Ad Fraud Researcher
17 min readApr 30, 2021

Impact of Loss of 3P Cookies on Publishers’ Ad Revenue

ABSTRACT

Previous studies that addressed the impact of losing third party (“3P”) cookies on ad revenue did not clearly differentiate between the impact on ad tech intermediaries versus on publishers. While a “dramatic drop” in CPMs was observed from the advertisers’ perspective, that drop cannot be used to represent a decrease in publishers’ ad revenue. This is because the drop in advertiser CPMs came from the elimination of the fees that advertisers paid to ad tech intermediaries for ad targeting rather than changes to the CPMs received by publishers. Publishers did not benefit from any of the ad tech intermediary fees, so the elimination of those fees does not lower publishers’ revenues. The previous studies “dramatically” overstated the negative impact of the loss of 3P cookies on publishers’ ad revenues.

The advertiser CPMs used in the previous studies included unknown amounts of fees paid to ad tech intermediaries for ad targeting, verification, and other services built on 3P cookies. So, instead of “advertiser CPMs” (what advertisers pay) this study uses “media CPMs” (what the publishers get) to better isolate the impact of tracking vs no tracking on publishers. “Media CPMs” were used as input data instead of publishers’ total ad revenues, because said revenues are impacted by other factors beyond the control of publishers — e.g. advertisers’ bidding strategies, mix of browsers visiting their site, and intensity of competition in real-time auctions for ad opportunities, etc.

Direct observations of the unduplicated distribution of media CPM values for pairs of browsers (one tracking browser vs one non-tracking browser) in like environments provided the clearest picture of the effect of “presence of cookie” versus “absence of cookie.” The observed data shows there is virtually no difference in the median (“middle number”) CPMs for tracking vs non-tracking browsers. This means the loss of 3P cookies may lead to modestly lower ad revenues for publishers and certainly not the “dramatic drops” implied by the previous studies.

OBJECTIVE

A literature review by Garrett Johnson tabulated five studies that show the loss of 3P cookies resulting in CPM declines of 50–70%; but it was unclear from those studies whether the lower CPMs resulted in more revenue loss for publishers or for ad tech intermediaries. Further, a study by Acquisti, et al showed that the presence of 3P cookies yielded only a 4% lift in ad revenues for publishers. How can such a large difference be reconciled or explained?

The objective of this study is to isolate the true impact of the loss of 3P cookies on publishers. To accomplish this, the present study will decouple and exclude the many factors that have confounded previous statements and conclusions about revenue loss, specifically conflating lower CPMs as seen by advertisers to mean lower ad revenue for publishers. Doing so neglects the fact that most of the incremental CPM from 3P cookies and tracking accrues to ad tech intermediaries instead of to publishers; and the loss of 3P cookies and tracking means loss of revenue mostly for the ad tech intermediaries instead of publishers.

The hypothesis is that the drop in CPM observed in the five previous studies came mostly from the elimination of fees for ad targeting paid to ad tech intermediaries, instead of from changes in the media CPMs that publishers receive. Therefore the elimination of 3P cookies will have a far smaller impact on publishers than previously implied.

BACKGROUND

Before the methodology of this study can be explained, a bit of background is necessary to illustrate the difference between 1) “advertiser CPMs” paid by advertisers, 2) “CPM+” that accrue to ad tech intermediaries, and 3) “media CPMs” received by publishers. See the diagram below.

Scenario A shows the shortest possible supply path between buyer (advertiser, represented by DSP “demand-side platform”) and seller (publisher, represented by SSP “supply-side platform”). In this scenario, using round numbers to illustrate, when an advertiser bids $1.00 CPM, the DSP extends a bid of $0.80 into the ad exchange, because they keep $0.20. The SSP on the other side of the auction accepts the $0.80 bid, and passes $0.60 through to the publisher, keeping $0.20. For the purposes of this study, the $0.80 accepted by the SSP on behalf of the publisher is the “media CPM” (the value of the ${PRICE_PAID} macro, supplied by the ad exchange) because we do not have further visibility into the terms that different publishers have with different SSPs.

Scenario B shows the addition of value-added services such as targeting and ad verification, enabled by the presence of 3P cookies. Advertisers pay extra for this, usually on a CPM+ basis, which means the fees for these services are added on top of the media CPM, and charged to the advertiser through the DSP. These CPM+ fees accrue to the ad tech intermediaries that provide the services, not to publishers.

