Some Second Order Thinking
Est. Reading Time: +40 min
Disclosure: the author and Partnerships managed by New West Capital hold positions in the entities discussed here. The article is meant for entertainment and/or educational purposes and does not constitute investment advice. Particularly due to the binary outcome of some of the positions discussed and the legal nature of the analysis.
- - -
Google, or Alphabet (Ticker: GOOG), is not a company I would typically dig into. There is an army of analyst eyes on it, and it has already achieved behemoth status in the business world, both in size and reputation. But, due a request for my thoughts from a colleague who has a non-trivial sized stake in the company, their recent string of legal headaches warranted a closer look. And after pouring through the suit regarding its ad tech antitrust behavior, I believe there is a 50x – 100x potential upside position sitting in plain sight.
To set the stage, last September the Department of Justice (DOJ) kicked off a trial against Google claiming that they abused their power as a monopoly to dominate the search market. This was the first major monopoly case to make it to trial in decades and the first in the postmodern internet era. Given the prevalence of Google Search, it was about as exciting as these types of legal occurrences get and kicked off a lot of conversations around what this might mean for Google, particularly for those invested in the Company.
And for good reason as Search, with an estimated 90% market share, is the linchpin that catapulted Google to the company it has become. But, thinking the Search case is the biggest risk to the business would be missing a key component of Google’s dominance and that is their ad business. In 2022, Google’s revenue exceeded $280B with about $225B of that (79%) coming from advertising. Year over year, this high contribution percent to revenue from advertising has been ever present.
Advertising revenues have been the economic lifeblood for successfully building an ecosystem around Search and it just so happens to also be in the middle of an antitrust headache of its own.
The DOJ, along with the Attorneys General of California, Colorado, Connecticut, New Jersey, New York, Rhode Island, Tennessee and Virginia are suing Google for violating antitrust laws claiming Google has monopolized the digital advertising market.
This case has far more disruptive potential for Google and, in my opinion, has a higher probability of success for the DOJ than the Search case.
There are a lot of moving pieces and no shortage of complexity in a case of this magnitude, so I will break this down into large material components to arrive at a hypothesis and potential positions to take. These are;
- Timeline and Choice of Legal Venue
- A Word on the Sherman Act
- Overview and History of Google’s Ad Tech Business
- Antitrust Complaints Overview
- What the digital advertising landscape looks like in a DOJ win scenario
- Potential Positions to Take
For those wanting to skip the brain damage and go straight to how to take a position, I’d suggest going straight to the Antitrust Complaints Overview section.
Timeline and Choice of Legal Venue
The official suit, U.S. v. Google LLC, 1:23-cv-00108-LMB-JFA, alleges monopolization of technology advertising markets by Google and was filed on January 24th, 2023, in the Alexandria Division of the United States District Court for the Eastern District of Virginia (EDVA).
This choice of venue is an important feature in the case as it’s famously (or infamously) referred to as the “Rocket Docket.” It has a history of expeditious resolution of cases averaging just over six months from filing to final resolution. Discovery phases typically span four to six months, and extensions for discovery are usually rejected. This reputation has, so far, been maintained in the case of Google with the Court’s denial of Google’s motion to transfer to another court (more on that below) and subsequent ordering of parties to file a proposed discovery schedule within fourteen days of that order.
From a historical perspective, as of 2022, the median time to trial in the EDVA averaged 18.6 months versus the median time across all district courts of 29.8 months. In 2019, before the current but now lessening COVID induced backlog of civil cases the median for the EDVA was even less at 13.4 months.
The implications of this reputation on the case are apparently not lost on Google as it filed a motion to transfer the case to the Southern District of New York on March 14th of 2023. The Court denied this motion citing the “strong public policy interest in expeditious resolution of government antitrust enforcement actions” and reasoning that if the case were to be transferred to New York it would certainly be delayed. The Court went on to further explain that Google’s concerns were outweighed by congressional intent to resolve any antitrust actions with speed and efficiency.
To keep this from turning into a slog of legal jargon, the timeline below will suffice to set the stage of what has occurred thus far. But the important take away is that the selection of the so-call Rocket Docket allows us to put an approximate timeline on the case. With trial set to begin on September 9th, 2024, we can expect a resolution in in Q4 2024 or Q1 2025 to be highly likely.
A Word on the Sherman Act
Before diving into specifics on Google, it is worth touching on the Sherman Act and what is and is not allowed with regards to anti-competitive behavior and monopolies.
Specifically, the Sherman Act outlaws “every contract, combination, or conspiracy in restraint of trade,” and any “monopolization, attempted monopolization, or conspiracy or combination to monopolize.” It is the law passed by Congress in 1890 to prevent the proliferation of monopolies in America.
However, there is an important distinction to be made here and that is that monopolies on their own are not illegal. You can absolutely operate a de-facto monopoly so long as it was arrived at due to your company’s superior product or service despite the attempts by competition to take market share.
What is not permitted is arriving at monopolistic dominance by unfair practices such as fixing prices or rigging bids. A historical example would be Standard Oil lowering prices in new or ancillary markets to bankrupt all local competition. Cushioned by the income from their other markets, they could wait however long for local competition to fold after which they would acquire 100% of the market and raise prices.
So, what is important in the case of Google is not that they have a dominant market share but that this was arrived at and has been maintained via anti-competitive actions.
Overview and History of Google’s Ad Tech Business
To start, the below graphic outlines the process required to buy and sell ads digitally along with Google’s share of the market as defined by impressions (not revenue). This graphic and following description is pulled directly from the DOJ’s suit. As this situation hinges on a legal decision, I’ll stick to defining the business in terms of how the case has defined it.
