Who is bidding more for your attention: Nike or Adidas? Unilever or P&G? Are advertisers paying more to reach readers of the Wall Street Journal, Fox News, or Breitbart? Are any advertisers competing against themselves for the same ad impression by placing redundant bids?

Adalytics attempts to explore these questions through an analysis of programmatic header bidding data from 3,627 different advertisers across 368 different websites. In doing so, Adalytics strives to give consumers and marketers greater transparency into the digital ad tech ecosystem. 

Global spending on programmatic display digital ads (wherein automated technology is used to buy ads) was $127 billion in 2020, accounting for 72% of all digital display ad spend. In theory, this system allows for advertisers to reach the "right audience, in the right context, and at the right time." Programmatic advertising encompasses several modes of purchasing, including open exchange real time bidding, guaranteed direct buying, private exchange buying, and preferred deals.

However, a number of studies have observed that it can be extremely difficult for advertisers to fully audit how their ad dollars are spent. A 2020 analysis PricewaterhouseCoopers conducted with ISBA (Society of British Advertisers) revealed an average of 15% of the money that went into the programmatic supply chain could not be accounted for, and only 12% of all impressions purchased could be matched with a corresponding display website.

The Association of National Advertisers (ANA) published a study in 2017 that noted that many marketers “do not properly understand the economics of programmatic media to manage the complex supply chain that underlies each transaction." Another report noted that CFO's and finance teams often lack granular, itemized receipts or a single source of truth to assess inefficiencies in their ad buy. The UK’s Competition & Markets Authority found that the market for both advertiser & publisher ad serving was one factor in explaining market participants’ lack of transparency. 

This write-up discusses several interesting observations, such as:

1. The New York Times is paying $20-$30 CPMs for ads to serve on other premium news publisher websites such as Vox and Wall Street Journal.
2. Google was observed paying $37.60 CPMs to show ads on an article about how “Google’s plan to get rid of cookies isn’t going well”.
3. A brand (inadvertently) used Google Remarketing to target its own employees with ads across the open web.
4. Market research companies such as Kantar, Nielsen, and Qualtrics are paying up to $10.34 CPMs for media purchased via programmatic ad auctions. Consumer goods companies like Unilever and P&G are paying between $0.01 and $21.56 CPMs for ad slots across many different publisher websites.
5. Advertisers such as Vrbo, Thomson Reuters, Xfinity, and Saks Off 5th were observed submitting large numbers of bids for the same ad auction, potentially overpaying or bidding against themselves due to a lack of ad supply chain optimization.

1. Methodology
2. How much are advertisers paying per ad impression?
3. How much are publishers receiving per ad impression?
4. Are advertisers bidding against themselves?
5. Lack of frequency capping in digital ads?
6. Which advertisers are using Google Remarketing?
7. Conclusion
1. Caveats & limitations
2. Discussion
3. Take away points & recommendations

Methodology

As described previously, Adalytics is running a continuous audience research pilot with a convenience sample of 25 volunteers. The study’s small sample size is not statistically representative, but is meant as a proof-of-concept for the types of insights that could be derived with greater statistical rigor provided more participants. The 25 audience research volunteers were asked to read a study description and fill out a consent form. Participants had the option of self-reporting a number of demographic attributes, such as gender, race, profession, and age.

Once a volunteer had filled out the study consent form, they were asked to use the Adalytics browser extension, which allows them to track the ads they are shown while browsing the internet. The extension allows users to see how much advertisers are bidding to show them ads via programmatic OpenRTB header bidding. The extension is able to capture both winning and non-winning bids submitted to programmatic ad auctions.

The Adalytics extension lets users review their ad history in a personalized dashboard, as well as see how much advertisers are paying per programmatic OpenRTB ad impression in certain contexts. After collecting data for at least two weeks, some study participants received personalized surveys to help contextualize some of the ads they were being shown.

How much are advertisers paying per ad impression?

This audience research study crowd-sourced ad auction bids from 3,627 different brands and advertisers. This dataset of auction bids encompassed 10,640 distinct auctions and 82,224 distinct bids. Of these, 3,543 (4.31%) were bids that actually won a given auction; the other 78,678 bids (95.69%) were non-winning. Some auctions only had one submitted bid, whilst others had up to 131 distinct bids submitted.

