Commentary

Programmatic Insights: A Close Look At Mobile Ad Fraud

As we ring in 2018, one thing is clear: Media quality remains a top challenge for digital marketers. With mobile ad spending accounting for 70.3% of total U.S. digital advertising spending, many of the techniques used to perpetuate fraud as well as detect it have migrated from desktop to mobile. Mobile ad fraud comes in several varieties; let’s take a look at several of the most prominent.

Audience Fraud: Mislabeled Inventory

One very common type of mobile ad fraud involves mislabeled inventory, often incentivized inventory — whereby visitors receive points for clicking — that isn’t labeled as such. Incentivized inventory performs strongly on click-through and install metrics but delivers users who are unlikely to engage in any meaningful way or provide real value to the advertiser. The profit margin for the fraudster comes because incentivized inventory is cheaper than non-incent, so someone who can buy inventory at incentivized prices and then re-sell it at non-incent rates can skim a lot of money off the top. Companies that look only at Cost-per-Install (CPI) are particularly vulnerable here and may even be inadvertently optimizing towards fraudulent inventory. 

Audience Fraud: Misleading Targeting Data

Other types of audience fraud include reselling impressions as belonging to a different demographic or in a different geographical location. This type of fraud is made possible because networks that can’t deliver enough volume to meet their clients’s demand often buy surplus inventory from others to fill the gaps. There is an entire ecosystem of brokers supplying that demand, and ad impressions can go through half a dozen sellers and resellers, each taking a cut and each with the potential (and motive) to obfuscate critical information. This can make it very difficult for a network to tell what it is that it is being re-sold.

Impression Fraud

Another major form of mobile app ad fraud is impression fraud. In this scenario, an app can either load multiple ads at once and “stack” them so consumers only see one, or it can load ads in the background when the app isn’t even active. In both cases, advertisers who buy those impressions hoping for a chance to engage the real user will be disappointed because the end user will never even see the ad, let alone click on it. This happens with both banners and video.

Click Fraud

Click fraud is another common tactic. This scenario might involve a bot clicking an ad that the user never sees in order to drive revenue for the app developer. In combination with impression fraud, it can be very lucrative for fraudsters and absolutely disastrous for brands and programmatic buyers.

Click Injection

Click injection is an extension of click fraud. On older versions of Android, it was (and is) possible for apps that had the right permissions to see which other apps an end-user had installed. If that app was running in the background, it could potentially see new apps as they appeared on the end-user’s device and then fire a fraudulent click on an ad for the newly installed app before the end-user had a chance to open it for the first time. Since the mobile ad industry uses last-click attribution, this fraudulent click would get credit for the install and the developer of the fraudulent app would get paid. 

Install Fraud

Install fraud is the most complex kind of mobile fraud (so far). In this scenario, a fraudster may have a large number of physical or virtual devices with sophisticated software that spoofs clicks, downloads, installs, and first opens. After each install, the virtual machine’s device identifier is changed so it can rinse and repeat. Advertisers that spend with the goal of driving installs who don’t look at or optimize toward post-install engagement are particularly vulnerable to this kind of fraud.

Vigilance and a Solution

Fraud is problematic for practically everyone in the space. For media companies, fraud undermines the ability to serve clients’s very demanding performance metrics. Brands also have significant skin in the game; poorly defined campaign metrics can result in phenomenal sums of money being spent to acquire users who don’t actually exist or who have no interest in actually engaging with the brand’s app or website.

Optimizing towards real engagement and not just installs is one of the most important things brands can do to limit the effect of mobile ad fraud. Ultimately, relationships are built on trust and the ad tech industry is no different. Staying on top of bad actors and the methods they use in order to combat fraud helps improve that trust.

2 comments about "Programmatic Insights: A Close Look At Mobile Ad Fraud".
Check to receive email when comments are posted.
  1. Mark Scott from Sage Projections, January 10, 2018 at 2:02 p.m.

    Interesting article. I wonder if the author would share what percentage they believes is actually is impacted by fraud? This really hurst businesses. They are paying for something they are not getting and in the process not getting their product messages out.

  2. Jed Wheeler from RhythmOne replied, January 10, 2018 at 3:10 p.m.

    The percentage varies wildly depending on the campaign and the optimization criteria. 


    As mentioned in the article, app install campaigns that do not look at downstream engagement are likely the most vulnerable to several of these types of fraud since they will likely be actively optimizing toward mislabeled inventory and install fraud inventory.


    Campaigns that are optimized strictly on click-through are also at risk for the same reason - optimizing just on clicks means the impression frequency will be highest on fraudulant inventory where clicks are plentiful.


    The solution in both cases is the same - optimize campaigns based on downstream actions.  If you give your budget to sites and apps that drive the most engaged users instead of the most installs you'll eliminate most fraud and end up with a stronger more engaged user base as a bonus. 

Next story loading loading..