Source
From the case filing;
“The digital advertising market is divided into three sections: sell-side inventory on the left, buy-side demand on the right, and an ad exchange in the middle. Sell-side inventory is made up of website publishers that that flow to Google’s “DoubleClick for Publishers” Publisher Ad Server, which has >90% of market share. Google AdExchange, which is greater than or equal to 50% of the ad exchange market share, receives bid requests from the publisher ad server, sends them to the buy-side demand, receives bid responses from the buy side demand and sends them back to the publisher ad server. Buy-side demand is made up of advertisers that flow to either: “Google Ads” Advertiser Ad Network (Small and Large Advertisers) which has +/- 80% market share; or Google’s “Display & Video 360” Demand Side Platform (DSP) Agencies/Large Advertisers which has +/-40% share.”
With that basic framework established, we can look at how these platforms came to be and the anticompetitive actions that boosted their success.
An AdTech Business Is Born
Google’s foray into the digital ad space began in 2000 when they launched Google Ads (originally called AdWords), a tool for businesses to have Google search users see advertisements alongside search engine results. This was an innovative, highly targeted, advertising technique and businesses flocked to Google Ads.
A few years following the successful launch of Google Ads, Google wanted to expand into those ads that were placed on third party sites, not just their search page. They created an advertiser ad tech tool for Google Ad customers that wanted to buy space on third-party websites as well and this effectively established them as a middleman on digital advertising transactions beyond their search engine.
If referring back to the infographic from the filing, Google is expanding from right to left in that diagram.
Naturally, moving all the way into the publisher side was the logical aim and Google sought to develop an ad tech tool called a publisher ad server where publishers could manage their online advertising sales. By gaining visibility into, and control over, the publishing side of this advertising landscape Google had access to both sides of the transaction and could effectively take what was an open bidding system and make it a controlled toll way.
This may take time to build up, but with end-to-end influence Google would gain the ability to charge pricing above what can be sustained in a competitive market while still excluding rivals from stepping in.
Returning to the Sherman Act, this is not illegal, so long as this is all accomplished on merit. Up to this point Google has moved into these industries by way of organic business expansion given the incredible usefulness of Search, and this is all to be expected.
However, Google’s publisher ad server failed to gain traction in the industry. Perhaps it was because this deviated too far from and was not reliant on Search, but the reason is not material. What is material was their pivot.
In 2008, Google acquired the market-leading publisher ad server from a firm called DoubleClick for $3.1B. The ad server (“DoubleClick for Publishers” or “DFP”) held 60% of the market share at the time of acquisition. Additionally, it also acquired an ad exchange (“AdX”) in the transaction giving it an auction house for digital advertising space. Following this acquisition, Google’s market share in the ad server/exchange market doubled to about 55%.
Now they occupied meaningful amounts of all three pillars in the infographic above and, most importantly, they could once again leverage the usefulness of Search to drive their success in the publisher side of the industry.
According to the suit, Google created an exclusive link between Google Ads and DFP through the AdX ad exchange. If you were a publisher and you wanted access to Google Ads’ advertising demand, you had to use the DPF ad server and the AdX ad exchange rather than equivalent tools offered by rivals. They were now a buyer, seller and auctioneer of digital display advertising.
This was extremely successful in cutting off any rival publishers’ ability to gain access to Google Ad’s and potentially grow into a threat. With the dominance of ads via Search as more and more businesses moved online it was near impossible for publishing rivals to grow with what was left outside of Search access. According to Google’s own estimates, by 2015 their publisher ad server market share had grown to 90%.
The primary mechanism by which this was accomplished was twofold.
First, Google configured Google Ads to bid on their own AdX exchange in a manner that would increase the pricing of advertising, to the benefit of the publishers and the detriment of the advertising customers. As observed by one Google employee, this was sending a “$3bn yearly check [to publishers] by overcharging our advertisers to ensure we’re strong on the pub[lisher] side.” In the near term, this locked publishers into Google’s publisher ad server with intentionally inflated revenues for certain inventory, at the cost of Google’s advertiser customers. But in the long run this drove out rival publisher ad servers and limited competition in that market. Despite one side of the transaction losing on this setup, Google still made a hefty transaction fee for its position as the middleman via its exchange.
Second, to bolster this middleman positioning, Google implemented a mechanism known as dynamic allocation. This was a means by which Google gave the Google-owned AdX (and only AdX) the opportunity to buy publisher inventory before it was offered to any other ad exchange. It also programmed DFP, its publisher server, to prevent publishers on the platform from offering preferential or even equivalent terms to other ad exchanges. By doing so, and with 60% of the publisher supply within their control at the time of acquisition, Google could decide what ad volume it wanted to allow to be released to rival exchanges. The result was unsurprisingly an increase in the reliance of publishers on Google and a reduction in exchange and publisher ad server competition.
Now, with all these features in place, Google had successfully created end to end control of the digital advertising landscape.
Innovation Steps In, and Capitalism is Brutal
As it tends to happen in capitalism, innovation begins to disrupt your hard-earned cash generating machine and by as early as 2010 other ad tech companies began to recognize that Google’s platforms were not operating in the interests of publishers.
Some companies began offering “yield management” functionality that gave publishers the ability to identify better prices for their inventory outside of Google’s ecosystem in real time. This posed a major threat as a “walled garden” structure that Google had sought to build can only successfully compete on the basis of real time pricing within those walls.
AdMeld, the leading yield management provider at the time, could receive bids in real time across multiple exchanges and demand sources. This allowed other ad exchanges an opportunity to compete in the same way Google’s publisher server allowed AdX to compete through dynamic allocation via a real time bidding standard. Publishers began adopting yield management as it not only gave them access to multiple exchanges, but AdMeld in particular also gave publishers the ability to connect with the advertisers who especially valued their inventory. In contrast, Google only permitted advertiser connections within their own AdX exchange.
In a 2010 strategy discussion, Google executives noted that “Yield Managers are a threat we need to take very seriously” with “AdMeld [being] the largest concern.” Specifically, if AdMeld’s success continued, Google worried about having to “pass real-time AdX pricing into a non-DFP ad server.” So, in 2011, Google acquired AdMeld and shut down the real-time bidding technology portion of the business.
