Richard Kramer, Founder
Arete Research Services LLP
1
CMA Online Platforms Review: Arete Researchs View
31 July 2019
Arete Research is the leading independent technology investment research boutique, serving over 150 global TMT investors
since its founding in 2000. We provide investment advice and services in the UK, US, and Hong Kong, regulated in each
market by the FSA, FINRA and HK SFC. We have devoted considerable effort to looking at all aspects of large technology
companies such as Google, Apple, Amazon, Netflix, and Facebook, which we actively research, despite their de minimis
disclosure around financials and operating metrics. We regularly visit or speak with 100s of companies monthly. We
communicate our findings to the largest investors in these companies (among them large UK institutional investors and
hedge funds). Our work is free from the conflicts of interest that skew research from investment banks and industry
analysts, who have a vested interest in serving the companies they cover or report on, as they are typically paid by them
in some fashion. We have no financial dealings with Google or Facebook, either as a customer, competitor, partner, supplier
or content provider, allowing us to take a wholly disinterested view. Our views are based upon decades of financial markets
and commercial experience, and are wholly independent, as we never seek or receive any compensation from companies.
Ample Evidence of Dominance from Financials
In reading the premise of the CMA inquiry, we find ample evidence of their market dominance from their financials. As of
recent 2Q19 results, Google had $117bn of net cash and generated $51bn of operating cash flow in the last 12 months,
while Facebook had $49bn of net cash and operating cash flow of $33bn in the last 12 months, before it pays its $5.1bn in
fine to the FTC and SEC. A further facet of market power stems from the sheer scale of Google and Facebook operations.
The two companies had global R&D budgets of $25bn in $12bn in 18. Even if only a fraction of that spend is directed
towards digital ad technology, this gives them the ability to out-spend rivals or copy their features or innovations rapidly.
Behind this is a legal machine which can intimidate smaller rivals and impose unequal business terms upon them. Until the
access to these near-unlimited financial resources is curtailed and we believe they have already accepted de facto
restraints on any large scale M&A for fear of regulatory scrutiny - it will be equally difficult to constrain the ability of
these companies to dominate the market for digital ads.
… Led By Dominating Time Spent
Digital advertising and content markets have seen a relentless consolidation of towards an ever-smaller number of ever
larger firms. The leading digital platforms Google and Facebook provide the properties and utilities (search, YouTube, Maps,
GMail, Facebook, Instagram, WhatsApp, etc.) are garnering the majority of online time spent (something which can be seen
in ample studies from Ofcom and others), along with their unique ability to know the identity of who is spending that
time (i.e. targeting), and block any third parties from accessing that data. This gives them a dominant market position in
the supply of digital ad inventory. Recent efforts by publisher consortia such as the Ozone Project to aggregate audiences
and share data on them with advertisers, outside of the platform ecosystems. There is no dobut this has allowed Google
and Facebook to grab the vast majority of growth in Internet ad spend; no other platforms are comparable to the reach
and scale that Facebook and Google enjoy, with Amazon and Apple as its only “rising” competitors. The ability of ad agencies
to direct this spend is diminishing, as the Internet giants go direct” to address advertisers large marketing budgets and use
self-serve platforms to aggregate ad spend from SMEs. There is no UK-specific equivalent business, and Google no longer
breaks out its UK-only revenues. We regard the documents filed by both companies at Companies House for local UK
operations as unreliable due to the range of transfer price arrangements which distort underlying business.
CMA Online Platforms Review: Arete Research’s View
2
Directing the Flow of Traffic
Google is also the main conduit through which the long tail of publishers get paid (i.e. those smaller sites which cannot
afford their own digital ad salesforce or can adequately track their audience and present it to advertisers). We refer to
the $27bn Google spent in 2018 traffic acquisition costs - $14.2bn paid to Google Network members or third party sites,
and another $12.6bn paid to Distribution partners, including Apple - paid annually as code of silence money for
publishers, who need their representation. Google aggregates advertiser spend and has the option to direct that spending
to its own properties or to third parties. One solution would be to oblige Google to separate the third-party ad network is
runs by far the worlds largest from selling ads on its own properties. This may be beyond the scope of the UK to insist
upon, but it could be proposed that in the UK, Google cannot represent third-party sites while also selling its own ad
inventory. This would open the market to others selling ads on third party sites and restrict Googles access to data on users
beyond its own Sites.
