Digital Ad Fraud is Huge and it Doesn't Matter to Advertisers

Despite all the clean up efforts, digital ad fraud still accounts for a huge amount of the ad economy.

How much?
No one knows.

When we buy ads in newspapers, or on TV we have no idea how many people see them. There are circulation and ratings guides, which may or may not be accurate. If a newspaper is delivered does anyone read it, and do they get to the page with your ad? When your tv commercial plays does anyone see it, or are they staring at their phones? There are no detailed numbers, but we are ok with that because thats the system.

Digital was supposed to be different. Impression level tracking. Attribution. Finally the measurement and accuracy that we craved. We wait for the day that digital/programatic style ad buying makes its way to TV.

But if I buy a digital ad on a major publication, are the impressions real? Or are they driving bot traffic to their own website to prop up their visitor numbers and to burn through my ad impressions?

And if I buy ads programatically or through an exchange, where my ad is matched with people or content across the web, are the websites, apps or people actually real?

How does ad fraud work? permalink

What is the simplest form of ad fraud? I build a basic website and put some content up and then install advertising through an exchange. I get paid for the impressions of the ads on my site, and to get more ads, I buy fake bot traffic to come and browse the website. As long as the cost of the fake traffic (which is just a computer program moving through my site from one link to the next) is less than my advertising revenue, then i’m profitable and can keep building more and more such websites.

If the bot traffic is more sophisticated, and has been systematically browsing particular topics and downloading the cookies from other sites, then the ad impressions can cost even more. Maybe the bot is pretending to be a young mother, visiting lots of parenting websites. A company targeting ads to mothers will pay more for that impression.

App ad fraud is just as simple. I make an app, put it on the app stores and run a cheap campaign to get installs. Then the app can just run in the background pretending to be used and generating ad impressions.

The fraudulent websites and apps are hidden behind layers of shell companies across multiple countries, making it almost impossible to identify who is actually making the money. There are often fake employees with fake social media accounts who are the ‘respectable face’ of the fraudulent sites. The legal cost of recovering the money or prosecuting the people responsible means that there is little reason to do so, which makes the fraudsters even more brazen.

In a case in Israel, after a network of fraudulent apps was revealed, the network tried to sue an ad vendor for withholding their payment for the fake impressions. They dropped the case when they realised that that would have to show up in court in person.

From 2015-2016 the ad exchange AppNexus cleaned up much of the ad fraud on their network. Ad impressions fell by 92%. Yes, almost all the reported impressions had been fake, or repetitions. This was great news for advertisers who would no longer be paying for ad impressions to bots, except that the new cost for the real impressions went up by 800%. Advertisers were paying the same amount, and were probably getting the same outcomes from the real people who were clicking and buying things.

We’ve even discovered lead-gen fraud for campaigns that are setup on a pay-per-lead basis. There are call centres of people who fill out website forms, accept the phone call from the hapless advertiser to confirm their interest and say they are busy and can speak later. They’ll drag the conversation out for weeks, insisting that they are interested but can’t speak just now. Finally their mobile number is disconnected and the email account ignored. But by then the advertiser has marked the lead as genuine and paid for the lead. These fake leads are blended into real lead-gen campaigns to boost the lead numbers while still ensuring some actual sales take place and the advertiser continues to use the service.

Who is incentivised to stamp out ad fraud? permalink

Not the marketing managers, who love reporting the lower costs and expanding reach that come with junk traffic. Who wants to report to their company that ad cost is up and reach is down because they’ve been paying for fake impressions for years? Advertisers are far more likely to take action when a single ad shows up on an unsavoury website, or is paired with a nasty YouTube video.

Not the ad exchanges, who make their commissions either way. They only need to do enough to avoid lawsuits that accuse them of not doing enough.

Generally speaking, not the publishers who like to know that they can boost revenue with some fake traffic.

The companies who sell anti-ad-fraud software are certainly incentivised to push for truth in advertising as it makes them money.

Small publishers who are genuinely trying to produce high quality content and make a living through advertising would love to see their earnings rise by stopping the junk impressions that devalue the system. But they are the least-powerful players in the ecosystem.

What is the future? permalink

To an extent, digital advertisers need to accept that they face some level of fraud. Performance marketers - generating leads or sales - have an easier time as they have performance metrics to validate their work. If sales are growing, then some fake ad spend doesn’t matter as much. Brand advertisers need to take a more proactive approach to their ad spend management, with in-house ad servers and ad-fraud tools. This will put pressure on the exchanges and ad networks to find and remove fake website and apps.

It is also important to have good measurement protocols in place. Offline advertisers measure their work though offline surveys. They go and ask lots of people if they saw the advertising and what they remember about it. Digital advertisers often miss this step. Impressions etc are great, but how are you tracking whether people have seen your marketing or not?