Abstract
The global food system is characterized by market concentration and oligopoly. In our article, we focus on the most powerful input supply and machinery companies and analyze how these firms create value, both economic and otherwise, from big data. In digital capitalism, data is valorized across sectors; personal data is aggregated into large-scale datasets, a practice that feeds economic concentration and monopolization. Big data also has become central to the business model for agricultural companies; it is a claim made by the companies themselves. Yet, little is known about their specific strategies to do so. We aim to fill this gap, asking how is agricultural data transformed into value by the most powerful agribusinesses and ag-tech firms?
Through the lens of assetization, we examine corporate strategies for transforming agricultural data into value. We draw on literature from food studies, specifically political economic analyses of the historical practices of agricultural corporations, as well as literature from critical data studies that investigates data as an asset. For our analysis, we rely on a variety of gray literature and public-facing documents: financial documents, sustainability and shareholder reports, terms of use, license agreements, and news articles. Our results contribute to the critical data studies literature on agricultural big data by identifying three main strategies of assetization: securing relationships and dependence, price-setting and data sharing, and product development and targeted marketing.
The strategies have socio-ecological implications; our results indicate the reproduction of asymmetrical power relations in the agri-food system favoring corporations and the continuation of long-standing dynamics of inequalities.
Keywords
Introduction
The global food system is characterized by intense market concentration and oligopoly; a small handful of firms exerts control over research agendas, policies, and the everyday lives of farmers and consumers (Clapp, 2021; Howard, 2021). Based on their market share, we focus on powerful seed and chemical (e.g. Syngenta, Bayer/Monsanto) and machinery (e.g. John Deere) firms and analyze how these firms create value, both economic and otherwise, from big data. Many of these companies appear to be shifting their business practices to focus on data as a key asset and driver of profit. In our era of digital capitalism, data is valorized across sectors; personal data is “harvested” and aggregated into large-scale datasets (Zuboff, 2020). The use of personal data for corporate gain has sparked lively public debates and academic interest not least because corporate data practices appear to feed economic concentration and monopolization. Google, Facebook, and Amazon's data-related business models and data valorization methods are well-known (Birch et al., 2021; Mayer-Schönberger and Cukier, 2013; Prainsack, 2019; Sadowski, 2019). Yet the data practices of agricultural corporations have yet to be empirically examined.
That big data has become central to the business model for agricultural companies is irrefutable (Bronson, 2022); it is a claim made by the companies themselves. Syngenta's CEO, for example, has said: “Before, we sold pesticides, seeds, and fertilizer. Now we’re a [digital] farm services company” (ETC Group, 2022: 21). John Deere expects that by the end of the decade, 10% of the firm's annual revenue will come from digital platform user fees (Tita et al., 2022). Bayer/Monsanto has been building up one of the largest agricultural databases in the world, and in 2014, Monsanto acquired the weather analytics company The Climate Corporation; this is likely why Bayer acquired Monsanto in 2018, in the largest acquisition in global history.
In recent years, Wysel et al. (2021) has put forward a descriptive analysis of corporate practices surrounding agricultural data, and Klingenberg et al. (2022) has conducted a case study on an agricultural value chain to evaluate the impacts of digitalization on valorization. One key finding of this work is that platforms operated by non-agricultural actors like IBM or Alphabet Inc. are well positioned to create value from farm data. Trail (2018) outlines ways this might be done, highlighting the data ownership issues that sensor-based tractors and “precision” equipment could raise and the resulting need for corporate uses of farm data to be regulated.
Aside from this work, almost no work has empirically detailed the specific strategies these firms will use to profit from farm data. The absence of empirical examination of the exact uses of data collected from farms is a problem; scholarship has indicated unique features of the agricultural sector in the era of big data that deserve unique critical data studies (see Bronson and Sengers, 2022; Stone, 2022a). Scholars note that while anyone can use social media platforms in exchange for their data being collected, farmers, even after contributing their data without remuneration, must pay to access the data-driven advice offered by agricultural firms. And though big tech companies like Facebook are now considered some of the most influential in the world, the oligopoly of agricultural corporations has influenced the fate of national agricultural policies, farmers, and consumers longer than any other power (Birch et al., 2021; Bronson and Sengers, 2022). Thus, the question of what exactly input supply companies do with big datasets configured from farm-level data is worth examining empirically.