Scenario C shows the removal of ad targeting, verification, and other services when 3P cookies go away. The absence of 3P cookies means the fees for these services go away too. The advertiser CPM drops dramatically when the CPM+ fees are eliminated. This was what was observed in the previous studies. The challenge is that those studies could not account for the varying amounts of these fees paid by advertisers; these fees were unknown and not broken out. Further, as is evident in the diagram, the lower CPMs observed by advertisers is not the same as a change in media CPMs, as seen by publishers. In this illustrative example, advertiser CPMs drop from $1.60 to $1.00 from the CPM+ fees going away, while the publisher’s media CPMs do not change.

METHODOLOGY

1. This study will use “media CPMs” (the CPMs publishers receive) instead of “advertiser CPMs” (the CPMs paid by advertisers).

The previous studies correctly reported large drops in CPMs, due to the loss of 3P cookies, from the advertisers’ perspective, using data supplied by Google’s ad platform (the DSP). But they incorrectly concluded that the large drop in advertiser CPMs means a large drop in publisher ad revenues. The loss of 3P cookies for tracking will lead to significant revenue loss for ad tech intermediaries selling value-added services based on 3P cookies, but will not cause significant revenue loss to publishers.

This is why the present study looks at media CPMs, from the publishers’ perspective, instead of advertiser CPMs paid by the advertisers.

2. This study focuses strictly on media CPMs instead of ad revenues of publishers, because publishers’ ad revenues depend on factors beyond their control, like advertisers’ bidding strategies.

This study uses “media CPMs” as input instead of publishers’ ad revenue, because said revenue depends on many external factors that cannot be controlled for — e.g. advertisers’ bidding strategy. It is generally acknowledged that publishers do get more revenue when they allow unbridled tracking versus no tracking. But the decrease in publishers’ ad revenues may be due to the buyers’ bidding strategy — for example, in the absence of 3P cookies, advertisers may bid less frequently and/or may lower their bid CPMs. These bidding strategies are unknown to the publisher, are likely to differ from advertiser to advertiser, and are beyond the control of the publisher.

This is why the present study will use only media CPMs, instead of publisher revenues, to more strictly isolate the true impact of presence of 3P cookies versus their absence.

3. This study compares the media CPMs of pairs of tracking browsers vs tracking prevention browsers, in like environments, to isolate the impact of loss of 3P cookies and tracking for publishers, which is different from the impact for ad tech intermediaries and advertisers.

Finally, to isolate and then quantify the effect of “presence of cookies” versus “absence of cookies,” this study will compare the media CPMs for tracking prevention browsers (Firefox, Safari) versus Chrome (which allows 3P cookies for tracking). The previous studies did not finely separate desktop CPMs from mobile CPMs, which can differ significantly; this study will isolate the effects of environment (desktop vs mobile) by strictly comparing three pairs of tracking vs non-tracking browsers, in like environments: 1) (desktop) Windows-Chrome vs Windows-Firefox, 2) (desktop) Mac-Chrome vs Mac-Safari, and 3) (mobile) Android-Chrome vs iOS-Safari. The study will also exclude ad impressions shown in mobile apps, so only CPMs for ads on desktop web and mobile web will be used for comparison.

This study uses directly observed media CPMs, no regression analysis needed. The values of media CPMs are collected using creative macros, on an impression level, specifically the ${PRICE_PAID} macro on the Xandr (formerly AppNexus) platform. For definition, please see https://invest.xandr.com/docs/creative-macros. The data comes from exchange-wide observations, across all advertisers and publishers. To minimize the effects of long-tail sites and ad fraud, the data excluded impressions shown to bots (visits originating from data centers) and impressions with irrationally low bids (bid prices under $0.10).

HOW TO REPRODUCE THIS STUDY

In order for others to reproduce these experimental results, the following are the specific searches used to isolate browser-operating system combinations. The input data are the values of the ${PRICE_PAID} creative macro from the Xandr ad exchange. A like number of data points was taken for each of the browsers, except that the data on Windows-Firefox was very scarce.

Windows-Chrome

platform:windows browser:chrome custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

Windows-Firefox

platform:windows browser:firefox custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

Mac-Chrome

platform:mac browser:chrome custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

Mac-Safari

platform:mac browser:safari custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

Android-Chrome

platform:android browser:android custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

iOS-Safari

platform:iphone browser:ios custom_BID_PRICE:greaterthan($0.10) custom_PRICE_PAID datacenter:0 -mobile_app

DISCUSSION OF PRIOR RESEARCH

1. Competition and Markets Authority of the UK — A study on the effects of tracking prevention (loss of 3P cookies) due to privacy regulations on publishers’ revenue was conducted by the Competition and Markets Authority of the UK, using data supplied by Google. The study found that “blocking cookies decreases publisher revenue by $0.00132 USD per query (equivalent to $1.32 CPM — cost per thousand) in the short term. This corresponds to approximately 70% of the mean in the control group (implied $1.89 CPM).”

