Skip to content
All posts

AdTheorent’s Take on the “ANA Programmatic Media Supply Chain Transparency Study”


The Association of National Advertisers (the ANA) issued a report on June 19, 2023, entitled ANA Programmatic Media Supply Chain Transparency Study,” highlighting certain inefficiencies and opportunities for marketers utilizing open web programmatic advertising. The open web is comprised of digital properties outside of the so-called Walled Gardens across web and app properties on the internet. The open web has been a valuable and growing place for advertisers to engage with consumers – and a needed alternative to the abusive retargeting and user profiling factories that comprise the mortar in the Walled Gardens (think Meta, Snap, TikTok). In fact, the open internet market represents $88B in global ad spend as noted by the ANA in its report, citing the “State of the Open Internet Report” by Jounce Media. For this immense opportunity to deliver on its promise for advertisers, programmatic ecosystem participants need to avoid the self-interested practices that have undermined user confidence in the Walled Gardens. ANA’s report outlines several key challenges related to programmatic media buying inefficiencies which impeded return on ad spend for advertisers. At AdTheorent, we have been saying the same thing for years: there are significant industry practices which benefit a handful of ad tech vendors more than advertisers themselves.

For more than 11 years the AdTheorent team has been designing a statistics-powered and ML-powered DSP that puts advertisers first – focusing on measurable business goals such as Return on Ad Spend or Cost Per Action. We believe that examining the ANA report and what it calls for is a useful context to evaluate the AdTheorent business and how it is additive and different. The ANA report highlights programmatic advertising market inefficiencies around which we have built valuable platform and machine learning tools. In the end, we believe we deliver for advertisers the truest benefit and value from open web programmatic advertising.

Size of Market… and Waste

ANA reports that there is $20B (23%) of waste in the $88B global open web programmatic opportunity. This is a startling number, and it is quite interesting to unpack. And as ANA has noted, the report does not recommend cutting advertising costs, but rather optimizing spend so that advertisers realize the full benefit.

Programmatic waste is caused by many factors, including obvious inefficiencies such as bidding on fraudulent impressions and other practices that are less obvious, but equally wasteful, such as bidding on so-called “made for advertising sites” – which is what it sounds like, sites that exist primarily to contain ads, with arguably second-rate content. The prevalence of these less than optimal practices relates back to the same flawed premise: many media-buying platforms prioritize collecting “clicks” – and the transactional act of buying media – rather than driving true and real campaign KPIs (such as sales or store visits) that drive actual business value for the advertiser customer. Another example cited by the ANA is the practice of serving ads on so-called long-tail web sites – which you can think of as more niche or less popular internet properties – that are designed to drive clicks resulting in low-cost campaign execution, but often the ROI is non-existent for advertisers.

Bringing this back to AdTheorent – we agree with all of this! In fact, we built our business to solve for these and other inefficiencies, and our superior campaign performance reflects these efforts. AI is a giant buzzword in the 2023 commercial zeitgeist, but for the past 11 years we have been using and operationalizing advanced machine learning to reduce and eliminate waste in order to drive smart media buying for advertisers. And in addition to campaign KPI performance – and this is table stakes in our view – our ML algorithms help us identify and eliminate fraudulent impressions, non-viewable impressions, non-measurable impressions, impressions that are not brand safe – all prior to bidding on them.

ANA Report Takeaways; How AdTheorent Solves These Challenges to Deliver the True Value of Open Web Programmatic Advertising

The ANA report cites seven Key Findings which are important to consider.

Takeaway 1: “Information Asymmetry is a serious issue for advertisers, leading to inefficient and unproductive media investment decision making – resulting in substantial loss.”

AdTheorent’s Take: This is a common issue in the industry and one that exists if the media buying platform has not invested on the data side – meaning, it has failed to augment and normalize the endless digital data signals in order to make them actionable. Many media buying platforms and networks simply have ID-lists, and they target those IDs blindly. Because user profiling and IDs are not our focus, we have made significant technical investments in approaching programmatic data more holistically.

A couple notes about AdTheorent’s different approach:

  • We do not overpay for inventory because our price and performance optimizers are the most sophisticated tools in programmatic, designed to identify data signals that correlate with performance. We use these insights to inform bidding and bid pricing decisions for the entire campaign. These savings are passed directly and completely to our customers. We do not have any incentive for the initial price to be high so that we can lower it for the customer and then take a cut. That's not what we do. We believe that there's a huge opportunity in this market for that type of transparency.
  • We do not have information gaps as noted in the ANA report because:
    1. For 11+ years we have invested in curating, augmenting, and normalizing the programmatic signals that can be used to decide whether to buy media – no DSP or media buying platform is doing this at the level we are. This allows us to use data more effectively (it can be messy and if work not put in, that data can be useless). We have put a lot of tech and development work into normalizing and augmenting the signals provided to us by supply side platforms (SSPs).
    2. At AdTheorent, we work very collaboratively with our SSP partners. Among other things, our SSP partners send us signals regarding why we don't win auctions and other valuable data points, all of which we use in our ML.
    3. Finally, to put the ANA comment in context, it is important to remember that programmatic media-buying is auction-based and the price you pay for media depends on the DSP’s capabilities. We ingest signals and invest in making them accurate and actionable in order to be informed and become smarter over time; there is no pre-set "price" for an impression. Demand sets the price; so, understanding the data (as we do using advanced ML) gives us a distinct (and growing) edge.