Not long after this, in 2013, Google once again changed the structure of its AdX exchange through a program called Dynamic Revenue Share in which Google Ads took its revenue share fee from a fixed 14% take rate on each impression to a floating fee that averaged 14% per publisher over time. On the publisher side, this doesn’t necessarily change much, over time, but for Google it allowed for strategic overbidding for higher value auctions (by accepting a lower fee) and underbidding on lower value auctions. This change in and of itself is not anticompetitive, any individual or business should be able to pay more or less for something based on their own perceived value, but it can be abused.
Later in 2013 Google altered this change a bit more by subsidizing bids, or bidding well above the advertiser’s willingness to pay, on the competitive auctions and sacrificing any profit whatsoever. To maintain margins, any losses shouldered by the subsidized bids were offset by charging much higher fee (usually over 50%) on the auctions where Google Ads faced no competition (i.e., the majority of the auctions Google had already been winning). This was internally dubbed Project Bernanke after former Federal Reserve Chairman Ben Bernanke because it resembled “quantitative easing on the Ad Exchange” or helicopter money for auctions Google deemed favorable. It also pushed Dynamic Revenue closer towards the practices of Standard Oil mentioned above.
In practice it would play out as follows;
From the case filing and Google generated internal documentation –
“After running its internal auction (as described above), Google Ads calculates its two highest bids on a CPM (cost per click) basis as $1.00 and $0.96 (the grey bars). These bids might be similar because they are based on the same Google targeting data. Applying a uniform 14% take rate (or “margin”) would result in bids equal to $0.86 CPM and $0.83 CPM. With dynamic revenue share, Google adjusted the bids to $0.95 CPM and $0.83 CPM (the red bars in the Figure). For Bernanke, Google raised the first bid even further (sometimes substantially), as the first bid determines the winner of the auction. By raising the first bid (here from $0.95 CPM to $1.20 CPM, the green bars), Google Ads won more auctions, either clearing publishers’ reserve price more often or winning against a rival’s bid for competitive impressions.”
Where an auction was not competitive, the majority of the Google Ads auctions, the lower price was the one Google Ads paid for the impression. Google then kept the margin (50% in the above example) to subsidize the competitive auctions. So instead of the website publisher receiving $0.83 CPM for the ad, it received only $0.48 CPM with Project Bernanke. While Google Ads’ advertisers won some competitive impressions, they did not receive the full benefit of the lower prices Goggle Ads paid for the less competitive inventory.
In 2014, Google recalibrated Project Bernanke to decrease Google Ads’ bids on AdX for publishers that allowed rivals an opportunity to buy inventory ahead of AdX. This tweak was called “Project Bell” and operated without any input from or awareness of the advertiser (i.e., automatic opt-in) reducing bids by about 20%.
Header Bidding, the Largest “Existential Threat”
By 2015 the industry had created another attempt at circumventing the walled garden Google had built with something called header bidding. It worked to circumvent the AdX exchange by placing code into the “header” section of the HTML code of a web page. As the page loaded the code would trigger a real time auction among ad exchanges before the publisher ad server was called. The highest bid from the header bidding auction was then sent to the publisher’s ad server. Because of the manner in which Google had configured DFP, and the last look advantage dynamic pricing gave AdX, the Google ad exchange received this winning bid and had to see if it could beat that price.
This effectively forced Google’s ad exchange to compete with other ad exchanges and against real time pricing rather than against historical average prices. In assessing this impact, an internal Google presentation noted “header bidding and header wrappers are BETTER that [Google’s platforms] for buyers and sellers” and went on to explain that the competition this created between AdX and buyers increased publisher revenues by 30% to 40%, and provided additional transparency.
What’s most notable is that header bidding, due to its application and deployment, was far more cumbersome to use and negatively affected a publisher’s webpage performance. Despite this, it continued to siphon volume and Google’s own internal analysis showed “header bidding removed AdX inventory exclusivity...[and] buyers shift[ed] spend as other inventory sources delivered equal/better value.”
Structurally, the headaches caused by yield management solutions had returned in the form of header bidding.
Realizing that it was unlikely going to be able to return to a fully close loop system, Google developed its own mechanism called “Exchange Bidding” which was later renamed to “Open Bidding” and formally launched in April of 2018. On the surface it looked like a publisher friendly tool that substantially reduced the additional headache of implementing header bidding and allowed to share with other exchanges that agreed to work with Open Bidding the last look advantage over header bidding traffic that dynamic allocation had previously reserved for AdX. But it was a bit of a Trojan Horse and came with drawbacks for publishers and participating exchanges.
First, Google required a 5% fee on any transaction won by a rival exchange through Open Bidding. This effectively lowers the net bid of all non-Google participants by 5% and amounted to a 25% increase (5% over Google’s 20%) in the rival ad exchange’s fee making it far less attractive to publishers.
Second, Google remained the entity of remittance and even if a rival exchange won a bid, they paid Google and Google in turn paid the publisher. On the publisher side, should a rival exchange provide any repeatable value it would be unknown to the publisher.
Third, if another exchange owned an advertiser buying tool (as Google did) and had a presence in the right portion of the industry infographic from above, that exchange was not permitted to allow its advertiser buying tool to participate. From the perspective of a publisher, not being able to bring the counterparty demand would decrease your exchange’s value proposition. The best brokers bring both sides of the transaction after all.
Forth, in Open Bidding, Google was able to obtain the data of the rival bids for each impression which would not have been possible through header bidding. Competing exchanges that opted into Open Bidding were not given the same luxury and they had to share their bid with Google with no reciprocity.
From internal emails, a Google partnership strategist phrased this approach as aiming to “dry out” rival ad exchanges that adopted header bidding. The idea was that with the in-place dominance of +90% of the publisher side and +50% of the exchange traffic, there would be enough responses along the lines of “if you can’t beat them, join them” that it would be sufficient to snuff out any competition from reaching a critical mass to compete.
But there existed two other sufficiently large and controlled sources of sell side demand, the opposite side of the transaction, that could still potentially provide enough of a foothold to keep header bidding progressing and those were Facebook (now Meta) and Amazon.