In this article, we ask: how is agricultural data transformed into value by the most powerful agribusinesses and ag-tech firms? To answer this question, we draw on critical food studies and science and technology studies literature on historical corporate practices related to profit generation. We analyze strategies of big data assetization as they intersect with iniquitous political and economic relations in the food system, which have been described at length. We rely on analyses of the historical practices of agricultural corporations, as well as critical data studies research that investigates data as an asset outside of the agricultural sector. We also draw on gray literature and public-facing documents related to input supply corporations: financial documents, sustainability and shareholder reports, terms of use, license agreements, and news articles. Based on empirical examination, we identify three valorization strategies for agricultural big data: Companies, we find 1) use data to lock farmers into relationships with particular firms; 2) use big data analytics for price-setting and hedging, and they also profit from the sharing of big datasets; and 3) use big data internally, to drive product development, and for targeted marketing. In the end, we argue that big agricultural data gives agricultural firms unique and incommensurate advantages over farmers and consumers and that harvesting agricultural data is thus likely to enable increased corporate concentration in agriculture.
Ultimately, our article fills a gap in empirical analyses of agricultural firms’ big data practices and thereby adds specificity to scholarly and public conversations on the societal implications of agricultural big data practices. A long-standing feature of the agri-food system is economic asymmetries related to new technologies, which agribusiness companies have for centuries used to dominate markets and create relationships of dependence with farmers (Clapp, 2022). Today, control over data flows may reproduce existing power relations and socio-ecological inequalities in agriculture at the expense of data and environmental justice (Hackfort, 2021). In the service of a more just and sustainable food system and a broader conversation in critical data studies, we believe that a systematic overview of the strategies involving big agricultural data is of the utmost importance.
Conceptual framework and methodology
The guiding framework for our analysis is the concept of assetization, which was developed to account for the political economy of “technoscientific capitalism” (Birch and Muniesa, 2020), where income streams are increasingly generated from financial resources and properties, which are intangible. An asset is “something that can be owned or controlled, traded, and capitalized as a revenue stream, often involving the valuation of discounted future earnings in the present” (Birch and Muniesa 2020: 2). This framework is appropriate to our research because we focus on processes of extracting value from agricultural data in both the present and the future. Birch and Muniesa (2020) comprehensively argue for assetization's suitability as a conceptual apparatus for studying the generation of value—both economic and otherwise—from data in relation to larger political economic relations. Our argument builds upon theirs, providing novel empirical work that demonstrates processes of assetization in the context of agricultural big data.
Assetization denotes the collection of data with potential future value; data assets are not necessarily bought and sold directly on commodity markets in the same way wheat is, for example. Rather, data is collected and deemed valuable and worthy of investing in based on its assumed ability to generate future income, revenue, and returns. To derive these revenues, firms must commit sustained attention and effort to data assetization—a process that Birch et al. (2021) delineate in the context of Big Tech firms’ assetization of personal data for the purposes of locking in future users, targeted advertising, and other monetization strategies.
Assetization also draws attention to the fact that data assets are legal constructs underpinned by ownership and control rights (for instance, intellectual property rights), regulatory and market devices, and practices such as exclusivity agreements (Birch 2020: 3). Access to, control of, and use of data are granted and can be limited through legal rights, agreements, and licenses. These modes of ownership and control further differentiate the process of assetization from the process of commodification.
Lastly, data assets are not easily reproducible or substitutable without access to the original data source; they thus encourage monopoly and oligopoly by putting new entrants to the marketplace at a disadvantage while incumbent firms grow increasingly powerful. In this context, as they mediate relationships between users and corporations, digital platforms provide infrastructure that is crucial for the assetization of data. Assetization processes have been analyzed in the contexts of personal data (Birch et al., 2021) and the gig economy (Doorn and Badger, 2020; Gregory and Sadowski, 2021). As we know from empirical studies of Amazon and Google, in the age of “surveillance capitalism” (Zuboff, 2020), such companies use online behavior data points to predict users’ preferences and future behavior; these predictions are then sold to third-party clients—for instance, advertisers. Irrefutably, both personal data and the platforms that collect it have become profitable assets. Zuboff (2020) has thus defined personal data as a “new asset class.”
Amid this breadth of high-quality scholarship, however, there is a lack of attention paid to the assetization of agricultural data. Unique to the context of agriculture, machinery companies’ sale of “precision” equipment (like the new John Deere tractors) provides a route by which they can collect data from farms. What is meaningful—and unique to agriculture—is that some of the most powerful, long-standing, oligopolistic corporations have recently been turning to these strategies of agricultural data assetization (Bronson and Sengers, 2022). The concept of assetization, however, has not yet been used to direct empirical examinations of these practices, so it is here that we focus our attention.
