Understanding Data Valuation: Valuing Google’s Data Assets

Digital personal data are increasingly understood as a key asset in our digital economies. But how should we value such data? Numerous policymakers, regulators, and stakeholders are trying to work out how to manage the collection, use, and valuation of data in order to balance the advantages and disadvantages of its collection and use. The negative implications of data practices may include privacy loss, data breaches, or declining market competition, while social and economic benefits include improved service delivery, more efficient welfare systems, or better products. Increasingly, data are conceptualized as an asset. To understand the value of data as an asset means understanding how data are configured as an asset; data value does not reflect ownership and property rights per se, but rather diverse modes of access and use restrictions (usually delineated by opaque contractual agreements). Data are increasingly controlled by a few, large digital technology firms, especially so-called ‘Big Tech’ firms. In this paper, we use Google as a case study of how Big Tech firms configure and value digital data as an asset. We analyse how Google understands, frames, values, and monetizes the data they collect from users. We qualitatively analyse an extensive dataset of financial documentary materials produced by and about Google to identify the different modes of access and use restrictions that Google deploys to turn digital data into a valuable asset. We conclude that, despite being highly ambiguous, Google’s approach to data value focuses on monetizing users.

more efficient welfare systems, or better products.These policymakers, standards-setters, and stakeholders include subnational and national governments and their agencies (e.g., U.K. Treasury, competition authorities; Province of Ontario); national and international statistical offices (e.g., StatCan, System of National Accounts); think tanks (e.g., Centre for International Governance Innovation, Bennett Institute); inter-governmental institutions (e.g., UNCTAD); supra-national institutions (e.g., EU, Eurostat); international institutions (e.g., World Economic Forum, OECD); professional organizations (e.g., English Law Commission); and private business (e.g., Deloitte).Increasingly, these 'stakeholders' conceptualize data as an asset, especially in national accounting standards produced by international policymakers (e.g., System of National Accounts).
Understanding data as an asset requires an interdisciplinary approach.Analysing its configuration and valuation as an asset entails drawing on the conceptual and empirical insights of disciplines like science and technology studies (e.g., [3], [4], [5]), economics and political economy (e.g., [6], [7], [8]), law (e.g., [9], [10], [11]), and policy studies (e.g., [12], [13], [14]).This interdisciplinary literature stresses the need to think about the techno-economic construction of assets in order to understand them as political-economic objects.As an asset, for example, data are constituted by a socio-technical configuration of legal rights, technoscientific devices, and policy regimes that create forms of de facto data exclusivity and control [11], largely through the construction of limitations and restrictions on data access [2], [15], [16].As such, actually existing data valuation does not reflect usual forms of ownership and property rights per se, but rather diverse modes of access and use restrictions created through economies of scale, network effects, intellectual property, limited interoperability, and contractual arrangements.As a result, data are increasingly controlled by a few, large digital technology firms, especially so-called 'Big Tech' [17].
For more information, see https://creativecommons.org/licenses/by/4.0/Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
earnings calls (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022); and court cases about the collection and monetization of digital personal data.Our starting point was to use qualitative data analysis software (NVivo and ATLAS.ti) to identify the different modes of access and use restrictions that Google deploys to turn data into an asset.However, in seeking to undertake this analysis, it became evident that there is no easy way to analyse the value of data held by Google.
Throughout our qualitative analysis, we use long quotes from the empirical materials to provide context and evidence for our claims and arguments.The empirical analysis begins (Section III) with an overview of the common definitions of personal data found in the business, data policy, and political economy of digital media literature.In Section IV, we draw from earnings calls and financial reports to elucidate the rise of programmatic advertising as Google's business model and the importance of traffic acquisition costs (TAC) to understand the company's approach to monetization.Section V connects data from Google's financial documents with different valuation approaches in order to understand the most likely approach to Google's monetization strategy.Finally, we end the paper with the policy implications concerning the collection and use of personal data by Big Tech firms like Google, focusing especially on the need to enhance transparency and accountability around corporate data holdings to spread the economic benefit of data use more fairly.