The chart below shows the largest drop in CPMs observed in Chrome (tracking browser) and the smallest drop observed in Firefox (tracking prevention browser). Firefox has had tracking prevention since 2014, while at the time of the study Chrome still allowed 3P cookies for tracking. The loss of 3P cookies in Chrome will lead to the largest drop in CPMs compared to the present. The drop in CPM for Safari, surprisingly, was in between those observed for Chrome and Firefox, even though Safari is a tracking prevention browser and the drop should have been similarly small, like that of Firefox. This unclear result is likely due to commingling CPM prices from Safari on desktop versus mobile. Mobile CPMs are usually higher than desktop CPMs; not separating mobile Safari from desktop Safari muddles the drop in CPM.

The present study does break out iOS-Safari and compares it to Android-Chrome (mobile), while Mac-Safari is compared to Mac-Chrome. A surprisingly large number of people use Chrome on Mac; however Safari on Windows machines is not observed, since it has not been supported by Apple for several years. And use of Chrome on iOS is virtually non-existent. Further note that users of Chrome that have ad blocking and tracking prevention extensions installed are naturally excluded because no ads would be served to them. The data in the present study is collected after the auction is won — with the {$PRICE_PAID} macro.

2. Google — Another widely cited study by Google concluded that “disabling third-party cookies’’ resulted in the following: “average revenue in the treatment group decreased by 52%, with a median per-publisher decline of 64%.” Although the title of the study was “Effect of disabling third-party cookies on publisher revenue” it was based on CPMs in Google’s ad platform — i.e. the DSP data, from the advertisers’ point of view. Disabling access to cookies resulted in the observed “dramatic drops” in advertiser CPMs. However these should NOT be interpreted as a drop in ad revenues for publishers.

This is because advertisers pay for value added services such as targeting and ad verification on a “CPM+” basis — i.e. added on top of the media CPMs. The total CPM paid by the advertiser includes both the media CPM and the CPM+ fees. When 3P cookies go away, the CPM+ component goes away with little to no direct change in the media CPMs that accrue to publishers. This means the large drop in advertiser CPMs came entirely from eliminating the CPM+ fees, not from changes to the publishers’ media CPMs. Further, since different advertisers pay for different amounts of targeting, the CPM+ fees are unknown and may differ widely. Therefore advertisers’ CPM, which include the fees paid for targeting, cannot be used as a proxy for changes to publishers ad revenue when 3P cookies go away.

This study uses only media CPMs as input; and does not use advertisers’ CPMs. for this reason.

3. WFA, ANA, ISBA — Three supply chain studies conducted by industry trade associations have documented that, for every $1.00 advertisers spend buying ads in programmatic channels, approximately 50 cents make it to the publishers for showing ads (see chat below); the rest accrues to ad tech intermediaries selling audience segments, targeting parameters, and ad verification on a CPM+ basis. These studies illustrate the portion of advertiser CPMs that go to ad tech intermediaries instead of to publishers.

Further, the latest study from the ISBA from 2020 also documented that, on average, 1/3 of the dollars that went to ad tech intermediaries “went missing” — 15% of the 50% that went to ad tech intermediaries could not be accounted for. This may not be ad fraud, but it is certainly ad waste in the programmatic supply chain that reduces advertisers’ business outcomes from digital marketing.

4. Pubstack — A study from December 2020 by Pubstack looked at “The Impact of Identity on Publishers’ Advertising Revenues.” Third party cookies are the anonymous identifiers used for ad targeting purposes. The study looked at the effect of the presence or absence of these IDs on publishers’ ad revenues. They specifically quantified that bid rates (bidding frequency) were higher by 2–10% with an ID, but that bid CPMs were virtually unchanged with or without ID. The study concluded that the overall impact of the presence of IDs/cookies on RPM (revenue per thousand auctions) for publishers was a 5–10% increase, driven primarily by higher bidding frequency rather than increases in publishers’ media CPMs. This small increase is similar to the 4% lift found in the Acquisti study, mentioned next. As a refinement to this Pubstack study, the present study uses the change in media CPMs when cookies are present versus when they are absent, thus eliminating the effects of bidding frequency (something that publishers cannot control).