Facebook had at one point attempted to build its own advertising stack with a publisher ad server acquisition in 2013 that was later shuttered having recognized that any full-stack ad tech strategy “is subject to one bottleneck and intermediary-Google. They ‘own’ the Ad Server, and hence the last mile relationship with publishers.” But when head bidding arrived Facebook, with their billions of users, was all too happy to begin using and testing the approach. As was Amazon.
In September of 2016, ahead of Open Bidding’s launch, Google began working on a strategy to bring the Facebook demand into Open Bidding as a better approach to slow adoption of heading bidding than “[a]ggressively mak[ing] [Open Bidding] much better than [header bidding].” As product leadership recommended to CEO Sundar Pichai, with a Google-Facebook deal, “[f]or web inventory, we will move [FAN’s] demand off of header bidding set up and further weaken the header bidding narrative in the marketplace.” Facebook was a bit more direct in their assessment of the deal, stating “What Google wants: To kill header bidding (us baptizing [Open Bidding] will help significantly).”
In September 2018, shortly after Open Bidding’s launch, Google and Facebook entered into a “Network Bidding Agreement” (“NBA”). The deal provided Facebook with unique terms not offered to anyone else. It included a contractual promise of no last look by AdX and direct remittance to publishes allowing Facebook to maintain its publisher-facing relationships. In exchange, Facebook committed to a minimum annual spend on Open Bidding and was incentivized through a tiered volume discount to shift spend away from possible header bidding. Facebook’s then-VP of Partnerships noted that “by doing this deal, we will cement [Google’s] position of power.”
Amazon, for their part, rebuffed offers from Google and continued to develop their own header bidding solution.
But, like Facebook and Amazon, Google had a large amount of buy side control with Display and Video 360 (“DV360”) controlling nearly 30% of gross digital advertising revenues when header bidding appeared in 2016. In conjunction with the Open Bidding efforts, they did not leave this lever untouched for addressing the risk header bidding posed.
DV360 tended to cater to larger, more sophisticated advertisers and ad agencies and as such Google had permitted greater control over where and how they bought advertising inventory. Due to this, more than half of the spend by DV360 advertisers went to rival exchanges by 2017 leading up to the launch of Open Bidding and DV360 advertisers often represented the largest buyers on rival exchanges utilizing header bidding.
The initial strategy was similar to previous changes made to AdX to prohibit utilizing other exchanges and a whitelist was created of those exchanges not suspected of implementing header bidding that DV360 clients may also transact with. But, given the nature and sophistication of the DV360 clients, internal experiments testing such a strategy showed the DV360 buying tool would lose approximately 30% of both impression and revenue as clients were highly likely to leave.
An alternative solution was adopted to target header bidding while minimizing damage to Google: “instead of stop bidding on HB [header bidding] queries, we could bid lower on HB queries.” When combined with the advantages inherent to its AdX exchange by way of its publisher ad server, this not only would allow Google to inhibit header bidding transactions but also allow for the redirecting of revenues back to Google’s ad exchange.
This project was dubbed Project Poirot and worked by lowering all DV360 bids to rival ad exchanges that had ceased to use second-price auctions, a proxy for header bidding usage. For each rival exchange, Google would reduce all bids to that exchange by a certain percent. This manipulation would ensure that Google’s AdX exchange would win the relevant auction due to the decreased bids supplied to rival exchanges and because AdX did not take part in header bidding.
In practice it would look like this;
From the filing -
“An advertiser using DV360 might configure an ad campaign to pay a maximum price (for example, $1 CPM) for a particular type of advertising impression. Under Poirot, Google would lower that maximum bid when bidding on rival ad exchanges that used head bidding by applying an ad exchange-level multiplier, for example, bidding $0.38, $0.42, and $0.40 on three rival ad exchanges. In this example, the $0.42 bid is the highest and wins the header bidding auction; that bid is then passed to the publisher’s ad server after the ad exchange deducts its revenue share fee (assumed here to be 15%, reducing the net bid to $0.36). Next, the ad server sends that price along with a request for a bid to Google’s ad exchange via dynamic allocation. The $0.36 serves as a price floor and is shared with bidders on the ad exchange. On behalf of the same advertiser as before, DV360 now bids the advertiser’s maximum bid ($1) and wins the impression. Because Google’s ad exchange ran a second-price auction, however, the publisher receives only the floor price, $0.36. Google charges the advertiser this price plus the applicable ad exchange revenue share (20%), which translates to $0.45. Ultimately, this means the advertiser pays more for the impression than it would have paid bidding via a rival ad exchange, Google is able to profit by extracting its revenue share fee at the ad exchange level, and the rival ad exchange that otherwise would have won (because it would have charged a smaller revenue share fee than Google’s ad exchange for the same DV360 bid) is denied the transaction.”
In prelaunch experiments Google found that Project Poirot would reduce the publisher side DV360 revenue by over 10% by the process outlined in the above example. On the other side, the predicted additional revenue generated by Google and advertisers was only about 1% and the total number of impressions DV360 purchased would drop by roughly 5%.
Project Poirot was formally launched in July 2017 ahead of the Open Bidding release under the name Optimized Fixed CPM Bidding and was released as an opt out update. Meaning no information was supplied to its advertisers regarding the nature of the program and clients had to submit a request to not be subject to Optimized Fixed CPM Bidding. The result was that 99% of all automated bidding campaigns on DV360 incorporated Poirot automatically.
The initial roll out of Poirot in 2017 was “quite effective, resulting in [DV360] spending 7% more on AdX and reducing spend on most other ad exchanges” noted Google’s Director of Product Management. It shifted roughly $200M of DV360 advertiser spend away from rival ad exchanges towards Google’s. With this success, it launched “Poirot 2.0” in September 2018 (the same month the Facebook deal was inked) which was simply a further decrease in bids made to rival exchanges by up to 90%.