A. Definitions of Personal Data
Today, when people use the terms like "personal information" or "personal data", they generally refer to the collection, use, and exploitation of specifically digital personal data.Recent digital and algorithmic technologies have enabled the massification of the collection, use, and commercial exploitation of personal data, entailing new technological and economic objectives of use (e.g., inferential analytics enabled by 'big' data) and new technological and economic structures of collection (e.g., collection of data on our online and cellular activities) [5].Personal data can be categorized in different ways, depending on its type; the main collection method; and its characteristics.
First, it is possible to identify different types of personal data.According to the OECD [18], these include: "User generated content"; "Activity or behavioural data"; "Social data"; "Locational data"; "Demographic data"; and "Identifying data of an official nature" [19]."Sensitive data", such as genetic, health or financial information, is also collected in some cases by large tech firms [20].It should be noted that this list of data types is not exhaustive, as it is difficult to identify and classify all types of data collected by large tech firms.Second, personal data can be characterized by its collection method.Again, the OECD [18], [21] is useful here as it differentiates data collection into: "Volunteered" by individuals willingly; "Observed" about individuals; and "Inferred" from the analysis of individuals.Finally, personal data can be understood by its characteristics and differentiated between: "Identifiable"; "Anonymous"; and "Pseudonymous" [22].

B. Personal Data as a Political-Economic Object
Academic, business, and policy debates about the commercial uses, exploitation, and value of personal data have been going on for more than two decades [23].Earlier discussions focused on the notion of personal data markets as markets for privacy, while more recent discussions have expanded into other areas.Generally, it is now commonly accepted that personal data are an important economic resource, asset, and/or commodity across this academic, business, and policy literature (e.g., [18], [21], [24], [25]).Recent discussions also tend to emphasize that personal data have certain qualities that make its commercial use and exploitation different from other economic resources, assets, and/or commodities, including: 1) Personal data are non-rivalrous; 2) Nevertheless, personal data can be excludable; 3) There are no property rights to personal data per se; 1 4) Personal data have emergent properties, or positive externalities; 5) Personal data are relational [5], [9], [26], [27].Personal data are an important resource for a range of different public and private entities, from government agencies to multinational corporations.Notably, however, they have become crucial to private sector business models in the digital technology sector, underpinning consumer services like online search; social networking platforms; online advertising, especially with individual targeting; analytical and marketing services for online (and in-person) businesses; and artificial intelligence or algorithmic technology products and services [3], [18], [19], [22], [28].
In particular, personal data are an important economic resource and asset for online advertising companies like Google. 2 For example, Germany's federal competition authority [30] noted in 2023 that "Google's business model relies heavily on the processing of user data" and that "Due to its established access to relevant data gathered from a large number of different services, Google enjoys a strategic advantage over other companies".This statement follows an earlier one by the EU's competition authority in 2021 that the European Commission "already considers data as an asset in merger assessments".This statement was a response to an EU Parliamentary question regarding an investigation of "the way data concerning users is gathered, processed and monetised by Google" [31].

IV. GOOGLE, DATA ASSETS, AND THE RISE OF PROGRAMMATIC ADVERTISING
Online advertising can be traced back to 1994 when advertisers and publishers negotiated directly with one another to buy and sell ad inventory (i.e., online ad space).Advertisers would buy ad space on a "cost-per-mille" (CPM) basis, where "mille" represents 1000 impressions or views of a website.Online advertising is split between search advertising 1 Some countries and jurisdictions allow property rights for databases, representing a particular arrangement and structuring of data equivalent to copyright [21]. 2 Personal data are also central to the business models of other Big Tech companies [29].