To guide our analysis and our selection of firms for study, we draw on critical food studies. As scholars in that field have illustrated, agribusinesses have become increasingly consolidated, and several mergers and acquisitions during the last two decades have resulted in enormous market concentration in the two most important agricultural sectors. Four companies now dominate the chemical and seed markets: Bayer/Monsanto, Corteva/DuPont and Dow, ChemChina/Syngenta, and BASF (Clapp, 2021; ETC Group, 2022). In the agrochemical sector alone, these four companies combined make up 65.8% of global sales, and for commercial seeds, ChemChina, Bayer, Limagrain, and Corteva account for 53.2% (Howard, 2020). Like the farm inputs and agricultural commodity industries, through mergers and acquisitions, the machinery manufacturing market has also become characterized by concentration in the last century. Today, just a few firms possess a major share of the market (Clapp, 2022).
Of these, the largest agricultural machinery companies are included in our analysis—John Deere, CNH Industrial, AGCO, and Claas. By now, every large agrochemical or agricultural machinery company has acquired or developed at least one big data-based farm management platform.
To investigate how these firms transform agricultural data into value, we analyzed a variety of documents, including license agreements, which empirically speak to ownership and control rights as key elements allowing for assetization. We also analyzed background material on the companies and their products and services on websites, including their sustainability reports, financial reports, and earnings calls. Financial reporting is only mandatory for public companies, so in some cases, especially start-up ag-tech companies, the financial documentation we were able to find and analyze was limited. Additionally, we reviewed academic work and gray literature, including NGO reports and newspapers, which detail historical economic practices among agricultural companies, such as discriminatory pricing and price-setting. We also analyzed the digital agricultural platforms and services themselves in depth wherever possible. While our analysis primarily concentrated on the largest agribusinesses, where relevant, we also included insights on other corporate actors, their subsidiaries, and start-ups they have acquired (see Table 1 for an overview of the analyzed material).
Analyzed background material.
The background material was analyzed using software-based qualitative content analysis (MAXQDA). Coding was an iterative process of deductive and inductive analysis, beginning with a list of key terms (such as “ownership,” “access,” and “third parties”) compiled largely from other studies of big data (critical data studies) and of data assetization outside of the agricultural context.
Corporate strategies of data assetization in agriculture
Through analysis, we have arrived at two overarching arguments. The first is that any attempt to systematically examine what agribusinesses do with agricultural data is impaired by legal mechanisms that obfuscate data practices, datasets, and algorithms: copyright, intellectual property law, trade secrecy law, and arbitration agreements all allow for proprietary technologies and a high degree of vagueness and opacity. This is a finding in and of itself; such obfuscation prevents critical analysis and the kind of oversight that the equitable governance of technology requires. Our second, broader argument is that data itself is very likely an asset for agricultural firms, which now uniformly include big data-based services in their portfolios. These companies are dominant in not just their traditional areas of technology, but also in offering digital farm tools, particularly big data-driven decision support platforms. Apart from a few start-ups that have gained a foothold in the market, the development and provision of digital technologies and services is dominated by the same handful of companies that supply agricultural seeds, biotechnology, pesticides, fertilizers (including Syngenta, Bayer, BASF, and the Norwegian multinational Yara), and farm equipment (e.g. John Deere and Claas). Syngenta has heavily invested in takeovers of farm management platforms, and it claims to be the only company with access to the leading farm management platforms in the world's top four agricultural markets: the United States, Brazil, China, and Eastern Europe (Syngenta, 2023). Bayer is also an important player here given its ownership of Climate FieldView (CFV), which in 2018 had more than 100,000 registered clients who together farmed on more than 120 million acres in the United States, Canada, and Brazil (Bayer's CFV, 2022b). We can confidently assert that for agribusinesses, data now generates revenue alongside input and machine supply. In line with our assetization framework, companies appear to be investing in data collection and analytics for their own sake, quite apart from using datasets to generate the advice provided by the farm decision support tools licensed to farmers.
Below, we outline three strategies that firms use (or are likely to use) to generate value from agricultural data (these are detailed, with examples, in Table 2). We also refer back to the conceptual underpinnings of our article to discuss how these strategies can be understood as assetization.
Assetization strategies of agricultural big data (references are given in the text).