5. Acquisti et. al. — This study leveraged a “dataset of millions of advertising transactions completed across multiple websites owned by a large media company. The average CPM (cost per thousand, paid by the advertising firms) across the sample is $1.14; transactions with a cookie have an average CPM of $1.18, while transactions without a cookie have an average CPM of $0.74 (37% lower).” The study eliminated the effects of geography, device type, and browser, and then isolated two subsets of data: 1) transactions for which the user cookie ID is available, and 2) transactions for which the user cookie ID is not available. The researchers found that “the mean for transactions without cookies is $0.93, while the mean for transactions with cookies is $1.02, a statistically significant difference. The results suggest that when a cookie is available, publisher’s revenue increases by about 4%.” [PDF of study]

STUDY RESULTS

The data of the present study shows that the median CPM for the non-tracking browser (line inside the boxplot) was higher than the media CPM for the corresponding tracking browser and the average CPM (“x” inside the boxplot) was also slightly higher. These are the small differences publishers should expect to see when 3P cookies and tracking goes away, not the 50–70% “dramatic drops” seen in the previous studies. Those drops in CPMs came from the elimination of fees for targeting and other services built on 3P cookies and sold by ad tech intermediaries. The impact on publishers’ media CPMs is far less, as seen here.

When individual pairs of browsers are broken out, the median and mean CPMs for the pairs are very similar with the non-tracking browser slightly higher. Only in the case of iOS-Safari versus Android-Chrome was the non-tracking browser lower. This is consistent publishers’ data which shows Safari gets far lower CPMs than Chrome.

  • Median and mean CPMs for Firefox were both slightly higher than those for Chrome on desktop.
  • Median and mean CPMs for Mac-Safari were higher than Mac-Chrome.
  • Median CPM for iOS were about 25% less than the median CPM for Android-Chrome on mobile

DISCUSSION

The results above appear to support the hypothesis that the loss of tracking from 3P cookies will have a smaller negative impact on media CPMs received by publishers than the 50–70% drops reported in the previous studies. The “dramatic drops” in advertiser CPMs observed in the previous studies came primarily from the loss of CPM+ fees that accrued primarily to ad tech intermediaries not to the publishers. Ad tech intermediaries rarely, if ever, shared any of the CPM+ revenues with publishers; so publishers will also not see a decrease in revenue due to the loss of CPM+ services. Publishers did not benefit from the addition of third party tracking on their sites; publishers added the tracking tags on their sites for free because they were led to believe that they would make more ad revenue with 3P cookies and tracking. Publishers sometimes even paid ad tech intermediaries to add the tracking. It is becoming increasingly clear that this is a net cost to the publisher, and did not lead to higher CPMs or more ad revenue. A survey among publishers revealed that “40% of publishers describe their digital ad revenue as shrinking or static.”

The present study has shown that when the effects of CPM+ pricing, buyers’ bidding strategies, and environment (desktop vs mobile) are removed, the impact of tracking versus no tracking is clarified. Similar to the Acquisti study, by strictly isolating the difference of “presence of 3P cookies” versus “absence of 3P cookies” this study corroborated the relatively small differences — e.g. 9% lower overall for non-tracking browsers. There are no “dramatic drops” in media CPMs for publishers even though there are “dramatic drops” in advertiser CPMs. Those drops came from the elimination of CPM+ that relied on 3P cookies and tracking. The CPM+ accrued entirely to ad tech intermediaries and did not benefit publishers in either CPMs (unit prices) or overall ad revenues.

Publishers may see a temporary decrease in ad revenues when 3P cookies go away, due to the existing bidding strategies of advertisers. Advertisers were conditioned to believe that the loss of tracking means worse outcomes. That’s why advertisers formulated bidding strategies that bid less frequently, stopped bidding, or bid lower CPMs for users with tracking prevention browsers like Firefox and Safari. These bidding strategies are beyond publishers’ control. They also explain why publishers have seen higher ad revenues with the presence of cookies and tracking — advertisers bid more often and higher CPMs in the presence of cookies, when compared to the absence of cookies.

The impact of advertisers’ bidding strategy is most clearly seen in the following chart, isolating one large advertiser doing branding campaigns in programmatic. Their bidding strategy DID NOT artificially penalize tracking prevention browsers like iOS and Firefox with lower CPM and less frequent bids. Even though there was variance in price paid, the data from their vast “run of network” campaign exhibited nearly identical median media CPMs (“price paid”) in each pair of tracking vs no-tracking browsers. This particular advertiser optimizes campaigns based on sales, not vanity metrics such as impressions, clicks, and traffic. They are deliberately not using behavioral targeting or ad verification CPM+ services. They are seeing no negative impact on business outcomes when showing ads to users with tracking prevention browsers, compared to users with browsers that allow 3P cookies.

