Programmatic Advertising: the Rise of Ad-Frauds and Lack of Brand Safety!
It’s estimated 50% of the $709 billion Digital Ad expenditure is lost to Ad fraud! Read this article to understand different types of fraud and its solution. What is Brand Safety and how it’s impacted!
In one of my earlier articles, I covered the fundamentals of programmatic advertising and how it’s the solution to the ever-growing complex Advertising world! But like every medicine, it also has side effects, two of those are Ad-Frauds and lack of brand safety. In this article, we will understand what is Ad-Fraud, and what are a few common types of Ad-Frauds. And what do we mean by brand safety and how programmatic advertising impacts it. Later on, we will see the rise of new entities in the AdTech ecosystem to handle these issues and where these entities fit in the whole ecosystem.
This article assumes basic knowledge of programmatic advertising. However, I will touch upon the basics wherever required. A very basic flow is shown in the below image:
Ad-Frauds!
As we know, Advertisers pay for Ads based on impressions, clicks, conversion, etc. But it’s implicitly understood these impressions, clicks, etc. should be by a HUMAN. Ad-Fraud is when these impressions or clicks are not generated by a human but a piece of software program popularly known as bots.
Advertisers pay for ads because the expectation is — humans will see or interact with ads and may visit the website, know about a brand, or buy the products. Since the bot is now making clicks and impressions, none of the advertisers’ targets are met.
What’s the Big deal?
One might think it’s not a big deal. But to understand the impact let’s see some numbers.
- As per Magna (leading US agency), in 2021, the global Digital Advertisement market expenditure was about $709 billion. And, expected to rise by 12% in 2022 and will hit the trillion-dollar mark by 2025!
- As per many programmatic solution vendors, for every dollar spent on online ads 40 to 50 cents are stolen or wasted due to Ad-Frauds! Now imagine the scale. No one can predict the exact number as it’s impossible to detect all fraudulent clicks and impressions.
- It’s the number one cyber-crime in the world as per revenue. And second most profitable organized crime after drugs!
Another interesting fact is around 45–50% of the internet users are bots, in one way or the other, generating 55–60% of the internet traffic.
These AdFrauds can happen with any advertisers or publishers, even the big and premium ones (we will see later).
Type of Ad-Frauds
There are hundreds of types, and it’s impossible to cover all those. Here we will go through the common ones.
- Botnet Traffic: This is one of the most common ad frauds. Malware programs (bots) are installed on the consumers’ computers. These bots run in the background and keep viewing ads, clicking links, and visiting various websites. These are hidden from the users. See below a snapshot from the ForensiQ (now Impact) showing that in 24hrs a bot generated around ~10K fraudulent impressions! These bots rerun every 5 mins and then become dormant. This fraud is not specific to the pc but happens in the mobiles too. Dishonest publishers generally use this method.
- Re-Targetting or Remarketing frauds: Re-Targetting or Re-Marketing refers to displaying ads to returning users, for example, you want to display ads to users who have visited your site in past or have downloaded your app, etc. Usually, bids for such users are a bit higher than the normal ones. In this fraud, bots visit the famous big commercial sites. So that they get added to the re-targeting pool. Now, these bots will visit the publisher site. Because the bot was added to the retargeting pool, the advertiser will bid for the impressions and will be tricked to pay for a bot impression! Dishonest publishers generally use this method.
- Pixel Stuffing: Ad is placed in 1x1 pixel ad space on the publisher page. Even though the ads are not visible but technically an impression is served. So, the advertiser has to pay. Dishonest publishers generally use this method.
- Ad Stacking: Multiple Ads are stacked on a single AdSpace. Only the top ad will be visible but all the ads will load on the page. Hence, advertisers for those hidden ads have to pay even though it was not invisible.
- View Port Area: Viewport is the visible area on the website. In this fraud, ads are loaded outside the viewport area.
- Pop-under: In this Ads are opened in the new pop-up window but behind the main webpage. Hence, the name pop-under. No need to say most of the time these Ads are not seen by the users.
- Video Auto-play: Many advertisers pay when a video is played by the user. But in this fraud, the video is auto-played in mute mode.
- Ghost sites: High-traffic websites contents are copied and a dummy website is created. Then using botnets traffic is generated for the site. So, this website now looks like high quality and high traffic website. Advertisers end up buying the ad spaces but with no real users to view them!
- Mobile device id reset: Bots download the apps and then reset the mobile install id, uninstall the app, and then re-install the apps.
Third-Party Frauds: Most of the above frauds are done by dishonest publishers. Below are the frauds committed by third-party.
- Impression Laundering and Bundle ID spoofing: Fraudster pretends to be legitimate website such as New York Times using the tech and labeling themselves as New York Times. During bidding, the advertiser will buy the impression thinking it’s the original website. In a similar way, the ads can be displayed on other apps which pretend to be some other legitimate app.