In the year following, the most active exchanges using header bidding had pronounced drops in advertiser spend via DV360. AppNexus/Xandr lost 31% of DV360 advertiser spend, Rubicon lost 22%, OpenX lost 42% and Pubmatic lost 26%. On the other hand, Google’s exchange saw an increase in revenues from DV360 advertisers from approximately 40% of overall spend to 70% due to Project Poirot despite having far higher fees than all the above listed exchanges.
While the success of this project, along with keeping at least Facebook’s demand on Open Bidding, was apparent there was an additional avenue Google identified that might keep rival exchanges sustained, and that was publisher pricing controls.
Controls built into the publisher ad server allowed publishers to set different price floors for different exchanges or advertiser buying tools. Previously this had been dealt with via the AdX ban on offering preferential terms to competing exchanges that was put in place shortly after the DoubleClick acquisition. But, as an “equal footing” contractual provision and not a hard coded feature, this had become difficult to monitor and enforce in the wake of Open Bidding.
In practice this behavior would look as follows; “a publisher could set a price floor of $2 for Google’s ad exchange and $1.80 for the competing ad exchange OpenX. If OpenX submitted a winning bid of $1.85, and Google’s ad exchange had a buyer willing to pay $1.90, the inventory would still be sold to the OpenX advertiser because Google’s AdX ad exchange failed to clear the minimum price set by the publisher for AdX.”
Google’s internal audit of the situation found that while its AdX ad exchange was winning 66% of auctions where header bidding is being utilized, for the amount it did not win just under half of those instances AdX had a higher bid than the rival ad exchange that did win.
In March 2019, Google announced a number of updates to the publisher server and ad exchange. Within these changes was the removal of granular publisher price controls and in their place, Google implemented Unified Pricing Rules or a single price floor for all ad exchanges and advertiser buying tools. Following the launch and full ramp up of Unified Pricing Rules the market share of Google’s ad exchange increased approximately 6% in 2019.
Notably when Unified Pricing Rules were released Google also ceded one of its biggest advantages, last look. Google Engineering Director explained that Google “paired this change [dropping last look] with other benefits to Google (fair access and uniform reserve prices), rather than being forced by regulators to remove last look under disadvantageous terms.” While technically correct in saying they were being forthcoming, what is also technically correct is this is far less useful under unified pricing anyhow.
Google’s response was to replace last look with a predictive model, “Smart Bidding,” that would estimate the distribution of bids of competing exchanges and bid accordingly. While this would be a fair and open market practice normally, the years of dominance and entrenched position of all aspects of the industry meant they had exponentially more data to work with than any competitor. When Smart Bidding launched, it resulted in an increase in revenues of 3% versus the expected drop of 30% and 10% in Google Ads and DV360 respectively.
To head off any complaints regarding data access, Google began sharing data with rival exchanges but with a few limitations. Only those exchanges that bid on Open Bidding (i.e., no header bidding) and data is limited to just the winning price (if the rival lost) or the second highest bid (if the rival won). Keep in mind Google continues to have access to broad pricing data via their dominance of the publisher side of the industry.
Antitrust Complaints and Analysis
If you’ve read everything up to this point my hat’s off to you and the analysis portion is hopefully a bit easier to navigate and get through.
Whether you feel Google is a bad actor or in the right is immaterial to an investment case. This hinges on whether the DOJ and collection of states that have filed suit have a legal basis to their claim.
We’ll take a look at each complaint, assess whether there is merit based on Federal Trade Commission definitions of antitrust infringement, and assign a probability of a Google loss. Taken together that will give a rough estimate of how likely an unfavorable outcome may be for Google.
The suit alleges that all 309 paragraphs in the suit (yes, 309) constitute the allegations.
But specifically, those claims are the following antitrust violations;
1. Google’s acquisition of DoubleClick to obtain not only a dominant publisher ad server, DFP, but also a nascent ad exchange, AdX, in order to pursue its goal of dominance across the entire ad tech stack;
2. Google’s restriction of Google Ad’s advertiser demand exclusively to AdX;
3. Google’s restriction of effective real-time access to AdX exclusively to DFP;
4. Google’s limitation of dynamic allocation bidding techniques exclusively to AdX;
5. Google’s providing AdX with a “last look” auction advantage over rival exchanges;
6. Google’s acquisition of Admeld to stop its yield management technology from promoting multi-homing across ad exchanges;
7. Google’s use of Project Bell, which lowered, without advertisers’ permission, bids to publishers who dared partner with Google’s competitors;
8. Google’s deployment of sell-side Dynamic Revenue Share to manipulate auction bids-again, without publishers’ knowledge-to advantage AdX;
9. Google’s use of Project Poirot to thwart the competitive threat of header bidding by secretly and artificially manipulating DV360’s advertiser bids on rival ad exchanges using header bidding in order to ensure transactions were won by Googe’s AdX; and
10. Google’s veiled introduction of so-called Unified Pricing Rules that took away publishers’ power to transact with rival ad exchanges at certain prices.
Alleged Violation #1 – The DoubleClick Merger
Section 7 of the Clayton Act prohibits mergers and acquisitions when the effect “may be substantially to lessen competition, or tend to create a monopoly.” This is forward looking and while it can be argued that the DoubleClick merger was the linchpin of everything to come, at the time the two companies were not horizontal competitors but rather occupants of two separate spaces within a broader industry.
As so much has happened since this acquisition, I find it doubtful it will be cited as a violation but rather something that likely should’ve been scrutinized more at the time. Perhaps altering the acquisition to exclude the ad exchange portion or similar.
Probability of being material – 0%
Alleged Violation #2 – Restriction of Google Ads to just AdX
Alleged Violation #3 – Restriction of real time access exclusively to DFP
Alleged Violation #4 – Dynamic Allocation Bidding
Alleged Violation #5 – Last Look Usage on AdX
For Alleged Violations 2-5, a suitable corollary to this might be when Microsoft came under antitrust review for the inclusion of its Internet Explorer on computers as default. As Search resembles something akin to an indispensably valuable product like Microsoft’s operating system, perhaps this behavior will also draw antitrust scrutiny.
