CTV that's NOT CTV - "see Fou yourself" FouAnalytics | Dr. Augustine Fou

Editor's Note

FouAnalytics is a syndicated column by Dr. Augustine Fou—one of the leading independent voices in ad fraud detection and digital marketing measurement—applying methodological rigor to the programmatic advertising supply chain. This piece exposes how cheap Connected TV (CTV) impressions are routinely spoofed on mobile apps, display slots, and data centers, giving marketers the forensic insights needed to eliminate wasted ad spend and verify true living-room delivery.

You bought a lot of CTV impressions for cheap CPMs through a DSP? Well, you just got ripped off. Let's take a closer look at a few examples as seen in FouAnalytics. All of this is happening at scale, and large advertisers are getting seriously ripped off, but they could not see this in the existing reporting from the platforms or the reports of legacy verification vendors.

CTV on websites and apps

Your CTV ads didn't run on a big screen in a household. They ran on websites (e.g., campaign 1 below) or mobile apps (e.g., campaign 2 below). When your video ads run on websites or mobile apps, they probably have sound off, probably in a display ad slot like 300x250, and probably not viewed at all by the users who left the page in under 15 or 30 seconds. Even if your reports tell you that 80–90% of your ads were viewed to completion, that's because the bad guys invoked the complete.gif tracking pixel to deliberately falsify the reporting.

Direct buy CTV, but with undisclosed audience extension

Don't assume that if you do a direct buy, it is fraud-free. Below is an example of a direct buy at full CTV prices, but yet 80% of the impressions went to websites and mobile apps, NOT CTV devices. This was because the seller was doing audience extension without disclosing that to the client. The client threatened to pull all spend from this CTV seller, so they finally did the right thing and turned off audience extension for this client's campaigns.

CTV on wireless, mobile

Large-screen connected TVs are usually in the living rooms of households. They are connected to Wi-Fi or Ethernet that gets internet access through the cable modem in the household. But in the following example, we see the ASNs (names of ISPs) are wireless carriers like T-Mobile, Verizon, Cellco, Surf Air Wireless, et cetera. That means your CTV ads did not run on the large screens in the household.

We typically see 5–25% of CTV impressions running on wireless IP addresses rather than household ones. Of course there may be a few instances where you'll let this slide — e.g., the kid on the iPad watching Disney+ in the back seat of a car. But then again, is that the right audience, does that deserve a high priced CTV ad, etc.?

CTV not to household

A related but different issue to the wireless one above is that some CTV ads never get delivered to a household at all. In CTV, there's SSAI ("server-side ad insertion"), which means the CTV ad opportunity is routed through one or more data centers so the video ad can be "stitched" into the CTV stream. That's how it's supposed to work. So in the FouAnalytics data, we expect to see more than one IP address AND one or more of those will be data center IPs. Here's how it looks in FouAnalytics — GEO_IP_0, GEO_IP_1, GEO_IP_2, GEO_IP_3, GEO_IP_4 (up to 5 IP addresses recorded in the service of a single CTV ad). At least one of those should be a residential IP address because the CTV ad should go to a household to be shown on the big screen connectedTV, right? If we don't see ANY household IP addresses in the chain, then the CTV ad didn't get delivered to a household. This is typical of CTV ad fraud where bad guys use server scripts to fabricate entirely fake CTV bid requests, no TV needed, no streaming stick needed, no streaming app needed. And they are getting away with it at scale. That's the cheap CTV stuff you're buying.

CTV without any detectable source

A final problem that you can easily see in FouAnalytics data is the following — e.g., 61% of the CTV impressions have no detectable source, regardless of CTV bundle name, website or mobile app. The data grid below shows the source is detectable in 39% of the impressions. The other 61% have no detectable source — i.e., we can't tell where the ad went. There can be many explanations for this, but it boils down to whether you want to continue buying CTV impressions which are not measurable at all.

For contrast, you can buy CTV ads where the source is 100% detectable, and a glance through the data shows normal CTV app and bundle names, some of which you may even have heard of before like Tubi, Plex, Xumo, Pluto, etc.

Supply sources that don't have CTV inventory

And last but not least, it is super clear which supply sources have CTV inventory and which do not. If you followed along so far, you will remember that ads shown on websites and mobile apps are NOT CTV and get an immediate red flag in FouAnalytics if the campaigns are supposed to be CTV (i.e., you're paying CTV prices). In the example table below, you can see a large number of supply sources in a CTV campaign. Note that only 2 have any blue. One quick way to ensure you are getting CTV ads when you're paying CTV prices is to cut off all the supply sources who don't even have legitimate CTV supply. If you leave those supply sources on, you're basically at risk of outright fraud (entirely faked CTV bid requests), mis-marked inventory (display and OLV sold as CTV), or wanton arbitrage (you're paying far more than you need to be because middlemen are arbitraging the supply and reselling to you while pocketing margin).

What does CTV measurement look like in FouAnalytics?

When you have these details you can further optimize your CTV campaigns. For example, if you see a percentage of your ads running on websites and mobile apps, you can block those. You can also edit your campaign settings to target only Connected TV, and not other deviceTypes. All of these will help you tighten up your expensive CTV campaigns.

So what?

Put simply, you can't trust the reporting supplied by the vendors you are buying CTV impressions from. You should always measure it yourself, for example with FouAnalytics, so you can see if you are getting what you thought you were buying. If the ads were detected to be running on websites and mobile apps, you didn't get the CTV ads you paid for. If the ads ran on mobile and wireless IP addresses, they were not delivered to a big screen in a household, which is supposed to be on a cable modem at home. If you see the data shown above, you can talk with the vendor you are buying from and get to the bottom of it or get a refund for the stuff that was not CTV. Or you can check your campaign settings to make sure you selected DeviceType=CTV, for example. "see Fou yourself" with FouAnalytics CTV pixel.

Bonus, what great CTV campaigns look like

The 12% light blue comes from one flag that shows these are wireless IP addresses. Normally, large-screen connected TVs are in the living room and connected to the cable modem in the house. When the IP address is wireless, it suggests that the CTV stream is being played on a mobile device, like an iPad. This is not bad or fraud, it's just not the large screen in the living room. Again, as analytics, FouAnalytics flags this for you to review. You can then decide if it's an ad running on Disney+ being watched on an iPad in the back seat, etc.

Why Subscribe

Because your "CTV" campaign might actually be running in a hidden mobile app slot—and the platform selling it to you won't tell you. Every week, FouAnalytics uses forensic campaign data to expose supply path waste, fake data center traffic, and audience extension arbitrage, giving media buyers and budget owners the independent proof needed to protect their investments.

Subscribe below to read more from FouAnalytics weekly here on State of Streaming.