Burberry, a British luxury fashion house, was observed paying $51.54 CPMs to show ads on womenshealthmag.com. Google bid $37.60 CPM for an ad on a vox.com article titled “Google’s plan to get rid of cookies isn’t going well”. AT&T’s business service division was observed paying $30.61 CPMs on a slide show on a complex.com article about the “The 40 Most Famous Celebrity Feuds. Procter & Gamble (P&G) bid $21.56 for their “Mr. Clean” brand on a minimalistbaker.com recipe page for Mediterranean baked sweet potatoes. 

The New York Times Company was submitting bids on several of their competitors’ websites, including $25.12 for an ad on Condé Nast-owned newyorker.com, $24.76 CPM on vox.com, $11.20 on businessinsider.com, $10.77 on sfgate.com, $4.77 on G/O Media-owned gizmodo.com, and $4.07 CPMs on theguardian.com.

An aggregate analysis of all bids generated by a given advertiser shows that Google had the highest average CPM bids ($37.60) from all the advertisers observed in this dataset. Procter & Gamble owned 9 Elements (a line of laundry and home care products), had the second highest average CPM bids ($12.13), whilst financial software company Intuit (maker of TurboTax) had the third highest CPM bids ($11.66).

Media companies such as the New York Times (average CPM $5.66), Wall Street Journal (average CPM $0.08), Thomson Reuters (average $0.25), and ESPN (average $1.53) were observed bidding on ad slots on other publishers’ ad inventory. 

In the pharmaceutical sector, Johnson & Johnson (J&J) was observed bidding an average $5.60 CPM for ad slots on websites such as minimalistbaker.com. These ads were for Stelara, a drug treatment for psoriasis and Crohn’s disease. J&J was also bidding $8.50 CPM for another autoimmune treatment, Tremfya, on websites such as businessinsider.com. Regeneron, the manufacturer of an antibody therapy that was used to treat former President Donald Trump, was observed bidding $3.62 CPMs on websites such as two Vietnamese cooking blogs, an Instant Pot recipe website, a personal travel blog, and elle.com. Boehringer-Ingelheim, a German pharma company, was observed bidding an average of $1.61 on websites such as usatoday.com, wired.com, and bjpenn.com. 

Financial services companies such as Chase (average $4.32), MasterCard (average $8.99), and Merrill Edge (average $5.83), the retail banking arm of part of Bank of America, were observed bidding for ad slots on news, entertainment, sports, and niche blogging websites.

Market research firm Qualtrics was observed bidding an average of $4.77 CPM, with a max bid of $10.34. Kantar was observed placing an average of $4.97 CPM bids, whilst Nielsen had an average CPM per bid of $2.11

Consumer goods giants Unilever and P&G were also observed submitting bids for many of their different brands. Knorr, a German food brand owned by Unilever, was observed placing $6.94 CPM average bids, whilst the company’s body wash and soap brand Dove was observed submitting an average of $6.33 CPM bids. Over a dozen different P&G brands were observed submitting ad bids, including cleaning supplies Mr. Clean (average $21.56), toilet paper Charmin (average $13.83), baby diapers Pampers ($3.69) and Pepto Bismol ($1.34).

How much are publishers receiving per ad impression?

This Adalytics audience research study crowd-sourced ad auction bid across 368 different websites that utilize programmatic header bidding. These ranged from high traffic websites such as cnn.com and espn.com, to niche personal blogs such as brainybackpackers.com or vickypham.com. The highest CPM bids were observed on the Express-Zeitung AG owned Swiss news website 20min.ch ($113.32 after conversion from Swiss Francs). In the US, a trio of Hearst Media owned websites — the San Francisco Chronicle, Women’s Health Magazine, and Runner’s World — attracted the highest observed individual bids (above $40.00 CPMs). Various high traffic news sites, such as vox.com, wsj.com, and businessinsider.com, all attracted a few bids over $30.00 CPMs.

The Meredith Corporation owned travelandleisure.com attracted the highest average CPM bids, at $7.31. The investigative journalism newsroom ProPublica attracted the second highest average bids, at $6.00, whilst financial advice and research website dqydj.com (“Don’t Quit Your Day Job”, which is run by working professionals who moonlight as finance bloggers), attracted the highest average CPM bids, at $5.63. 

Other sites in the top 10 by highest average CPM included hungryhuy.com (a Vietnamese cooking site run by an individual blogger), gringajourneys.com (an individual travel blog), and crunchycreamysweet.com (a healthy cooking blog run by a mom of three).