Similarly, Google regularly directs search queries to its own properties, either where it promotes its own apps (in its
PlayStore) over rival alternatives, whether it favours its own travel or shopping advertisers over organic results, or
whether, via its partnership with Twitter (whose Chairman is Googles former Chief Business Officer) which displays its
tweets in search results, an option not available to other social networks. The complaints of failing businesses like Yelp are
well known, but more able rivals have also suffered from lack of exposure in Google search results, especially when Google
owned Zagat, a leading restaurant review site. Google also aggregates results from crawling publisher sites, and may
present these as its own, with limited reference to, or path for monetisation for the publishers who have done the work.
In many instances, Google sits on top of multiple other content creators, directing whether they get paid (or not),
something Facebook has not matched by establishing its own third-party network.
Limiting IDs to its Own Networks
Another issue is that Google (and Facebook) restricts the passing along of log-level IDs, so that advertisers have no way
to independently confirm that they reached an audience. In this way, Google services are opaque by design they
stubbornly resist outside 3P measurement. This also prevents cyber-security researchers from testing the level of bot
activity or other ad fraud schemes on Google platforms. There have been numerous instances where Android apps were
found to be allowing extensive over-permissioning in terms of data collection. This is done by a range of utility apps, such
as battery monitors, keyboards, which have been found to run click injection schemes or collect keystroke data to sell apps
install ads. This is equally a clearly well-documented problem with Facebook, which has now, staggeringly, been
indemnified by the US FTC for any mis-steps in applying its privacy policy prior to its recent settlement. Facebook has on
multiple occasions quietly updated its policies on data collection, while reserving the right to grade its own homework
when advertisers demand audits of its data via third-parties, releasing only limited data on the audience segments it
presented.
Distribution of APIs via Chrome and Android
Many of these business arrangements which Google exploits to sustain its $150bn sales business are built on the
distribution of its Chrome Browser and Android smartphone OS and many of its specific features, embedding Google
properties as default APIs (applications programming interfaces); for example, most Android compatible applications will
call up Maps, YouTube or Google Search as a default option within the functions of the apps themselves. Any business with
an application will be obliged to show its location using Google Maps, or embed video content which resolves to YouTube.
Google dictates terms to developers and encourages them to use their systems at set-piece events like its Google I/O
developers conference. These front-ends for consumer experiences allow Google to nudge consumers towards its own
services and create barriers to building other services alongside Googles. Facebook has a similar library of APIs, holds its
onw developer conference (F8) and tightly controls its own Partner Network of authorised advertising resellers.
CMA Online Platforms Review: Arete Research’s View
3
Digital Ads A Rigged Market
Google owns every component of the “ad tech” stack. This creates a scenario that would be unimaginable in comparison
with how financial markets operate, and leaves advertisers and publishers alike at a significant information asymmetry
and commercial disadvantage relative to Google. Fig. 1 shows that Google’s take rate is between 30-50% (although we
estimate it is closer to 30% in most cases) in comparison to similar processes in financial markets where typical rates are 5-
200bps. Beyond Google owning every part of the value chain, it also has dominant market shares: AdX, its exchange, has
~$7bn of spend flowing through it and 50%+ market share, whilst Double Click Bid Manager has 30%+ and Google’s Ad
Server has an 70%+ share. These figures are hard to measure with precision due to Googles wholly insufficient disclosure
(other leading Internet players are no better: Facebook does not disclose its Audience Network sales, nor does Amazon or
Apple directly disclose their ad businesses).
The financial analogy would see a scenario where JP Morgan or Barclays owned the stock exchange (i.e. the NYSE or LSE),
acted as the “market maker” – i.e. aggregating bid and offer quotes for stocks, while also acting as the exclusive broker for
its own shares (i.e. advertising inventory on YouTube, Search, Maps, etc.). In this case, Google also acts as the “broker” for
other shares (publisher inventory of display ads or video units on their pages, bought and placed via Google’s Network).
