Best Practices for Buying CTV - FouAnalytics "see Fou yourself" | 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 shows CTV advertisers what a well-run campaign looks like in the data, what a broken one looks like, and how to fix both before your next budget cycle.

What are some best practices buying CTV. Show me, don't tell me. OK, here's some data to show some great campaigns, and others that can be improved, and how.

Did your CTV ads deliver in the right geo?

In the following example, 99.9% in the header of the data grid means we have data in 99.9% of the impressions. That's good measurability. Of the data that was collected, 97.5% of the impressions went to the U.S. which is great, because this campaign was meant to run in the U.S.

What do you think of the following 2 campaigns? The one on the left was meant to run in France and the one on the right was meant to run in Germany. If there are other countries in the data and it is greater than a few percent, you can negative target the unwanted geolocations to tighten up the campaign.

If you need more detail -- for example for political campaigns, the right country is not enough, it has to be in the right city and state too. See the following example. The GEO_IP gives you a sense of frequency (the number of ads per IP address - household). The GEO_AS tells you the ISP or cable modem service the user has, like Charter, AT&T Enterprise, T-Mobile, etc. The GEO_REGION tells you the state. This campaign was meant to run in Ohio. GEO_POSTAL is the zip code, and LAT LNG are the approximate geolocations (derived from IP addresses, so not as accurate as GPS, but that is what is available).

Did your CTV ads deliver on ConnectedTV?

For gods' sakes check your campaign settings for device type. This was a major whiff, where only a third of the CTV impressions ran on connected TVs. The other impressions were wasted on PCs, Mobile, Tablet, etc. The simply fix is to have your agency check the campaign settings to target only ConnectedTV device type, or actively block other device types that are not CTV. Again, this is why you need independent measurement to check the work, what you are getting. This could well have just been a mistake or oversight on the part of overworked agency trafficking.

Did your CTV ads deliver in CTV apps?

In the following example, you can see a pretty good CTV campaign, where source is measurable in 51% of the impressions (obviously higher measurability is better). The country is U.S. which is where it was meant to run. And the IP addresses give you a sense of frequency per household. For more examples see this article https://www.linkedin.com/pulse/checking-frequency-caps-ctv-yes-fouanalytics-dr-augustine-fou-ylb6e/

Here's one more good example, with data from FouAnalytics. Note the DeviceType is ConnectedTV, great, right? As opposed to phone, tablet, laptop, etc. CTV ads should run on connected TVs. Note the CTV app names in the Domain and DSP_URL macros (passed by the DSP). And the ONE supply source.

For contrasting examples of CTV campaigns where large portions of the impressions did NOT go to CTV apps, see the following article https://www.linkedin.com/pulse/ctv-thats-see-fou-yourself-fouanalytics-dr-augustine-fou-lxsae/ The CTV ads went to crappy websites and mobile apps, even when the advertiser bought direct from a legit seller. (That mainstream TV manufacturer selling CTV ads was doing audience extension without disclosing it) By measuring with FouAnalytics, the advertiser could confirm whether they were getting what they paid for, and reduced their reliance on reports provided by the platform they were buying from.

Save money by doing your own SPO (Supply Path Optimization)

The above are examples of well-run CTV campaigns. You use FouAnalytics to check what you got from the platforms. And we already saw an example of ONE supply source, plus TWO deal IDs. That is ideal and it simulates a direct buy as much as possible, even though programmatic pipes were used to buy the ads.

The following is an example of unnecessary supply paths AND leakage which costs you more money than it should. If you see this in the data, the solution is also very straightforward. Reduce the supply paths and focus on a few deal IDs to hone in on real CTV inventory.

This was a CTV campaign running in Germany. de.exaring.waipu is a CTV seller in Germany. Notice the supply sources -- 64% came from "waipu" and the other 4 are unnecessary paths. Each of those 4 other sources - appnexus, adscale, yieldlab, google - are arbitraging and reselling. That means they are making margin on your dime. If you want de.exaring.waipu, you should buy from the direct source itself -- custom_SUPPLY=waipu. With this data, the advertiser edited the campaign settings to specify ONE supply path and ONE bundle ID to save themselves money. This is how you do SPO (supply path optimization) yourself, without having to pay for other services.

Finally, don't forget how easy it is for bad guys to pretend to be CTV, just by falsifying the data in the bid request. Years ago the Grindr mobile app and many other mobile apps were simply fabricating CTV bid requests when they were not supposed to. Since that time we have seen smart refrigerators, connected street lights, python code on servers, and even javascript code in ad slots fabricating fake CTV bid requests to the tune of 10s of billions per DAY, and getting away with it. That is the scale of CTV fraud when you buy from any DSP, and not direct from real CTV sellers (who still have unsold inventory).

Conversions and website activity

One final note on CTV campaigns. There's a "performance CTV" vendor that is selling 100% fraud. They write false data into their own customers' Google Analytics (on the landing page, ecommerce site). Those visits did not actually occur. This was discovered by adding FouAnalytics to the same site(s). If the data showed up in Google Analytics -- utm_source=[this vendor] -- but it did not show up in FouAnalytics, it did not occur. That is because FouAnalytics has basic cybersecurity measures to prevent tampering and false data being written into it. This vendor is tricking the attribution reporting to make it look like they were driving a lot of sales from their CTV campaigns. And there's been evidence over the years that they did not even run the CTV ads, or just ran low cost display ads instead.

There are MANY MANY examples of attribution fraud, where adtech vendors are claiming credit for sales/conversions that were not actually caused by running campaigns on their platforms. They have an incentive to do this, so it appears the campaigns are working well so you keep spending money with them. Don't be so easily fooled.

Be sure to pair your CTV in-ad impression measurement with on-site measurement with FouAnalytics so you can verify the on-site activity resulting from your CTV campaigns. And keep in mind you should expect to see NO actual clicks from CTV because how do you click from a CTV ad to your website? (I know there are QR codes, etc. now, but those are still not in use in many campaigns).

If you need any assistance with your digital campaigns, please reach out. I am here to help.

Why Subscribe

Because your CTV campaign might be running on phones, tablets, and fake smart refrigerators — and the platform selling it to you has no incentive to tell you. Every week, FouAnalytics shows advertisers what they actually bought versus what they paid for, with real campaign data, not vendor reports. If you buy media, this is where the independent measurement case gets made before your agency catches the error. If you allocate marketing budgets, this is where CTV fraud, supply path waste, and attribution manipulation get named and priced before your next upfront.

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