In the Microsoft case, a monopoly was found to exist due to Microsoft’s dominant operating system presence and the default inclusion of, and the difficulty to switch from, its internet browser kept competing firms from developing alternatives. In that case, not all routes were shut to a user switching systems, but it was sufficiently difficult to have courts force Microsoft to end certain practices.
In this instance it is a judgement call regarding Market Power and the level of Exclusionary Conduct but the outright requirement to use Google’s AdX exchange to access Google Ads is an even tighter constructed use-path than the Microsoft example.
This might fall under the Exclusive Dealing or Requirements Contracts or Refusal to Deal definitions of the Sherman Act.
In the former, exclusive dealing or requirements contracts between two parties are common and generally lawful. In simple terms, an exclusive dealing contract prevents a distributor from selling the products of a different manufacturer, and a requirements contract prevents a manufacturer from buying inputs from a different supplier. These arrangements are judged under a rule of reason.
They cross the line of anticompetitive when a firm with market power uses these types of arrangements to keep smaller competitors from succeeding in the marketplace. As an example, the Federal Trade Commission found in 2015 that a manufacturer of pipe fittings unlawfully maintained a monopoly in domestically produced ductile iron fittings by precluding the purchase of any non-domestic competitor products, who were attempting to enter the domestic market. A very similar scenario to Google’s mandate to use their AdX exchange. Albeit less complicated than ad tech.
On the latter, any business – even a monopoly – may choose its business partners. However, it becomes a Refusal to Deal matter for a firm with sufficient market power and the focus becomes whether the refusal helps the monopolist maintain its monopoly, or “allows the monopolist to use its monopoly in one market to attempt to monopolize another market.”
Through either an Exclusive Dealing or Requirements or a Refusal to Deal lens, I would think the argument would hinge upon how much power Google Ads had in the ad tech marketplace at the time of the exclusivity requirement’s implementation since both take Market Power into account.
That is unfortunately subjective once in court, but we can use a back of napkin data point to get a sense. Here is the historical online advertising revenue versus what Google Adwords brought in.
As US online advertising spend is a far more reliable metric during this timeframe, I’ll adjust these figures for their US based advertising revenue.
For the years of 2005, 2006, and 2007 the total spend of all US online advertising was $12.5B, $16.9B and $21.2B respectively. This would put Google’s US based revenue at 29.6% ($3.7B), 35.4% ($5.98) and 40.2% ($8.53B) of the US online market during those same years. Which I would qualify as being sufficiently market dominant and, if driving nearly 40% of US online advertising spend prior to the DoubleClick acquisition, their subsequent requirement to use their exchange would force me to fall in line if I were a publisher.
Additionally, if going back to the Microsoft example, in that case Microsoft did not fully preclude a user from going to another internet browser but simply made it difficult enough to have the FTC force a change of behavior. In this case Google built the system (allegedly) as to mechanically preclude rivals from growing not by the merit of Google’s solution but by wielding their market share to limit any behavior by their customers that would allow rivals to gain market share.
Probability of being material – 30%
Alleged Violation #6 – Acquisition of AdMeld
Similar to the acquisition of DoubleClick, there is likely too much about this claim that will hinge on subjective opinion such as how much of a threat yield management would or would not have posed had Google not shuttered the technology. For instance, a claim could be made that if it were truly as disruptive as the suit alleges, then why did no additional material players follow AdMeld’s lead.
Unlike header bidding, where the tide was so great Google was relegated to working within/around it, yield management may look like a flash in the pan if Google truly stamped it out with a single acquisition. For conservatism’s sake, this Alleged Violation is probably best thrown out of the analysis.
Probability of being material – 0%
Alleged Violation #7 – Project Bell (and Project Bernanke)
Alleged Violation #8 – Dynamic Revenue Share
While the Alleged Violations 2-5 seem pretty damning, during the building phase of any business there is a lot of opaqueness around what did happen compared to the nearly infinite paths that could have happened. Hence it is always difficult to make these claims during the highly competitive phase of the business cycle and the lower expected material probability of those Alleged Violations than it may seem they deserve is a form of conservatism in this regard.
However, Alleged Violations 7 & 8 introduce a unique aspect that could tip the scales towards the ire of antitrust officials and that is potential violations of the Robinson-Patman Act that pertains to Price Discrimination in antitrust cases.
While proof of a violation involves complex legal questions, illegal practices under the Act include;
Below-cost sales by a firm that charges higher prices in different localities, and that has a plan of recoupment;
Price differences in the sale of identical goods that cannot be justified on the basis of cost savings or meeting a competitor’s prices; or
Promotional allowance or services that are not proactively available to all customers on proportionately equal terms.
If Alleged Violations 2-5 are unsuccessful in gaining traction, possibly due to the ambiguity of what the early online advertising market would grow to become, these fall under a similar argument with the added layer of pricing differentials across customers. True, Alleged Violation 4 supposedly subsidized publisher profits in the early days at the expense of advertisers, but proving that was not product quality driven may be difficult.
By comparison, this scheme was aimed at exchange volume specifically and is more akin to taking a revenue hit in one market to gain local dominance and making it up with super normal profit in another in which dominance is already established. Using another past antitrust example, that behavior was one of the key reasons Standard Oil was broken up.
Probability of being material – 55%
Alleged Violation #9 – Project Poirot
Alleged Violation #10 – Unified Pricing Rules
These final two are, in my opinion, the most material. Not from an operational standpoint as the Alleged Violations leading up to these are the very necessary steps to arrive at a monopoly but by the time these two occurred it is easy to establish that Google was and is in fact a de facto monopoly in the online advertising space.
The risk in antitrust cases is that, while the monopoly point is being reached, it is hard to prove whether or not the monopolist’s success is due to “the willful acquisition or maintenance of that power as distinguished from growth or development as a consequent of a superior product, business acumen, or historical accident.”
Given the short history of the online advertising industry, the complex nature, and sheer volume of transactions and data, even if one reads through the behavior to this point and feels that they acted in an anti-competitive nature the task still remains to prove they did not arrive at this point due to the reasons listed above, and that is a risk if taking the plaintiff’s side. Or if investing/trading according to it.