Under this system, Google can see “both sides” of the trade they know how deep "bid density" is for certain types of
audiences or inventory, and how deep the pool of that inventory might be to satisfy market demand. This allows them to
set price accordingly, to maximise profits. It also allows them to “toggle” or “steer” demand from advertisers towards the
type of content which carries higher margins (i.e. YouTube creators that have a lower of revenue share than others). Since
YouTube inventory can only be purchased through Google’s DoubleClick platforms, Google can effectively determine the
clearing price of certain types of inventory. In this way, Google “runs” a rigged ad market. If such a market were established
in the financial world, it would present clear conflicts of interest and ample opportunity for market manipulation (for
example, acting as sole market maker in one’s own equity, while retaining the information advantage of knowledge about
the underlying state of that equity). In the financial markets, there are floor limits on the cost of transactions, and many
“indicative” bids from parties looking to determine the depth and liquidity of markets for any particular instrument. This
cannot happen as easily from the outside of Google’s ecosystem (though we do see this behaviour in other parts of Ad tech,
across Demand Side Platforms and Ad Networks). Generally, the costs for transacting in a stock that sells for £1 or £1,000
are broadly similar. Many brokers offer “per share” fees, not linked to the value of the share. In the case of Google or
Facebook’s ad tech buying systems (embedded into the CPM) or the “take rate” of e-commerce vendors like Amazon, the
costs for transacting in more highly valued shares are higher, even though the underlying transaction involves the same
stages (of execution/fulfilment, reporting, and reconciliation).
Fig. 1: Comparing Google to the Financial Markets
Source: Arete Research
CMA Online Platforms Review: Arete Research’s View
5
GDPR Power Play by Platforms
In the run-up to the implementation of GDPR legislation in 2018, Google introduced its own consent framework at the 11
th
hour and insisted that agencies or advertisers that wanted to match their own 3P data with Google did so within Google’s
Ads Data Hub, which gave Google visibility of the data. Google also hosted its Ads Data Hub internally on its own Google
Cloud Platform (GCP), giving a latency advantage to anyone performing media buying with the data. Under the pretext of
GDPR, they removed the practice of passing on log level IDs, further limiting advertiser’s visibility into what audience they
were buying and brought together all of its attribution products (to determine whether the audience was reached as
promised) into a single Unified Attribution offering. Marketers now receive an aggregate view rather than on an individual
basis, making it more difficult to track who has been served ads and forcing marketers to commit more spend to a single
platform (previously, marketers could see who they had targeted across platforms, whereas now they receive aggregate
data and hence the same ad could have been served to the same person across many platforms). These changes require
marketers to have “blind faith”, that Google can serve ads which drive the best ROI vs. which ones might be the most
profitable (for Google). There is no independent auditing of Google data (it works with a number of partners, but they are
sharply curtailed by domain, e.g. viewability, mobile attribution, e-commerce, etc.), leaving Google in the enviable position
of “grading its own homework.” Regulators have yet to show they understand the practice of programmatic i.e.
algorithmic advertising technology, whether it has a dramatic impact on politics or the wider commercial world.
Click to Donate Data
Another aspect of online platforms is their ability to coerce customers to accept what we call data donation policies,
without understanding the value of their data, if not at an individual level, then in aggregate, sold as part of consumer
segments or audiences. Google and Facebook have precise data on how many users opt-out or try to limit data collection,
how many take the time to read complex legal jargon that constitutes their privacy policies or EULAs. At the very least the
UK government should oblige digital platforms to expose their opt-out rates, and seek to value the data they collect on
end users, but examining the price differential between ads bought with data and those bought on an un-targeted
basis. Our view is that relatively few consumers will take advantage of data mobility; the blanket permissions asked for by
multiple applications (location, patterns of use, access to photos and contacts) are poorly understood by consumers, a
loophole which is abused by many companies, not just the largest players.