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
(i.e., matching search keywords with advertising bids) and display advertising (e.g., text, image, or video displays on websites, apps, etc.).In the second half of the 1990s, online advertising became dependent upon the advertising technology sector ('adtech'), which covers a range of intermediaries that sit between advertisers and publishers [32], [33].These intermediaries included DoubleClick, an early adtech company that was founded in 1995 and acquired by Google in 2008 [34], [35].This acquisition was the starting point for Google's emergence as the most powerful player in digital advertising, which was cemented with the rise of programmatic online advertising.
Today, online advertising reflects the rise of programmatic advertising from the late 2000s onwards.Programmatic advertising has automated the online advertising process and is "fuelled by various categories of user data" [34].Programmatic advertising moved online advertising away from a context-based approach (i.e., dependent upon website content) to an approach based on the characteristics of individual users [34].A common process of programmatic advertising is real-time-bidding (RTB), which involves an automated auctioning of a publisher's ad inventory in the milliseconds before a webpage loads.
Since its emergence, programmatic advertising has quickly dominated online advertising, representing 86 percent of online ad revenues in 2022 [36].Programmatic advertising is highly dependent upon the collection and analysis of personal data.In turn, it has driven the expansion of the collection of these various personal data through a range of data collection mechanisms (e.g., cookies, APIs, SDKs, etc.) and architectures.
Google's 2022 10K Annual Report notes that Google comprises Google Services and Google Cloud; the majority (around 80 percent) of Google's revenues come from (online) advertising.As noted in several of their earnings calls, Google generally presents itself as a platform or ecosystem connecting advertisers and service users, where the former can serve ads and the latter can create content and access Google products.Advertising enables Google to monetize users in their ecosystem: "• • • if we could bring out a product that will cause people to use Google and its various applications that much more and they spend more and more of their day using Google services, that allows us to eventually monetize that.So we do not insist on a direct link from say a product that does not get revenue to one that does."[37] Google is able to monetize users through the collection of their personal data, which underpins programmatic advertising.On their website, Google states that "we use data to show ads that are useful to you, whether they are on Google or they are on websites and mobile apps that partner with us.We do not sell your personal information to anyone" [38].Monetizing personal data through advertising has been Google's long-term business model; for example, their 2004 10K Annual Report noted that: "Concerns about our collection, use or sharing of personal information or other privacy-related matters, even if unfounded, could damage our reputation and operating results.Recently, several groups have raised privacy concerns in connection with our Gmail free email service which we announced in April 2004 and these concerns have attracted a significant amount of public commentary and attention.The concerns relate principally to the fact that Gmail uses computers to match advertisements to the content of a user's email message when email messages are viewed using the Gmail service.Privacy concerns have also arisen with our products that provide improved access to personal information that is already publicly available, but that we have made more readily accessible by the public."[39] [Emphasis ours] Privacy concerns from users have continued to proliferate in the past two decades with the release of other Google products that now use artificial intelligence, like Google Home and Google Assistant.These technologies provide customized user experiences by providing information or carrying out tasks, usually activated by the user's voice.
Google's advertising business model (i.e., monetizing users) has remained relatively consistent throughout their corporate existence, even if their adtech products have changed.Today, Google has a suite of adtech products and services that connect advertisers and publishers (see Fig. 1).Google operates on both sides of the online advertising market with products for both publishers and advertisers.Briefly, its publisher products include Google Ad Manager which combines an ad server (helping publishers manage the sale of their ad inventory), supply side platform (helping publishers sell their ad inventory), and ad exchange (operating real time auctions for ad inventory); and its advertiser products include Campaign Manager (helping advertisers manage their ad campaigns) and Google Display and Video 360 (helping advertisers automate the buying of ad inventory).
Since its initial public offering (IPO) in 2004, Google has made most of its revenues from advertising.As their 2022 10K Annual Report states: "We generate revenues by delivering relevant, costeffective online advertising; cloud-based solutions that provide customers with infrastructure and platform services and collaboration tools; sales of other products and services, such as apps and in-app purchases, digital content products, and hardware; and fees received for subscription-based products such as YouTube Premium and YouTube TV" [42].
