may actually be $4 when reaching human beings is taken into account. Ad campaign analysts may want to take alleged bot and fraudulent activity into account when calculating ROI and whether display advertising is worthwhile.
- Demand full disclosure. Clients should ask agencies and media buyers if they are getting paid directly or indirectly by ad networks. Agencies and media buyers should ad networks how they are combating bot activity and any fraudulent behavior. Ad networks should not turn a digital blind-eye to publishers who intentionally use bots to make profits off of advertisers. If anyone gives vague answers or otherwise disparages such questions, then that is a red flag. Advertisers should demand and receive full, verifiable information in light of what has allegedly been occurring.
- Block certain countries from campaigns. According to a report in Ad Week, China, Venezuela, Ukraine, and Singapore have "suspicious traffic" rates of between 86% and 92%. (The rate in the United States is 43%.)
- Use ad-fraud detection platforms. Companies such as Forensiq, SimilarWeb, Spider.io (which was bought by Google), Telemetry, and White Ops compare visit patterns with industry benchmark behavior as well as check for malicious software proxy unmasking, verify devices, and detect any manipulation.
- Run manual campaigns as much as possible. The only way to reduce wasted impressions significantly is to research and implement digital ad campaigns manually rather than use programmatic ad buying. Digital advertisers should research potential websites on which they want to run advertisements to see if they are legitimate—potentially even running ads on only the largest, well-known sites but doing so continuously. This way, it might be best to focus your ad campaigns on quality viewers rather than trying to maximize the quantity of viewers by also including lesser-known sites.