But Not to Pay Tax
Clearly there are also a dimension around international tax avoidance and transfer price agreements but we do not see
this as an issue to be resolved solely by the UK, which bears responsibility in how its many Crown Protectorates are offshore
tax havens, which are used to house intellectual property and other valuable assets. The notion that all of the ads sold by
Google in the UK are actually sold by its Irish entity and based solely on intellectual property developed in the US but held
in other low-tax jurisdictions, is simply indefensible by any common-sense test. At the EU level, a system of unitary taxation
would end the system by which all Google ads sold to UK clients are executed by its low-tax Irish arm. The reason why this
is important is that it gives digital platform companies greater control over their cash flows and larger profit pools on which
to draw for investment, to further cement their market dominance.
CMA Online Platforms Review: Arete Research’s View
6
Can Regulation Work?
In terms of regulation, there are many scholars and legal experts working on the problem of “re-setting” antitrust law away
from the basis of “consumer harm” and instead considering monopoly or monopsony power. This is before considering the
social implications of filtering news and broader consumer attention through a decreasing number of outlets, with the
ability to practice predatory pricing (just ask the BBC, competing with Netflix for talent, but unable to run a deficit of $16bn
in unrecognised content liabilities or sustain losses of $3bn of annualised free cash flow). There is also the problem of
politicians and regulators lacking an understanding of a complicated ad tech market, which has evolved into a sprawling
mess of 1P and 3P data applied to a vast array of content, and rife with issues of measurement and ad fraud. As to specific
policies we have a few suggestions:
1. Keep GDPR. It would be the height of folly for the UK to abandon the GDPR requirements it has recently imposed and
try to create a different regime. This also applied to forthcoming ePrivacy rules. The more globally or regionally
harmonised these are, the better for all concerned. These rules should soon to be clarified in case law defining key
terms such as "legitimate interest" or informed consent.
2. Give Everyone a Data Briefcase. Consumers need to be made more aware of the value of their data, both individually
and in aggregate, and allowed to “monetise” it in a market where they have the right to withhold it from the large
players. Making data “portable” is one step, but assigning a clear monetary value would create interest to overcome
“switching costs” of shifting that data around or denying access to it.
3. Open Access to Platforms? Aggregated pools of data from services like Search or social networks could be considered
“public goods” whereby all firms would get equal access to the data at similar costs. There is no way for advertisers or
sellers to effectively audit the activity on platforms. A structural separation of the “stock exchange functions” and the
media assets of large Internet players would be a start, but this may be beyond the remit of any one country. At the
very least, there should not be exclusive inventory (e.g. YouTube) which one is forced to buy through closed platforms.
This is at the heart of the current case before the US Supreme Court about the 30% take rate on Apple’s AppStore.
4. Require Improved Disclosure. Given that the UK is unlikely to be able to force any structural separation of elements
of large US Internet players (forcing a breakup of Google and YouTube, for example, runs counter to their ad buying
systems that find audiences and places ads on Maps, Search, Gmail and YouTube, as well as on 3P sites), one could
require greater disclosure of commercial terms, following similar rules as would govern capital markets; this should
be an urgent priority for the US SEC, but it has dropped the ball numerous times on this issue. Likewise, the basis of
transfer price arrangements and taxation should be broadly disclosed, not shrouded in secret. It is nearly impossible
for financial analysts to calculate the likely tax rate of Google operations outside the US.
5. Open Access to Selling Inventory? The problem with getting publishers or other small ad tech firms to complain about
leading players is that they rely on those same players as the primary conduit for revenue and allow them to act as
“sales agent.” Breaking this link is critical allowing anyone to access Google or Facebook inventory, the same way
as anyone can use the LSE to buy any share, and removing the opaque by design nature of sales on this closed
platform, outside of where HM Treasury can observe this large pool of economic activity current invisible to it.
There is much more detail we could go into to document the market dominance from our position looking at these
companies financials and assessing them as investments. Please feel free to reach out to us if we may be of service.
Richard Kramer
Managing Partner