Securing relationships and dependencies
Agribusinesses have historically used a variety of mechanisms to corner the market on agricultural inputs and machinery, establish exclusive relationships with farmers, and create different kinds of lock-ins and forms of dependency from within their economic–technical ecosystems. They have created relationships of control with farmers by cultivating their dependence on proprietary technologies—Monsanto, Syngenta, and DuPont—for instance, offer discounts and rebates to those who purchase their seeds and use their complementary herbicides and pesticides (Clapp, 2021; Curry, 2023). Agribusiness companies have also used intellectual property rights (in the form of, e.g. patents and trademarks) to tie farmers to seed systems by making it nearly impossible to switch systems without legal repercussions (Bronson, 2015; Shiva, 1991). Firms use leasing and financing arrangements to create lock-ins and dependencies among farmers. For example, John Deere, one of the world's largest manufacturers of farm equipment, offers financing for its equipment, making it easier for farmers to purchase its products. However, this also creates dependencies, as farmers are more likely to purchase future equipment from John Deere to maintain their financing relationship (John Deere US, 2023b).
One well-known example of dependency in the digital era is the “digital lock” on John Deere tractors, which contain sensors for collecting data and sending it to cloud-based infrastructures. Legally, this makes them digital rather than mechanical tools, like mobile phones. As such, their use is controlled by a license agreement under which they are leased, not owned, and may only be repaired by certified mechanics; farmers may not fix tractors themselves, nor may they access the data the machines generate (Solon, 2017). After years of contestation by activist organizations (see IFIXIT.org), John Deere has reached an agreement with the American Farm Bureau Federation to partially remove the lock. Yet, as observers argue, this partial concession might erode efforts to bring about broader, legally binding measures that would be stricter in terms of data sovereignty (PIRG, 2023). The current legal architecture, consisting of lax regulation and the absence of binding protections for farmers, is the basis for data assetization, as it leaves room for unknown future uses of data, which might benefit corporations by prioritizing their access to datasets over that of farmers.
Legal tools and mechanisms like the copyright restrictions built into terms of use or license agreements are not the only techniques that appear to further relationships of control and dependence between farmers and firms. Our analysis shows that a lack of interoperability among the most powerful companies’ data-based decision support platforms also works in this way, erecting barriers that prevent farmers from changing to (or even trying out) different systems. Companies such as John Deere have direct access to the data collected by the sensors and GPS systems installed in its machines, and increasingly, they also control that data. Farmers who buy new tractors cannot opt out of collection—crop data is automatically transmitted (along with other data) to a cloud-based infrastructure and used by the company's decision support platform and its proprietary algorithm. The company combines data from the level of individual farms with broader environmental datasets on weather and soil to generate recommendations that are sold to farmers. But notably, once a farmer's data has been uploaded, even if they were granted access to it, the lack of interoperability between companies’ platforms would make it difficult for them to retrieve it and switch to another system (e.g. Semios, 2022). The lack of interoperability and compatibility among software and hardware systems from different providers is a persistent problem for farmers in many countries (e.g. Atik and Martens, 2020; Higgins et al., 2017; Stock and Gardezi, 2022), one that constrains their ability to choose systems based on their changing needs and locks them into a relationship with a single tool and company—and that company's inputs. This lack of interoperability drives the monopolization of data, which is a key feature of assetization.
Alongside data lock-ins, companies use “soft” lock-in strategies to create path dependencies for farmers. While some companies can access data directly through the devices built into their machinery, others dangle carrots in front of farmers in exchange for access—free “starter packages,” for instance, or BASF's xarvio Field Manager app, which is offered for free for a limited time (BASF xarvio, n.d.). Through such incentives, as with biotechnology seed-chemical packages (Shiva, 1991), farmers are likely to become path-dependent on the technological systems of the companies whose machinery and software they initially purchase. Such strategies are not entirely new: elaborate incentive structures have long encouraged farmers to rely on bundled corporate products. As Kelloway (2022) reports, through its BayerPLUS Rewards program, Bayer offers farms a per-acre cash rebate for buying its products; buying chemicals in addition to seeds can result in savings of several dollars per acre. Such rewards can save commercial farms operating on more than a thousand acres thousands of dollars (BASF, 2023a). Similar rebate programs are offered in relation to data. With its Grow Smart Rewards program, BASF offers cash-back rewards to farmers using its digital agricultural platform in combination with the company's pesticides (Arc, 2018). Bayer's CFV, a big data analytics platform, is offered “free” to users who also sign up to BayerPLUS Rewards (Bayer's CFV, 2022a; Stock and Gardezi, 2022). In 2018, CropLife reported that to widely distribute the platform and achieve a high level of market reach, Bayer/Monsanto used rebate incentives to encourage not just farmers but also retailers to adopt digital technologies. Because the latter are offered rebates should they reach certain sales targets, it likely makes financial sense for them to occasionally sell products like CFV at a loss (Kelloway, 2022).
