Adsystem Overlap, Implications for Programmatic Media Buying
Many marketers assume that they are getting more reach when buying through programmatic channels. For example, instead of buying direct from a handful of good publishers, marketers and their agencies have been allocating more and more digital budgets to programmatic so their ads go to many more sites, sight unseen.
But if marketers actually got detailed placement reports that showed them the sites and the number of impressions that went to each site, they will quickly realize that they were not getting anywhere near the "reach" they thought they were getting. Check your campaigns and get back to me to tell me I'm wrong, if I'm wrong.
Furthermore, using data from Deepsee.io, we can see that most adsystems sell exactly the same inventory. That's right. When we analyzed the data in public ads.txt files and cross referenced them to entries in public sellers.json files, we found that the vast majority of the domains that had legitimate ads.txt sellerIDs that could be matched to sellerIDs in the adsystems' sellers.json were overlapping. Taking just the top three adsystems by size -- A. Google (493,544 domains), B. Rubicon (128,878 domains), C. Pubmatic (128,500 domains) -- the domain overlap was 26%, 97%, and 95% respectively. As you can see in the Venn Diagram, all of B and C fall within A. And B and C also overlap for the most part. For 50 more adsystems see the table below, sorted by largest to smallest size.
The implication? Advertisers can buy ads on all programmatic domains by buying from one adsystem - Google. Google is 4X larger than the next largest ad exchange. By buying through one channel, advertisers also reduce the risks of fraud and reduce supply chain costs. Any good SPO (supply path optimization) effort would have come to the same conclusion. If there is literally NO difference in the domains that any ad exchange offers, what value add is each exchange bringing to the table, for the fee they are taking. By reducing the number of intermediaries in your programmatic buying, you are reducing risk and also reducing the 50% adtech tax, so more of your dollar goes towards placing the ads.
If you stuck with me this far, then wouldn't it be better to buy direct from good publishers that have real human audiences, instead of buying cheap ads in programmatic channels and then trying to detect your way out of trouble by paying more for fraud detection, viewability detection, brand safety detection, etc. All of that money goes into the pockets of the ad tech companies and not to the publishers for showing your ads.
As always, if you have questions or if you want to discuss, debate, or refute any of the above, please get in touch. I am always open to hearing that I am wrong and to learning from more accurate information and data.
GTM Customer-Centric Strategy & Analytics | $120M+ Generated for AdTech & MarTech SaaS
3 年Great analysis on overlapping. While this is very true, a few nuances came to my mind: 1. While domains might be the same, the context of the ads (and goals) can differ and not be achievable in one single DSP. For example, you have content brand awareness campaign, and you need to buy Native Ads. You can't get it in Google, you have to go to SSP like Taboola or Revcontent. Rubicon or Pubmatic won't have it either, as SSP never sells 100% of inventory (often remnant) to DSPs due to frequent QPS limitations that isn't easy to maintain and yet supplying existing advertisers to keep ad revenue directly. However, if your campaign's goal matches display type, then yes, Google is your choice. Although Google's most premium inventory is not available openly, you have to go through Google Partner in case you need a significant reach. Which is impossible for smaller buyers on self-served basis. In this case, again you have to go this time to display inventory providers, like Rubicon. Difference is huge. And also availability. 2. When it comes to inventory overlap and decision what DSP to choose: then you consider how this particular DSP handles SPO itself. Every DSP is different and is good or bad in different ways. The best DSP is the one that does most of the work for you yet providing quality and maintaining reach, but there is no best DSP at the same time. That's why marketer's stack usually consists from around 5 platforms on average. But even then it's not enough to just have access to all inventory. It also comes to consideration customer service, platform features, how easy to use, etc. There is still plenty of great opportunities for AdTech companies to compete for ad budgets, and plenty of room for growth. Also, if you have one single DSP, then such factor as competition stop to move industry forward. That's why these ad exchanges will continue to exist and always find their buyer.
Advertising | Arbigo.com and Supernova.inc
3 年What should investors of the non-google companies do? Should marketers allocate all of their marketing budget to Google for programmatic media? Does the industry want to support only certain amount of companies? Analogously speaking, why should we support local businesses when Amazon, Whole Foods, Walmart, Costco, etc. have better pricing and availability for the most of the same products? Should the world only drink Coca-Cola or Pepsi? Should we de-duplicate the potato chip market? Should we de-duplicate the bottled water business? I believe, people do business with people they like. Until robots take over, this is a human relationship driven industry still.
Data Privacy Researcher ?? Partner @ ObscureIQ ?? Co-host of YBYR
3 年Really love this, Augustine Fou, PhD. And more more companies I need to add to my Broker database. ??
Engineering leadership at The Trade Desk.
3 年"Any good SPO (supply path optimization) effort would have come to the same conclusion. If there is literally NO difference in the domains that any ad exchange offers, what value add is each exchange bringing to the table, for the fee they are taking." The whole point of SPO is to find the best path to any piece of inventory. It's complicated to pull off but someone who is looking for the best price on like-for-like inventory is going to need to buy from more than one exchange to get it. I'll agree that this probably has a point of diminishing returns. You don't want to be managing SPO through 40 different exchanges.