- Ad Injection: When ads are inserted/displayed on the publisher’s websites without authorization. It may display an overlapping ad, replacing the original ad or displaying ads outside the viewing area. Needless to say, the revenue goes to the fraudster.
- Click Hijacking: When a user clicks on an ad instead of that ad user is taken to some other webpage or URL. Ad is served but the advertiser loses the user. This is usually done by malware on the user’s machine or compromising the publisher website or user’s proxy.
Obviously, one or more of the above frauds can be done at once, and more than not that’s the case.
What to Do?
This is a big problem! But for an Advertiser, it’s not possible to detect all these frauds on their own. It’s said every customer problem is an opportunity for a startup or a product. A new entity or tool came into the ecosystem — AdFraud Protection and Detection (AFPD) platform. Their sole purpose is to identify fraudulent behavior and notify the DSPs. If an impression is flagged by these platforms DSP doesn’t participate in the bidding process. These platforms/tools use Machine Learning models to detect anomalies or fraudulent behavior. This is similar to the ML models implemented by Visa or Master card but instead of users’ transactions activities, this model analyzes users’ ad viewing patterns. Generally, these platform charges a few cents per CPM.
There are two approaches to implement this solution:
- Manual: AFPD platform will provide the additional codes. Advertisers will wrap their ads with these additional codes.
- Automated: Another approach is to establish a connection between the AFPD platform and the advertiser’s DSP. This way additional code will be added automatically. But this depends on whether the DSP can establish a connection with the AFPD.
Advertisers generally add tracking code snippets to track user behavior on the advertised website. A few of the famous AFPD are — Double Verify, Confiant, Adjust, Anura, etc.
As we saw in many cases ads being not viewable was the issue. Hence, IAB (Interactive Advertising Bureau) and Media rating council developed the below standard.
Facebook guarantees — 100%, Youtube guarantees 95% whereas programmatic advertisement guarantees only 54% of viewability (that means authentic viewability).
Most of the time AFPD platform caters to the viewability needs too. And nowadays most of the DSPs come with their viewability platform/features. Advertisers can set their viewability criteria/parameters.
Brand Safety
By now it’s pretty much clear, in programmatic advertising, advertisers do not have complete control over where the ads will be displayed. If the ad is displayed at any place which doesn’t align with the brand image, value and messaging of the brand then that negatively impacts brands.
As per IAB “ Brand Safety is the practice of keeping a brand’s reputation safe when it’s advertised online.”
It’s estimated around 10% of ads are displayed next to risky, controversial ads. That means 1 out of 10 Ads!
Few example, imagine a user is in the brand’s retargeting pool. Later, that user visits some adult site. There are very high chances advertisers will bid for the impression and the ad will be displayed at the adult site. This can impact the brand’s image.
Similarly, if the ad is displayed along with the content which promotes or is hate speech, it may give an impression brand is supporting the same. There are many such scenarios.
Few sites which most of the common brands avoid can be — adult sites, drugs, military conflicting sites, crime, death, obscenity, hate speech, and many more.
Brand Safety Solution
As we saw, AFPD solutions will not be of much help in brand safety. It may prevent a few redirections to such sites but most of the time it’s a genuine user. Hence, AFPD will not block those.
AFPD providers understood this and most of them extended their functionalities and many new players came into the ecosystem to provide the service known as Brand Safety Solution (BSS) providers. Nowadays, most of the DSP has a connection with at least one brand safety solution provider, or have built these tools/platform themselves.
When an impression is generated and reaches BSS providers/DSP it uses ML and Deep Learning models to understand the website’s content. For example, it can use sentiment analysis using the BERT model (just an example) to understand what kind of textual content is displayed on the website. If the content has negative sentiment, or it’s hate speech it can block the bidding and the ad will not be served on such site. Similarly, the Computer Vision model can be used to detect nudity, crimes, death-related images on the website, and block ads. These parameters are generally given by model on some scale and advertisers have the option to choose the threshold. For example, for an innerwear brand, it’s okay to display the ads on fashion-related websites which may have a person in a bikini or innerwear. But it might not be okay for children or religion-related brands (just an example).
This is not important only for advertisers but for publishers too. If a publisher can guarantee the ads will be displayed as per brand image. It can attract more brands and may charge higher CPM.
Just to summarize the flow with these new entities in the picture. Our initial sample programmatic advertising flow will look like below:
Please remember Programmatic Advertising ecosystem is too vast and complex. And there are various permutations and combinations of the flow and entities involved. I hope this article gave you some insight and knowledge about AdFrauds and its scale. And how programmatic advertising can impact the brand’s image and the available solutions.
In the next article we will touch upon a few new players in the ecosystem — Create Management Platform, Dynamic Creative Optimization (DCO)Platform, Web Analytics Platform, Digital Asset Management, Tag Management System, and Ad-Order Management System. And we will touch upon Marketing Attribution Platform.