We see evidence of a creeping adoption of data technologies tied to rebate programs, mirroring Bayer's offer of tiered CFV subscriptions beyond the free basic platform, which incentivizes incremental adoption by providing additional features and services at marginal cost (Bayer's CFV, 2023). BASF's xarvio Scouting app is also free, and xarvio Field Manager is free for two fields (BASF xarvio, 2023). Companies in the agricultural industry commonly offer basic, entry-level tools for free to attract farmers and introduce them to products; these new clients then prove a return on investment. However, after they have already bought into, adopted to, and fully incorporated particular technological systems, early adopters may be required to pay for assistance or to access more advanced features and services.
Furthermore, corporations such as Syngenta, Corteva, BASF, Bayer, John Deere (partnering with Cargill), and Microsoft (partnering with Trimble) are increasing the marketability of their platforms by integrating current trends around climate-smart agriculture, sometimes using terms like “regenerative farming,” which are meant to signify sustainable food production (BASF, 2023b). For example, Bayer offers an “annual cash payout” for participation in “carbon farming” programs that use the CFV platform to “verify” farmers’ practices (Bayer AG, 2022). Many firms have launched programs to pay farmers for adopting carbon sequestering techniques. These emerging carbon credits are designed to be closely tied to corporate digital monitoring and control tools, thus enabling the flow of data directly into specific company platforms. Proprietary business models can mean that farmers enrolling in a company's carbon program simultaneously sign up to use its products, as in the case of Nutrien (FarmRaise, 2022). Moreover, data collected from farmers provides valuable information on their production practices and yields. To verify carbon credits, detailed information about planting is collected not just from technology at the level of farms and their equipment, but also from satellites. Some private carbon programs expand their user base by requiring farmers to directly upload information to proprietary digital platforms. These companies may take ownership of farmer-generated carbon credits, paying farmers either a portion of their sale price or at a fixed rate. Once they have sequestered carbon, farmers may need to sign long-term contracts to store it, giving payment platforms exclusive access to their data. This consolidates farm data among the largest agribusinesses, which already control large datasets and use them to create and optimize their algorithmic decision support products (Kelloway, 2021); it also creates barriers for newcomers, who may not have access to large datasets. On the surface, the implementation of systems of verification appears to be a win–win situation, a shift toward “green practices” that pays back farmers and ensures companies future revenue from data. However, such practices further strengthen ties between farmers and specific companies, eventually working as lock-ins for farmers and further strengthening oligopoly within agriculture. It is these processes of collecting large amounts of data via established, monopolistic platforms that make data assetization possible in the first place, as the basis upon which new business models can then be developed.
Another example of a soft lock-in is Bayer's “outcome-based” pricing program, in which the firm sells seeds and agrichemicals based on performance guarantees like specific crop yields or levels of weed reduction (Gullickson, 2019). Similarly, farmers who pay for BASF's xarvio Field Manager buy not a product but guaranteed, predetermined yields linked to specific, data-based recommendations (Burwood-Taylor, 2021). While this strategy presents a kind of risk-sharing between farmers and corporations, if the product outperforms the firm's predictions, the farmer loses a portion of the additional profits—potentially around half, according to one report (Gullickson, 2019). While this is sold as a tool to help farmers “manage risk” more effectively, it also helps BASF build stronger business relationships with farmers and allows for better pricing control. There is mounting concern over this corporate strategy: “Farmers and antitrust scholars worry that goliaths such as Bayer will use this data-driven pricing program to further squeeze farmers and to lock more growers into the behemoths’ product bundles and digital agriculture platforms” (Kelloway, 2020, n.p.).
The practices outlined above all help link farmers to agricultural inputs and the firms supplying them, and in the digital era, they confer an additional advantage on the firms: digital decision support tools are typically based on machine learning, which uses big data, and it is claimed that the more data they collect, the more their predictive capacity improves. Rebate and incentive mechanisms thus might translate into additional value for companies like Bayer, which collect ever more farm data, leading to a competitive advantage in this new market. Thus, in the agricultural context, firms are investing in data collection and infrastructure (including legal infrastructure) on the basis of the (partly unfulfilled) future “promise of precision” (Hackfort 2023: 7) that more data allows for more accuracy in predictions and recommendations generated from big data. These predictions are not just sold to farmers; they also have unspecified future uses for the corporations holding the data (via copyright restrictions), which is a key characteristic of data assetization.