Google generates these advertising revenues by monetizing the "traffic" (a term used to define its users) attracted to (or generated by) a suite of products and services (defined as "properties" by Google) and partner companies in the Google Network.Google generates advertising revenues from its own properties and from other non-owned properties within its Google Network: this Google Network includes companies that sign-up to products like AdMob (for advertising on mobile apps) and AdSense (for display advertising on the Google So it is important, but it's one of the many that we use.In terms of assets that apply to that, we do have a very, very large number of users coming to our door every day.A considerable percentage of them are logged-in users that are using multiple of our products.So there is a large variety of signals that we'll be able to use with user support and users seeing value from it to make the overall experience better."[43] [Emphasis ours] A key and ongoing concern of Google since at least its IPO, has been something called "traffic acquisition costs" (TAC) [44].According to their 2022 10K Annual Report, TAC consists of: 1) "Amounts paid to our distribution partners who make available our search access points and services.Our distribution partners include browser providers, mobile carriers, original equipment manufacturers, and software developers.2) Amounts paid to Google Network partners primarily for ads displayed on their properties" [42].In 2022, TAC amounted to U.S.$48,955 million and is defined as a "cost of revenues", which represents the direct costs of selling a product or service to a customer; in Google's case, the customer is advertisers, and the cost of revenues entails the attraction of user traffic and the collection and analysis of their personal data (e.g., to create user profiles).TAC is an important business metric for Google, as traffic (i.e., users) reflects a core resource/asset underlying its monetization strategy through advertising.Their 2006 10K Annual Report emphasized the possibility that TAC will increase if the company cannot improve monetization: "In particular, traffic acquisition costs as a percentage of advertising revenues may increase in the future if we are unable to continue to improve the monetization of traffic on our Web sites and our Google Network members' Web sites, particularly with those members to whom we have guaranteed minimum revenue share payments" [45].Reducing the proportion of TAC relative to advertising revenues has been a key goal for Google, which is evident across their annual reports and earnings calls; in 2022, TAC represented around 22 percent of advertising revenues [42] compared with 32 percent in 2006 [45].TAC includes "revenue sharing" with mobile carriers and original equipment manufacturers (OEM), especially after the advent of smartphones and their growing importance to online advertising, which becomes apparent to Google between 2008 and 2012 according to their annual reports and earnings calls.

A. Data Valuation Approaches
Although it is possible to identify personal data as an important asset for Google (and other adtech firms), it can be difficult to calculate the value of that personal data to the firm.To illustrate this, we deploy different data valuation approaches that have been used to value digital data assets, while recognizing that there is no agreed or standard method for valuing personal data.Data valuation approaches can be split into different categories, Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
each of which reflects a set of theoretical and methodological assumptions about personal data.
Business performance approaches use performance measures.First, recorded assets methods focus on a company's balance sheet (e.g., intangible assets, goodwill); these methods are not usually possible to do with personal data, however, as data are generally not recorded on company balance sheets [1].Second, market capitalization (or private valuation) methods focus on investor valuations; they tend to reflect market sentiment (i.e., what investors are willing to pay) which can be highly subjective and oscillate widely [5].Third, revenue methods use calculations like average revenue per user (ARPU) to assess the value of personal data; however, they require precise information on the revenues aligned to specific personal data uses [49].
Business cost approaches use cost measures.First, calculations of the cost of producing or replacing personal data, using this as a proxy for the value of personal data which can be calculated from the historical cost to produce it or to replace it if it is lost or destroyed; limitations of this method include the fact it reflects the lower-bound of valuations (because it is 'historical' rather than current market 'fair value') and is often based on aggregate calculations rather than calculations of specific datasets [50].Second, damages reflect the regulatory or legal costs incurred if a company experiences a data breach which can act as proxy for data's value; however, these methods can be inconsistent and different between jurisdictions [46].Third, a version of relief of royalty, representing the calculation of the costs a business would incur if it had to license data externally instead of producing it itself; this could be calculated from the cost of personal data in data markets (see below on limitations of this), or a calculation of the cost of production [1].Finally, cost of revenues methods, reflecting the costs incurred in generating revenues, including investment in the collection of personal data.