However, by this point in the timeline Google occupied over 90% of the publisher side, over 50% of the exchange volumes, and a significant portion of the advertiser side. While the last portion is a bit harder to define, and Google has attempted to define it as anything up to and including what happens in individual apps, we can go back to a simple back of the napkin gut check as before.
Using the year before Open Bidding and the deal with Facebook was reached, 2017, global online advertising spend was $229.25B (emarketer) versus Google’s advertising revenue of $95.375B, or 41.6% of total.
Looking ahead to more recent numbers, and it would seem Google’s market dominance has remained intact.
For the years 2019, 2020 and 2021 Google’s worldwide online advertising brought in $134.8B, $146.9B and $209.5B respectively. Comparing Google to the overall global online advertising marketplace spend of $522.5B in 2021, Google still holds a 40.1% global market share of online advertising. On a US online advertising basis this is 50.9%.
The suit references +80% in the space that Google Ads occupies and +40% in the space pertaining to Display & Video 360 but given the above I believe it is safe to claim that no matter how you divide up the advertising ecosystem, Google controls an amount that would quality as Market Power, per the Clayton Act.
So, regardless of whether the behavior along the way to this market influencing level is thrown out, these two Alleged Violations occurred to maintain market dominance. And this is a bit easier to make a case against and perhaps why most antitrust cases succeed at breaking up monopolies rather than preventing them. Although the latter cannot be proven or disproven.
Additionally, Project Poirot brings in, similar to Alleged Violations 7 & 8, potential violations of the Robinson-Patman Act as it in effect is altering pricing based on market and not product or service.
With market dominance clearly established, and questionable practices to maintain and further that dominance combined with instances of measurable effects and staffing reductions from competition immediately after these changes, I believe this is where the DOJ will spend the most time and get the most traction if any.
Probability of being material – 70%
- - -
This is, admittedly, not a scientific approach. But, should any grouping of the above Alleged Violations be material, we’ll prescribe a simple 50% chance that the DOJ and participating states get it over the finish line.
Taken together, the weighted average of these probabilities would put us at a 13.95% chance of a DOJ win.
A word on antitrust atmosphere during this trial...
I believe this may be an overly conservative assumption given the current climate for antitrust cases. For those who followed the Spirit Airlines and JetBlue case, it is clear that the current administration is drawing a harder line on corporate issues. With a presidential election occurring two months after this trial is set to begin, I am of the opinion that it is more likely than not that this may serve as a talking point for the in-place party. Additionally, an outright monopoly violation ruling followed by a company break up and hefty fines are not necessary to profit should the DOJ prove successful, which I’ll touch on. A simple change of behavior such as in the Microsoft example above would be sufficient which also probably makes the above assumption on the conservative side.
A DOJ Win – How Ad Tech Would Change
While this may read like a short analysis on Google, it’s not. And I don’t really ever short a company. Afterall, following the breakup of Standard Oil the constituent parts became worth more than the previously combined whole. There is a scenario where one shorts Google aiming for a DOJ win, it comes to pass, and you still lose.
The way I think about this is like the releasing of a pressure value.
Google has controlled three defined sections of an industry (the publisher ad server, exchange and advertiser demand components) since its inception and growth into maturity phase. Should this be broken up, or even forced to change their behavior, there will be a shift within each of these pillars accordingly.
On the advertiser or demand side, the only two publicly traded material competitors are The TradeDesk (Ticker: TTD) and Viant (DSP). This side of the market has a broader spectrum of competition, and is the section where Google has the least amount of clout. Competitors report similar margins to Google’s 20% with most estimates coming in about 15%. Given this, and Google’s legitimately wonderful businesses like Search and YouTube that help drive this side of their dominance, it is likely the area of least “disruptive response” to a Google loss scenario.
On the other side, the publisher or supply side of the market has been whittled down to just Magnite (Ticker: MGNI), Xander (a Microsoft product), PubMatic (Ticker: PUBM), Nexxen (Ticker: NEXN) and a slew of privately held players. This is where Google has the greatest market share at 90% and the area where the suit focuses and alleges the most violations. This area of the industry has also been the beneficiary of the most aggressive tactics employed by Google to grow and maintain dominance. Margins here are lower hovering around the 10%-15% range and subsequently these companies receive lower multiples than the demand side players like TTD and DSP. Given Google’s dominance and behavior towards competition in this space, it is fair to assume that if this overhang were removed these companies should see better margins and thus better multiples.
Regarding these two areas of the market, in the pressure valve release metaphor the publisher or supply side would see the most meaningful shift in business prospects should Google be forced to compete on merit as the few remaining players have had evolve to do.
But should Google be forced to decouple its ad tech stack, what tailwinds might we be able to reasonable expect for the publisher and exchange side?
For fiscal year 2023, the Google Network revenue was $31.5B. They unfortunately don’t break out the distribution amongst their ad tech stack components but given the requisite market share and expert estimated total available market share of the constituent pieces a 60/40 split between the supply and demand side is a reasonably accurate estimation. Using that we can expect that Google will generate $18.9B from the ad server and exchange side of the business with the advertiser or supply side making up the balance.
With the four publicly traded competitors of MGNI, Xander (although this is Microsoft), PUBM and NEXN generating about $1.5B collectively on a trailing twelve month basis (Q3 2023), there would be significant revenue increases up for grabs should Google cede any market share.
Below is a table summarizing the other players, their revenues, and approximate market share with and without Google present.
As the DOJ defines Google’s 90% market share in the publisher space as impression based, not revenue, we must adjust for this. Total expected programmatic ad spend (ex Search) for the 2024 calendar year of $120B was used, rounded down from an expected $121.9B per eMarketer. This was then adjusted to a TAM of $23.1B using a blended take rate between Google’s 20% and 12.5% to reflect the 10%-15% range of competitors stated earlier and using the impression-based weight of 90% for Google.