Further details in the spreadsheet below shows how advertisers’ bidding strategy accounts for nearly 100% of the change in average CPM (price paid). In the columns marked “cookie” the distribution of bids were 62%, 19%, 12% in the 0-$1, $1-$2, and $2-$3 price ranges, respectively. The bids significantly concentrate in the 0-$1 range for the “no cookie” columns — 72%, 20%, 4%, respectively. Advertisers’ bidding strategies reveal that they bid lower or don’t bid at all when “no cookie” is present. This is due to their assumption that lack of cookies means lack of targeting, and lack of targeting will produce worse outcomes. However, that may be a false assumption. Humans use non-tracking browsers like Safari and Firefox. Further, bots don’t pretend to be Safari or Firefox; bots pretend to be Chrome, because they can earn higher CPMs by doing so. Advertisers should realize from this study that targeting non-tracking browsers like Safari and Firefox may yield better outcomes simply by the fact that they are used by humans, even without any targeting parameters.

When 3P cookies and tracking go away in Chrome as well, advertisers will be forced to change their bidding strategies. They will no longer be able to bid lower CPMs or less frequently for browsers with no 3P cookies and tracking — because all browsers will no longer have 3P cookies and tracking services sold by other adtech companies. Advertisers will still spend their digital ad budget. So, over time, publishers should expect to see ad revenue increases when advertisers update their bidding strategy in the absence of 3P cookies. The recent moves by Apple and Google support this trend. Apple is doing away with identifiers in iOS 14; and Google is removing 3P cookies from Chrome by 2022 and eliminating targeting based on cross-site tracking (“behavioral targeting”) in its ad platform. Cross-site tracking relied on 3P cookies. Doing away with 3P cookies and identifiers should lead to increased ad revenue for publishers.

Publishers’ ad revenues will recover, because they will no longer be artificially depressed due to advertisers’ bidding strategies, which is beyond their control. When advertisers adjust to the new cookieless environment and fewer advertisers employ a bidding strategy that bids less frequently and with lower CPMs for non-tracking browsers, publishers’ ad revenues will go back up. When Google does away with behavioral targeting in their ad platform, advertisers will be forced to further wean themselves from the over-use of behavioral targeting (using 3P cookies for cross-site tracking and ad targeting). This reduces the number of ad tech intermediaries selling value-added services on a CPM+ basis, diverting large portions of the advertiser CPMs into their own pockets. More of the advertiser CPMs will flow through to publishers, raising media CPMs (unit prices) as well as overall ad revenues for publishers.

The shift towards better privacy will ultimately benefit publishers, as long as they resist using new forms of tracking that circumvent the privacy protections that did away with 3P cookies in the first place. Publishers also need to urgently educate advertisers that behavioral targeting, hyper targeting, and all the ills that came along with 3P cookies and tracking did not yield the anticipated benefits that were promised by ad tech intermediaries selling CPM+ services based on those 3P cookies for targeting. Doing away with behavioral targeting that was not yielding incremental business outcomes, reducing waste and fraud in the supply chain, and shortening supply paths will lead to dramatic increases in performance and efficiency for advertisers, too.

OTHER RESOURCES

Digiday Research: Most publishers don’t benefit from behavioral ad targeting

https://digiday.com/media/digiday-research-most-publishers-dont-benefit-from-behavioral-ad-targeting/

Behavioral Advertising Might Not Be As Crucial As You Think

https://adage.com/article/datadriven-marketing/behavioral-advertising-crucial/291858

The case against behavioral advertising is stacking up

https://techcrunch.com/2019/01/20/dont-be-creepy/

Behavioral advertising: The mirage built by Google

https://digitalcontentnext.org/blog/2019/06/06/behavioral-advertising-the-mirage-built-by-google/

Targeted ads offer little extra value for online publishers, study suggests

https://techcrunch.com/2019/05/31/targeted-ads-offer-little-extra-value-for-online-publishers-study-suggests/

What The Hell Happened to Good Publishers 1995 2015

https://www.slideshare.net/augustinefou/what-the-hell-happened-to-good-publishers-1995-2015

What Is “Surveillance Capitalism?” And How Did It Hijack the Internet?

https://www.linuxjournal.com/content/what-surveillance-capitalism-and-how-did-it-hijack-internet

Ad Fraud Researcher
Ad Fraud Researcher

Written by Ad Fraud Researcher

Former Group Chief Digital Officer, MIT PhD (Materials Science), McKinsey Alum, http://t.co/E9MgBvoF

No responses yet