Market approaches are based on the sale of data in markets [22].This includes the legal sale of personal data in 'data markets' (e.g., by data brokers): however, there are limited examples of direct or open data transactions and personal data may be more valuable than the direct price for specific data points in such data markets (e.g., by combining new datasets with existing datasets).Data market pricing tends to reflect business-to-business (B2B) transactions, but not consumer-tobusiness transactions (C2B) which are more relevant for personal data [3].It also includes the illegal sale of personal data on the 'black market', such as the sale of information like credit card details; this information is difficult to obtain, however.
User or data subject approaches calculate personal data value from the 'stated preferences' of individual users (for privacy) through surveys or experiments [25].These methods can assess both: a user's willingness to pay (WTP) to protect their personal data against disclosure, which can be assessed by asking them how much they would pay for their privacy; and a user's willingness to accept (WTA) payment for personal data, which be assessed by asking them how much they would accept to sell their personal data -this sale price tends to be higher than WTP.Both WTP and WTA methods are stated preferences, however, meaning that they reflect people's statements rather than observable decisions [25], [46], [47].

B. Valuing Google's Data Assets
In this section, we apply these valuation approaches to Google.Both market and user approaches provide limited means to calculate the value of Google's data assets.Google clearly states that they do not sell personal data, including having a "security and privacy principle" on their website to "Never sell our users' personal information to anyone" [51].This means that market approaches are not useful for understanding the value of the company's data assets, as data are not being sold in a market.It might be possible to survey Google's users to work out their willingness to pay (WTP) for privacy and willingness to accept (WTA) payment for their personal data, but this suffers from the limitation that any valuation would reflect a stated preference rather than actual user behaviour.
Business performance or cost approaches seem more useful, but not all of them.From our analysis, it is evident that Google does not record personal data as an asset on their balance sheet and does not frame it as an asset in their annual reports and earnings calls.Google's market capitalization could be a better proxy to understand personal data value, but market capitalization reflects investor expectations about a company (and its assets) and does not necessarily help our analysis.In particular, the market capitalization of Google has risen and fallen quite significantly over the past few years, which is unlikely to reflect a similar rise and decline in data value in the same time period.
Google's revenues can be used to calculate the value of personal data with metrics like ARPU.Two relatively old studies identify an aggregate ARPU for Google: one puts ARPU at U.S.$40 (in 2012) [52] and another puts ARPU at U.S.$59 (in 2017) [53].Neither is very clear about how this amount is calculated, however.It is easier to identify an ARPU for specific Google products, although there are few mentions of this in the empirical materials.For example, there is only one mention of an ARPU in Google's own annual reports and earnings calls (2004-2022), suggesting the ARPU for Google Play is around U.S.$7-8 (in 2019) [54].It is possible to do our own calculations of the ARPU of products like YouTube: this can be calculated at U.S.$11.63 (in 2022) by dividing YouTube's ad revenues (U.S.$29,243 million) by the number of users (2,514 million).Despite the two aggregate studies mentioned above, it is difficult to calculate an aggregate ARPU for Google without better access to specific (and confidential) data on business segment revenues and user numbers.For example, it might be possible to calculate a ARPU for "Google Search & Other" (in 2021) by dividing revenues of U.S.$148.95 billion by the 2,852 billion searches in the same year.Here, each search would generate around U.S.$0.05 in revenue.A major caveat is that such calculations do not take into account what constitutes the "Other" in these revenues.
