Ask a Centro Expert is a new blog series from Centro where we break down the complicated tools, tech, and trends you’ve been hearing about in the trade pubs and around the office. We reached out to some of our in-house experts to sort through the tough questions and turn them into bite-sized, palatable Q&As for your reading pleasure.
This week’s topic: Ad fraud. We talked to Ian Trider, Centro’s Director of RTB Platform Operations, for the break down.
1. In the simplest terms, how would you describe ad fraud?
Ad fraud is the practice of serving digital ads that a) have no chance of being viewed by a human user or b) are misrepresented.
2. A lot of the time ad fraud is described as non-human or robotic “bot” traffic. That can mean a lot of different things. Can you give examples?
Googlebot is an example of a legitimate bot. It is Google’s “crawler” or “spider” – the software that automatically visits websites to build Google’s search engine. Legitimate bots identify themselves, and discounting that traffic is a standard requirement in industry guidelines for impression and click counting.
Illegitimate bots attempt to simulate human web surfing for the purpose of generating paid ad impressions.
It is important to understand, though, that there are types of traffic fraud that are technically served to humans, but the traffic is still unacceptable. Examples include: impressions generated when web sites are loaded as pop-unders on porn and piracy sites, or domain spoofing, where impressions claim to be on a respectable, known site, but are actually served somewhere else. This “human traffic fraud” is a significant problem too, especially in video. Bots are a part of the problem, but not the entire problem.
3. Does bad inventory constitute as ad fraud? Does ad stacking count as ad fraud?
Bad needs further definition. Inventory that is bad because it doesn’t achieve the performance goal of a campaign is not fraud. Inventory that is bad because it has undesirable characteristics (fake news, adult content, etc.) is not fraud.
Ad stacking (placing multiple ads on top of each other) is fraud, because the hidden ads are impossible to view. An important distinction to note here: Unlikely to be viewed and impossible to view are not the same thing. A placement may have very poor viewability, such as a 728×90 banner at the very bottom of a news site, but it is possible to view it – most users just won’t scroll that far. A stacked ad cannot be viewed under any circumstances.
4. So, there’s a connection between viewability and ad fraud?
Yes, but not the one that might be expected. Fraudsters want fake impressions to look desirable, and making them appear viewable is one such way. There is actually a positive correlation between gross viewability and fraud: Highly viewable inventory is more likely to be fraudulent.
This is why viewability data should always be reported net of fraud – but even so, fraud detection is an inherently imperfect science. While it should not be assumed that high viewability is a sign of fraud, extremely high viewability on unfamiliar sites that seem unlikely to have the traffic they have is certainly a sign to be cautious of.
5. Who creates the bot traffic?
It depends. Websites are sometimes set up containing fake or low-quality content, and ads are placed on the sites with the intention of sending bot traffic there. The creators of those sites undoubtedly know exactly what they’re doing.
In other circumstances, especially when significant volumes of bot traffic appear on what appears to be a legitimate site, the publisher may have engaged in traffic buying (increasing the traffic to their website by paying other companies to bring visitors to the site). This is a fairly common practice, and there are ways to do it that are quite likely to result in real, human users (i.e. content ads via Outbrain, Taboola). However, these sources are relatively expensive. When the goal is to make more from advertising than it costs to bring the user to the site, publishers find themselves pursuing cheaper and more questionable sources of traffic. Breaking even requires visits that cost more than a few pennies each – and at this price, there is a substantial danger of bot traffic. Often, this traffic flows through multiple traffic brokers, and these brokers won’t necessarily ask many questions, resulting in selling of traffic with unknown origins.
Innocent publishers can also be victims of a drive-by – bots are designed to look human. Only visiting one site is not very human looking, so bots may visit popular publishers to appear more realistic in their behavior. This can cause bot traffic to occur on sites that are not engaging in traffic buying at all.
6. What are some of the negative impacts from ad fraud and why should advertisers care?
People buy things. Machines do not. The point of advertising is to get people to buy things, so serving ads to machines instead of people is a waste of money. Fraud also tarnishes confidence in the industry overall, affecting even very clean sources – buyers’ skepticism about traffic quality will be reflected in the CPMs they’re willing to pay.
7. How prevalent is ad fraud in the industry and how bad is the situation?
Estimates vary substantially. There is a baseline of suspicious traffic that will occur everywhere. Again, fraud detection is imperfect. It isn’t always possible to have smoking gun evidence of fraud, however this baseline level should be well under 5% of impressions. Significantly elevated levels of fraud can be found in pockets – certain sites, certain sellers, etc. Sometimes common sense will indicate where fraud might be a concern, but in other cases it’s not obvious.
Fraudlogix recently reported that 68.2% of fake impressions come from 3.2% of publishers.
8. How can marketers detect ad fraud and proactively avoid it?
Here are a few strategies:
Audit the media you’re buying. Visit the sites and see if they plausibly have a human audience. Build blacklists of sites you don’t trust.
Use pre-bid segments available from comScore, DoubleVerify (DV), Integral Ad Science (IAS), Peer39, and others, to help avoid bidding on fraudulent impressions.
Use fraud measurement vendors. The above list of vendors with pre-bid segments generally provide measurement services, too. There are some vendors who will offer free trials so buyers can get a sense of their fraud exposure to help them determine whether they want to pay for measurement services.
9. What are some best practices for buying quality inventory?
In addition to the avoidance tactics above, here are a few strategies:
Don’t buy just anything. Whitelist sites you want to buy.
Avoid middlemen. Use PMP deals to ensure you are receiving inventory directly from the publisher. (This helps avoid misrepresentation.)
Be willing to pay real prices for real users. Low CPMs are less likely to meet publisher price floors on quality inventory.
Interested in other Centro resources that will help you understand and combat ad fraud? Reach out to firstname.lastname@example.org.