Whereas high traffic major news websites do on occasion receive individual high CPM OpenRTB bids, their average bids are usually lower in magnitude compared to those of high quality bloggers like the four examples provided above. This is likely because major media organizations receive a majority of their ad revenue through non-Real-Time Bidding (RTB) channels, such as direct orders or private marketplace. These premium publishers often reserve a certain portion of their ad slot inventory because they prefer not to sell on the open auctions, as they can demand a premium price from advertisers.

Are advertisers bidding against themselves?

In 1,476 distinct auctions, one advertiser was observed submitting multiple bids with different prices to the same individual ad auction. In some of these cases, there were more bids than distinct ad slots on a given page, which may signify that the given advertiser was inadvertently driving up their own media buying costs by bidding against themselves for a single ad slot.

When an Adalytics user visited a specific webpage in June, the vacation rental company Vrbo submitted 22 bids for the same ad auction, which significantly exceeds the number of ad slots available on that given webpage. In this case, Vrbo was placing bids through three different Supply Side Platforms (SSPs) for a variety of different size ad creatives. The company may be intentionally placing different bids for different creatives, but it is also possible that a lack of information exchange and communication between different ad tech platforms leads to an advertiser effectively double, triple, or quadruple bidding for the same specific ad slot being shown to an individual user in a single ad auction. In this case, the difference between the various CPM bids may have inflated Vrbo’s media buying costs for a 300x250 ad slot by 221%.

In programmatic header bidding, a publisher offers inventory to multiple ad exchanges simultaneously. According to Sellers.guide, the “result is that there is no longer a single SSP or exchange that has exclusivity over one impression. Instead, an impression is represented by numerous ad calls to various SSPs, creating an overhead of ad calls for end buyers.” In other words, the ad buying opportunity sent to exchange A, exchange B, and exchange C are all the same. 

In another instance, the department store chain Saks Off 5th (owned by the Hudson’s Bay Company) was observed submitting 18 bids for the same 728x90 pixel ad creative through Index Exchange (an SSP) for the same ad auction for a page that only has 7 ad slots, potentially inflating their own media buying costs by 41%. The eye glasses retailer LensCrafters was observed submitting 11 bids to one ad auction for the same size creative via 4 SSPs. The difference between the lowest and highest CPM bids may have increased LensCrafters media buying costs by up to 122%.

Media conglomerate Thomson Reuters submitted 25 bids via two SSPs for a single ad auction, where the difference between the highest and lowest CPM bids may have inflated Reuter’s media buying costs by up to 40%.

Lack of frequency capping in digital ads?

A previous Adalytics study noted that, although many digital marketers think an ad should not be shown to a given consumer more than three times a day, there were many observed examples wherein frequency capping was not taking place. This present audience research study extends upon those observations with further details. 

One Adalytics user was targeted with LendingTree ads 26 times across 15 distinct publisher pages on July 3rd. In each case, the ad bids were served via the same SSP, whose cookie should have been tracking the number of impressions served to this individual. This individual reported via a survey that he was not in the market for Lending Tree's financial services anytime in the last few years. 

LCA-Vision, a provider of laser eye surgery, targeted another Adalytics user 25 times on one single page view session. In this example, the publisher had configured auto-ad refreshes, and each of the 25 LCA-Vision ad bids came in through the same SSP. It is unclear whether LCA-Vision was not utilizing frequency capping, or the SSP’s cookie was not properly tracking how many times the given consumer saw the LCA-Vision ads. 

Lastly, Apomeds — a European mail-order pharmacy service — targeted an Adalytics user 47 times across 10 distinct websites. Each time, the ad was served via the same SSP, whose cookie should have been tracking the number of ad impressions served to the user. The given Adalytics user indicated in a survey that he would never consider using an online pharmacy service like Apomeds’.

Which advertisers are using Google Remarketing?

108 different advertisers appear to be using Google Remarketing, which lets them target consumers who have previously visited a brand’s website. Google Display Network offers cross-exchange remarketing campaigns, which can run across inventory from multiple different ad exchanges. It appears that each advertiser is grouped (by either Google or individual third party ad exchanges) into various verticals, such as “Google Remarketing Retail Grouped Brands” for Capital One Shopping.