Google's costs seem to be a better proxy for calculating data value, although not all business costs.Production costs and relief of royalty are difficult to calculate because it is difficult to disaggregate tangible and intangible costs of Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
data collection and monetization (e.g., data centres versus intellectual property).An alternative would be to use cost of revenues because this identifies costs incurred in generating specific revenues from advertising, Google's primary business activity, and it is specified in Google's annual reports.The empirical materials we analysed highlight the importance of "traffic acquisition costs" (TAC) in this regard.TAC represents the costs of attracting users (and their personal data) to Google's adtech ecosystem in order to monetize those users by selling access to them to advertisers.
It might therefore be possible to calculate the aggregate value of Google's data assets by using the TAC rate that Google pays each year.TAC reflects the value Google itself is willing to pay for attracting users and their personal data to the company's properties in order to monetize them through advertising.To do this properly, however, would require more disaggregated information on user numbers and revenues generated from Google's different properties (e.g., Search, Gmail, Play, Maps, etc.).
We have tried calculating the value of Google's data assets by using the specific payments Google makes to other companies for access to their users; for example, Google makes significant TAC payments to Apple to set Google Search as a default on iOS devices.According to court documents, Google paid Apple U.S.$1 billion in 2014 to set Google Search as iOS default [55].Apple sold 192.7 million smartphone units that year [56], so Google paid U.S.$5.18 per new iPhone user.If we use Apple's cumulative smartphone unit sales, which were 442 million by 2014, then this would mean that Google paid U.S.$3.35 per iPhone user.According to a 2020 U.S. Congressional investigation into competition in digital markets, "Apple also reportedly made $9 billion in 2018 and $12 billion in 2019 to set Google as the default search engine on the Safari browser" [17].Apple's cumulative smartphone unit sales were 888 million and 948 million in 2018 and 2019 respectively, meaning that Google paid U.S.$10.13 and U.S.$12.65 respectively per iPhone user.
More generally, assuming that Google's TAC rate, which was around 23 percent of advertising revenues in 2014 and 22 percent in 2018 and 2019, reflects the cost of revenues for all users and their personal data to Google, then this would mean that the value to Google of each iPhone user and their personal data (cumulative) can be identified as: ) While these numbers provide a useful indication of the value of personal data to Google, there are caveats to these calculations: (1) they only reflect iPhone devices, not all Apple devices; (2) it is probably better to use cumulative users in the calculation, since Google benefits from access to all users; and (3) total value to Google reflects total advertising revenues from those users and their personal data.
As can be seen from this attempt to work out the value of Google's data assets, there is a significant range in possible valuations across different approaches.
VI. DISCUSSION: POLICY IMPLICATIONS Given the market dominance of Google in online advertising, clarity about the value of Google's data assets is a matter of public and policy importance [1], [7], [17], [24], [28], [57].Several reports by policymakers have alleged anti-competitive advantages that Google benefits from when it comes to its collection and use of personal data, especially in the adtech sector [28], [30], [31].They also make allegations about the impacts of Google's market dominance on consumers and citizens.How much Google benefits, however, is not always clear as there is so little transparency around data value.
To improve market competition and consumer well-being, we recommend improving the transparency and accountability of personal data holdings by firms like Google.
1) Transparency: It is currently difficult to assess exactly how digital personal data are valued economically by firms like Google.Clarifying these valuation processes is important for the public and for policymakers: international standards setters are currently trying to do this by establishing a consistent way to value digital data in national accounts [58].We think clarity about data value will force greater transparency about where our personal data sits and who controls it, as well as better privacy and economic oversight of data use.2) Accountability: One way to reduce Google's market dominance and ensure that the benefits of personal data use flow to users and citizens is by ensuring that companies are held accountable.Researchers have argued that Google's adtech practices distort competition and are monopolistic [35].Transparency would ensure that firms like Google could be better held to account for their collection and use of our personal data, especially where it impacts competition and privacy.3) Taxation: Data collection and analysis can lead to significant social benefits if those benefits are shared responsibly.A mechanism for ensuring a more equitable distribution of benefits from our personal data follows from transparency and accountability: namely, taxation reform.The widespread use of digital service taxes has been suggested to ensure that digital services income of large companies like Google can be taxed [59].These policy implications reflect broader efforts to adapt our societies and economies to the environmental, social, and governance (ESG) risks resulting from global challenges like climate change.Moreover, they build on current international policy initiatives to standardize the treatment and valuation of digital data as an asset, especially in national accounting [21], [58].All these efforts are likely to provide further momentum to working out how to better govern the economic and social impacts of Big Tech firms like Google (e.g., privacy).