As a pro rata redistribution of Google’s market share is unlikely to play out in the real world, the final line is the expected redistribution of Google’s market share on a 30/70 basis. That is, for each dollar that is up for grabs, the “All Else” category only converts at 70% what would be a strict pro rata redistribution. This is to approximate the fact that this long tail of privately held companies are less dominate than the publicly listed competitors here and are unlikely to convert at the same rate.
Going forward, we will only consider Magnite, PubMatic and NEXN as Xander sits within Microsoft and the remaining are not publicly traded and/or do not have enough concentration of revenues in the supply side to be considered attractive in a Google loss scenario.
Potential Positions to Take
Before considering any potential positions, we’ll start by estimating a few scenarios according to that redistribution analysis above. As Google will not completely vanish from the market, even if forced to divest by the DOJ, the weights above will guide various revenue outcomes for each company should Google lose 10%, 15%, 20%, etc. market share to competition in a DOJ win scenario.
For this we’ll use market share loss levels of 10%, 20% and 30% coming from Google and going to competition at the handicapped redistribution rates above.
Considering how scrappy the remaining competition has had to become, should Google be forced to compete on merit even the high end of 30% market share loss may prove to be conservative. This would still have Google control 60% of the market as measured by impressions. Still above the 50% monopolistic level referenced in the Clayton Act fact pages. But, conservatism is advisable in these types of analysis, so we’ll stick with that.
Below are the new revenue expectations according to the above assumptions.
For a valuation analysis, we’ll take a Sales and EBITDA multiple approach. I typically side with the late Charlie Munger in the opinion that EBIDA is synonymous with bullshit earnings, but in this particular corner of the investing world it will be used as a multiple by most interested investors, so we’ll have to go with the tide here. This is also an area where the use of “adjusted EBITDA” (read as really bullshit earnings) is also used liberally, especially in the case of NEXN. So, we’ll adjust back to traditional EBITDA.
For Magnite, recent EBITDA margins have ranged between 19% - 22%. For analysis 20% will be used.
For PubMatic, recent EBITDA margins have ranged between low to mid 30’s. For analysis 30% will be used.
For NEXN, an adjusted EBITDA of 47% is presented. RE-adjusted and we land at 23%. Which, go figure, is comfortably in-line with peers and what we will use.
Now, with the addition of new revenue we cannot simply apply the same EBITDA margins to new dollars coming in. It does not take much to process additional transactions on an already built platform and incremental margins in this space are very attractive ranging between 50% and 70% for the companies under analysis.
For conservatism, we will use the low end of 50% across all companies in assessing the new EBITDA figures in the three scenarios we are considering.
With that in place the updated numbers look at follows.
The final step is to approximate a valuation for each of the Sales and EBITDA outcomes across our various scenarios and compare that to present market capitalizations. Typically, EV/EBITDA multiples have ranged between 10-15 for this segment and we’ll use 12 across all companies. For Price/Sales, the average of the past few years (i.e., post COVID), will be used. These come out to 5.05, 6.54 and 3.09 for Magnite, PubMatic and NEXN respectively.
The equity value increase across our three scenarios would look as follows along with the upside to their market capitalization.
Now, whether a Price/Sales or EV/EBITDA multiple is more relevant is probably not worth stressing over. So, we’ll adjust the above assuming each are equally likely and take the average of both.
With this we can approximate that each of these companies, in a DOJ win outcome, could be worth these multiples over their market capitalizations depending on how much market share Goggle ends up ceding to the open market.
For those who have read any of my material in the past, you’ll know I don’t outright endorse any defined positions as I don’t know any of the readers’ specific situations. But we can at least discuss broad approaches. And, for what it is worth, highly out of the money options with +1 year to expiration was my decision in the last weeks of 2023 and first few weeks of 2024.
The Conservative Basket Approach
I’ll state the obvious, NEXN is a bit of a speculative position. They report via IFRS, absolutely LOVE adjusted EBITDA and have less reporting history than Magnite or PubMatic. That being said, they likely have the greatest upside if everything goes perfectly.
The simplest and most conservative approach here would be to go long a basket of these three, with weights according to their “legitimacy.” So heavier towards Magnite/PubMatic, and NEXN to a lesser extent or not at all.
These companies have already had to figure out to survive in Google’s shadow, so even if no breakup is forced, they should not have that much downside as they all are trading at historically low multiples.
There is also the motivation of hedging a long Google position should this come to pass.
For those who have read this far, and are interested in this topic for that reason, I have a hedging size calculator if you’d like to take a look. Send me an email through the website and I am happy to share the Excel.
The Binary, Asymmetrical Bet, Approach
Whenever you have a calculated outcome of potentially 5x the current value (average of all scenarios across all companies) it’s worth looking into an asymmetrical bet opportunity. And I did mention 50x-100x on the onset of this writeup.
If looking at LEAPs, this enters the realm of possibilities.
LEAPs, or Long-Term Equity Anticipation Securities (terrible acronym), are options that expire >1 year from the time they are purchased. With these, they are so far out they behave closer to equity with a lot of exponential behavior in regards to the underlying.
In simplified terms, a single unit move in the stock price produces a move in the option value multiples of that move in the stock price. But they also go to $0 in value as the expiration approaches. So, position sizing is of the utmost importance.
I am publishing this, somewhat unfortunately, on a large post earnings announcement jump in the market caps’ of Magnite and PubMatic of +10% and +20% respectively. So speaking to any specific option contract value is probably going to be irrelevant as soon as I type it.
But, on sufficiently out of the money calls for both companies over one year out, the upside even in the low end of the above calculated equity value increase scenarios was hovering in the 80x-100x range prior to this earnings release. Considering the details of the suit, and that I feel the probability of a DOJ win is comfortably above 10%, likely closer to 20%, this is a bet I’ll consider taking every time.
Back to the importance of the Rocket Docket, options are highly time dependent, and timing is a very hard game to play. But, with a September 9th trial date an expiration date within six months of that date is probably sufficient given the forward looking/extrapolative nature of the public markets.
- - -