A wide variety of advertisers appear to be use Remarketing, including ad tech companies such as Facebook, market research firms like Kantar, media organizations like the Wall Street Journal, and political advocacy groups such as stop-republicans.com, dccc.org (the Democratic Congressional Campaign Committee), and nrsc.org  (the National Republican Senatorial Committee).

Some of the ads from one of the aforementioned Google Remarketing utilizing brands were served to two Adalytics users who reported that they are employees of the given brand. In this case, the company was paying for Google Remarketing to effectively follow their own employees across the open web with advertisements. One of these employees was re-targeted with ads 74 times from the given brand, on websites such as reuters.com, bongino.com, and hannity.com. The two Adalytics users confirmed via survey responses that they had no need for their own employer’s product, as they already receive it as part of their jobs.

Conclusion

This study relied on programmatic Open Real Time Bidding header bidding data gathered through a browser extension used by a panel of 25 volunteers. Due to the small sample size, many of the observations presented herein should be considered as exploratory samples rather than representative estimates of advertisers’ bidding. As this rolling Adalytics audience research study grows in number of participants over time, more robust estimates will be possible in the future.

This study relied entirely on programmatic OpenRTB header bidding, which does not show the supply chain costs of various ad tech intermediaries, such as ad exchanges, demand side platforms, or audience data suppliers. Therefore, when the Adalytics browser extension reports, for example, a $5.00 CPM for a given ad impression, the advertiser likely paid more than that (in some cases, twice as much). Therefore, the CPM costs presented herein should be interpreted strictly as the cost of media buying from a specific publisher (i.e., what an advertiser pays to a publisher to show an ad), rather than the full cost incurred by an advertiser to deliver an impression (which would include additional costs from exchanges and audience data vendors). In some cases, this lack of visibility on supply chain costs may explain the differences in CPMs observed in auctions with duplicative bids from a single advertiser — different intermediaries take variable cuts of the advertiser’s payment for a given ad impression, resulting in the publisher receiving different bids.

Furthermore, the prices that the Adalytics browser extension can record are only OpenRTB, which for large premium publishers, primarily manages remnant (low price) ad slots. For sites such as wsj.com, high priced ad inventory is often sold directly and comes without pricing information present on the web page.

Lastly, this study was entirely based on desktop measurements of ad activity, and does not include any mobile or CTV bid data.

Caveats & Limitations

This study offers a proof-of-concept for how crowd-sourced ad impression data, gathered from a panel of volunteers, can offer insights into how much advertisers are paying for individual ad impressions. The study looked at how much pharmaceutical companies were paying to promote drugs on non-medical websites, and how much companies like Google were bidding for publishers to show their ads alongside negative press coverage. The study observed examples where a brand was using Google Remarketing to serve ads to their own employees.

The study identified examples where an advertiser paid for multiple ad impressions in a single page view session, where advertisers may have been inadvertently bidding against themselves, and other examples where frequency capping did not appear to be engaged.

As alluded to in the introduction, a lack of transparency and a single source of truth can make it difficult for advertisers (and corporate finance teams) to fully dissect how their ad budgets are being spent. Previous studies by PwC and ISBA, Digiday UK, and ANA have shown how difficult it can be to account for where every ad dollar is being spent. 

The adoption of header bidding, wherein the same ad impression may be made available for bidding across distinct exchanges simultaneously, necessitates that advertisers look to perform supply chain optimization (SPO). A careful interrogation of DSP, SSP, and ad exchange logs from ad campaigns may enable advertisers to identify the optimal way to allocate their spend.

Furthermore, crowd-sourced data like the ones collected above may offer an orthogonal methodology to understand what is really happening with an advertiser’s campaign. Whilst marketers may rely on ad tech platforms to enforce frequency capping or viewability standards, empirical data shows that those requirements are not always upheld. Identifying a small number of anomalies can be used to diagnose more wide spread patterns in log-level datasets.

Discussion

Take away points & recommendations

1. Advertisers are paying between $0.01 and $113.32 CPMs via programmatic header bidding for media inventory on different websites.
2. In some cases, advertisers appear to be submitting large numbers of bids for the same auction and potentially bidding against themselves, or paying for excessive impressions to be shown to one user.
3. Some advertisers use Google Remarketing to serve ads to their own employees.
4. Crowd sourced header bidding data can help identity anomalies in advertisers’ supply chains. This data can contextualize and help guide analyses of advertisers' private campaign log-level data. If you'd like Adalytics to provide you with a full view of your ad activity, get in touch.