Takeaway 2: “Data Access is Lacking, Resulting in Transparency Issues, Leading to Inefficiencies and Waste”

AdTheorent’s Take: The ANA report cites that the majority of companies participating in the study did not or could not share log level (or impression level) data, and that this data is not shared with advertisers. This is another differentiator for AdTheorent: we regularly share log level data with customers because we believe they have a right to the data – they paid for the campaign.

Takeaway 3: “Misaligned Incentives: Advertisers Prioritize Cost Over Value”

AdTheorent’s Take: Everything AdTheorent does is tied to ROAS and performance - ML models correlate impression data with conversion data, and we score impressions based on that performance data (tied to customer KPIs). No DSP has a more sophisticated cost optimizer framework than does AdTheorent, but the topic of cost cannot be detached from value delivered. For AdTheorent, the primary factor is KPI performance and return on ad spend . . . and obviously cost optimization is a component of that. By way of contrast, some other DSPs treat all User IDs within an audience segment as equal, and then try to pay as little as possible for each impression. AdTheorent has a foundationally different approach, evaluating the predicted performance for every impression, and then seeking to purchase only those impressions and in the process of doing that, maximizing cost efficiencies. This allows us to deliver strong performance for advertiser-specified KPIs in the most cost-effective way possible.

Takeaway 4: “More Waste – The Average Campaign Ran on 44,000 Websites”

AdTheorent’s Take: We think ANA is correct to point out the risks of waste related to so-called “long tail” or less premium internet properties. At AdTheorent, we only serve ads to properties in our approved list of sites and apps. We have a highly curated publisher list and only serve ads to properties on that curated list. ANA correctly points out that there are exceptions where long-tail properties can drive value. We believe with our ML capabilities we are able to locate valuable campaign conversions from ads served outside of the most popular properties, and this is another advantage for AdTheorent’s statistics-based impression scoring methodology.

AdTheorent also utilizes ML models in advanced ways to filter waste and fraud. Our real-time anti-fraud infrastructure detects fraud before the impression is served on all campaigns across all devices. Additionally, we have several third-party partners; we partner with DoubleVerify for Pre-Bid Filtering and Post-Bid Monitoring, and we utilize IAB’s Bots & Spiders Block List.

Takeaway 5: “Made for Advertising (MFA) Websites Represent 21 percent of Impressions – Indicating Advertisers are Not in Control of Media Placement Decisions”

AdTheorent’s Take: We agree that MFA websites are dubious, and we do not place ads on these sites. AdTheorent prioritizes high-quality inventory over click-bait. Again, this is because our focus is delivering measurable advertiser value, not vanity ad tech metrics like clicks. Again, rather than relying on just an “exclusion list,” AdTheorent runs all campaigns on a platform-wide approved publisher list that is manually reviewed and updated frequently.

Takeaway 6: “Sustainability Efforts Can Be Enhanced by Productive/Non-Wasteful Programmatic Media Buying”

AdTheorent’s Take: AdTheorent’s approach to supply path optimization is designed for efficiency in terms of both cost and infrastructure usage, leveraging a multi-pronged approach to ensure that our programmatic supply is efficient. We work directly with our 20+ SSP partners to ensure that any inventory sent to our DSP generally matches our demand. For example, we throttle down international inventory in countries where our presence is lighter, as well as certain creative sizes where we may have less campaign demand at any given moment.

 We continuously analyze signals in the bid request which indicate whether the impression opportunity is originating directly from the publisher, or whether that impression opportunity is going through multiple different intermediaries. Our machine learning models evaluate on an ongoing basis the optimal and most efficient path to an impression, ultimately ensuring that we are only delivering impressions that are most likely to lead to a client's business outcome, using the minimum infrastructure footprint necessary for the job.

Additionally, 95% of AdTheorent’s cloud infrastructure is powered by renewable energy with the goal to reach 100% by 2025 and we partner with green PMPs with Carbon-Neutral & Climate-Friendly Publishers.

Takeaway 7: “The Previously Identified ‘Unknown Delta’ can be Virtually Eliminated with a Full Path Log-Level Data (LLD) Analysis”

AdTheorent’s Take: AdTheorent’s advertisers know exactly where their media investment is going. Even this critique is cloaked in euphemism. AdTheorent does not engage in kick-backs that divert dollars from working media and we have fully transparent pricing. No DSP offers more pricing transparency than AdTheorent and we fully agree with ANA CEO Bill Dugan’s comment: “where there is mystery there is margin.” Also, as I mentioned before, we pass on any price optimizer savings directly and completely to the customer nor do we have any incentive for the initial price to be high so that we can lower it for the customer and then take a cut. That's not what we do. We believe that there's a huge opportunity in this market for that type of transparency.

I want to thank the ANA for putting a spotlight on these important industry issues. We agree with the faults cited, we believe the ANA study and commentary validates the work we have been doing for 11+ years, and we are excited to continue our work of driving the solution for our ecosystem and industry. Please reach out to share your perspective or with any questions!