VII. CONCLUSION
Despite the possibility of calculating the personal data value of Google's data assets using a range of valuation approaches, our analysis in this paper showed that these valuation approaches are always (and perhaps necessarily) partial.
In particular, our analysis ended up being based on using proxies and estimates triangulating across a range of empirical sources (which are often highly ambiguous or also partial).It is notable that Google's annual financial reports and earnings calls rarely contain information about personal data and its value, presumably because of the focus Google has on monetizing users rather than data [5]; it is also likely that Google's own approach (alongside interested social actors like investors and financial analysts) to personal data is framed by assumptions about what personal data are and what makes it important or not as a political-economic object [57], [60].
Consequently, it is worth considering what other empirical methods could be employed to get at the value of personal data (when conceptualizing them as assets) as understood by social actors themselves: data's 'vernacular value'.For example, we initially wanted to get access to the purchase price allocations / agreements (PPA) of mergers and acquisitions (M&A) by Google, since these should reflect the value allocated to different assets by Google during the M&A process.PPAs are filed with the USA's Internal Revenue Service (IRS) as they can have significant impacts on corporate tax rates and tax planning for firms like Google [61].A PPA consists of a document allocating the value of the assets of a merged or acquired firm, divided into seven different asset classes.We would expect that the value of digital data would be allocated to one of these asset classes as part of the M&A negotiation (probably Class VI).However, while Google's annual reports do discuss "purchase price" of acquisitions, there are few details about the allocation across different asset classes and the IRS documents themselves are not public documents.
While our attempt to access PPAs was unsuccessful, we considered other possible ways to investigate digital data value.These methods might include investigating -through qualitative interviews -how digital data are valued during private investment deals, the initial public offering process, M&A deals, loan arrangements where data is used as collateral, and in bankruptcy proceedings.We think all these 'moments of valuation' represent useful empirical sites for further examination of data's value and valuation.

Fig. 1 .
Fig. 1.Google's Advertising Ecosystem (post 2018).Sources: various, including ClearCode.ccAdtech Book, available online at https://adtechbook.clearcode.cc/, and[34],[35],[40],[41].Red lines indicate Google properties.Network).Critical to monetizing traffic (i.e., users) is the use of personal data to provide better targeting of individuals by advertisers, which Google facilitates.In a 2011 Q1 earnings call, Google executives stressed the importance of the "signals" coming from user traffic across its suite of products and services.These signals represent the user and personal data collected about users that underpins its online advertising business.Notably, Google executives define this user/user traffic as an asset in and of itself."Wedo see social as very important.Google uses well over 200 signals in terms of how we think about [Search] ranking today.And when we think about identity and relationships, those are our key signals that can and should be integrated in the experience.So it is important, but it's one of the many that we use.In terms of assets that apply to that, we do have a very, very large number of users coming to our door every day.A considerable percentage of them are logged-in users that are using multiple of our products.So there is a large variety of signals that we'll be able to use with user support and users seeing value from it to make the overall experience better."[43][Emphasis ours] A key and ongoing concern of Google since at least its IPO, has been something called "traffic acquisition costs" (TAC)[44].According to their 2022 10K Annual Report, TAC consists of:1) "Amounts paid to our distribution partners who make available our search access points and services.Our distribution partners include browser providers, mobile carriers, original equipment manufacturers, and software developers.2) Amounts paid to Google Network partners primarily for ads displayed on their properties"[42].In 2022, TAC amounted to U.S.$48,955 million and is defined as a "cost of revenues", which represents the direct costs of selling a product or service to a customer; in