If you’ve been around digital advertising for a while as I suspect you have if you’re reading this blog post, you’ve heard of click fraud on search engines a few years ago and might have heard of display ad fraud.
Fraudulent display ad impressions have been around for a little while and a few very bright people have started identifying how these occur as there are many different ways.
What is display ad fraud?
Display ad fraud is when ad impressions are billed to a client, and may have been delivered somehow but no human actually every saw the ad.
There are several methods in which this can occur.
One method uses spyware to run virtual browser windows open on specific sites (those who’ve paid for extra traffic) which will track your real browser window activity (time spent, page views, sessions, impressions, clicks, etc…) – but you would not know or see this is happening unless you run an anti-spyware scan. To be clear, those websites count new traffic with all the metrics, although no real human is paying attention to it. The site would also deliver ads to these virtual visits which might be clicked on depending on what you do on the real open browser window.
Another method is hidden within complex ad tags that can hide several, if not many dozens, of different creative from different advertisers stacked overtop one another. The user only sees the top one while the others are all on layers below the website’s content. Every ad counts metrics from the exposure as though it was the one on top – see by the user: impressions, time spent, interactions / dwell / engagement, etc…). Being on lower layers of the web code, any extension would occur behind the site so the user wouldn’t see it. Sound would also not be played – as though the user’s sound is off.
Check out these links for much more on display ad fraud and how they occur:
How big is this threat?
That all depends on which source you want to look at.
comScore released a study on this in 2013 indicating they detected (through their vCE / validated campaign essentials) up to 0.16%% of ad impressions were delivered to non-human agents. http://www.comscore.com/Insights/Press_Releases/2012/3/comScore_Releases_Full_Results_of_vCE_Charter_Study comScore further estimates that 50% of all web traffic is generated by non-human agents – however you should probably already know that ad servers already filter traffic based on a constantly updated list of known bots and non-human agents (most of which are quite legitimate and necessary to the good running of the web).
A study from Integral Ad Science indicated up to 20% of total ad impressions on ad exchanges could be fraudulous while only 2% is deemed suspicious on direct buys. http://integralads.com/about-integral/news-press/2013-09-11/integral-ad-science-releases-its-q1-q2-2013-semiannual-review
This is a huge range and the truth is likely somewhere in the middle as we could probably improve on both methodologies. Nevertheless, it is clear that display ad fraud is very real and is of a scale that is not negligible.
Follow this link to read up on the Carat – Red Bull display ad fraud investigation: http://www.adweek.com/news/technology/amount-questionable-online-traffic-will-blow-your-mind-153083
How can you spot this?
This all depends on where you’re sitting in the industry: advertiser, ad agency, publisher.
I myself have received a few “buy attempts” from unknown media agencies / ad buyers requesting a quick and substantial buy. If you are a seller and the new buyer isn’t asking anything about your company, audience and capabilities, that should be a red flag right there. Any new (smarter) buyer who’s in a hurry to buy but won’t proceed with a credit check is another huge red flag. I have colleagues in the industry who’ve received similar requests both by email and over the phone and were suspicious of them – my main reason for writing this article. If it looks too good to be true… = Red Flag!
IAB (US) has issued an advisory on display ad fraud which you should read. It outlines what it is, how it works and how to watch out for it. Download the advisory here: http://iabcanada.com/guidelines/best-practices-traffic-fraud/
The following metrics are easily gamed or affected by display ad fraud: ad views, clicks, clickthroughs, video completions and cookie counting. Avoid using these as your primary success metric as you can easily be fooled by fraudulent metrics.
Here are a few “red flag” identifiers of potential display ad fraud. A publisher’s traffic has skyrocketed overnight and you’re not seeing news / trends that would explain this. High degree of audience duplication on distinctly different content websites. Browser stats inconsistent with known industry averages. Sites with more than 4 ad creative / tags per page represent a high risk for ad fraud.
What can you do about this?
IAB Canada and the IAB (US) in their advisory on display ad fraud recommend the following: set goals that are clear and measurable; be willing to pay the real price for media; document your goals clearly – ensure both buyer and seller agree on them; don’t optimize on cost alone.
By setting goals you should ensure you identify success metrics that clearly indicate human behavior or interactions – something a “bot” or machine doesn’t do. For example, clear indicators of human behavior are purchases, subscriptions and verifiable brand survey results.
If you are performing some type of attribution modeling, ensure you are tracking users to one of these metrics to ensure you’re attributed value for actual human behavior as opposed to attributing value on robot interactions – which at the end of the day will not generate any revenue for your company.
Authored by Samuel Parent