Price-setting and data sharing
From historical research, we know that companies supplying farmers with seeds and chemicals have engaged in discriminatory pricing, arbitrarily setting higher prices for demographics or regions which are seen to depend on these products (Clapp, 2022). Farmers’ organizations like the Canadian National Farmers Union have collected evidence that input suppliers set prices to increase their profits at the expense of farmers, such that they capture any marginal returns a farmer may have seen in years where certain commodities traded high on the international market. This organization highlights how farmers are put at a disadvantage by the “corporate industrial complex” (NFU, 2020) and calls for an investigation into the practices of firms that are part of fertilizer oligopolies (Nutrien, Yara, CF Industries, and Mosaic; NFU, 2022). In Germany, 1500 farmers have joined a class-action lawsuit against agricultural wholesalers for illegal price-setting in the sale of pesticides; among the defendants will be Germany's largest agricultural retailer, BayWa (Schaal, 2023).
With agricultural data collection, it seems likely that powerful agribusinesses will use farm data (e.g. on yields) to generate additional predictions about particular commodities and, following the same logic they have historically employed, use this insight to set prices (Harris, 2022; WSJ, 2012). Although we did not find explicit evidence that corporations are using agricultural big data for price-setting (or selling it for this purpose), our research shows that their legal policies generate enough room for them to do so without disclosing their actual data practices. Bayer's current privacy statement forecloses seed pricing based on the personal data of farmers using the CFV platform but does not prevent the use of farm data, which suggests that the company is keeping open the possibility of selling farm data to third parties and using it for price-setting: “We collect information about you and your farm operation for the following reason: (…) For advertising and marketing purposes. (…) However, we will NOT use your personal data to price seed products to you or others, other than to offer you product discounts or make you aware of other offers or programs for which you may qualify, and we will not share your personal data with any non-affiliated third-party vendor or website owner for their own marketing purposes without your consent” (Bayer CFV, 2022). Bayer seems to have changed its data policy recently; the text of the policy from 2019 clearly states that the company will not use any data (aggregated or otherwise) for seed pricing or to make speculative commodity trades (Bayer CFV, 2019). The 2021 version for the US market, however, states that Bayer does not use aggregated personal or farm data for any other purpose than to lower its business risk. This keeps the door open for assetization through speculative financial activities generated from the use of both types of data—personal and farm. These activities are not made transparent in the legal agreement. Notably, Bayer's End User License Agreements (EULAs) further allow the company to “access, use, reproduce, display, modify, and prepare derivative works based on your Customer Farm Data in order to provide the CFV Services and related support to you, for our internal operations and research and development purposes, and for other purposes set forth in this Agreement” (Bayer CFV, 2021). One of these “other purposes” is to use both non-aggregated and aggregated personal data and customer farm data for hedging: “It is our policy not to use Customer Farm Data or Aggregated or Anonymized Information derived from Customer Farm Data to make speculative commodities trades, other than hedging we may do during the normal course of business to manage risks associated with our own seed/commodity production operations” (Bayer CFV, 2021). Hedging involves investing using various financial instruments—including future contracts, purchases, and commodities sales—and, apparently, agricultural data, which the company uses without explanation. These speculative financial operations based on collected farm data are the most obvious form of agricultural data assetization.
Our analysis of the use agreements for data-based tools and platforms from several companies shows that, along with the internal use of data for pricing and investment, data sharing seems to be common across ag-tech companies, even if they confine sharing to aggregated and anonymized datasets (e.g. John Deere, Claas 365 Farm.net; Corteva Agriscience, CNH Industrial, Syngenta, and Farmers Edge). For instance, the CFV Privacy Policy from 2019 explicitly states that the company will not share or sell any data (Bayer CFV, 2019). But in the present version of the policy, such statements are omitted, the wording instead allowing for the sharing of any non-personal data—farm data, for instance—with third parties without farmer consent (Bayer CFV, 2022). After aggregation, or if technical specifications require an exchange of data with external service providers, companies retain the right to pass on anonymized data to a third party without informing the farmer or obtaining their prior consent, beyond their agreement to the technology's license agreement. John Deere's service agreement states: “We may combine your anonymized data with data from others and include your data in anonymized datasets. We may also share, in aggregate, statistical form, non-personal information with our partners, affiliates or advisors” (John Deere US, 2022). Who these partners, affiliates, or advisors might be is not elaborated on. Corteva Agriscience states: “We may use and disclose Other Information for any purpose, except where we are required to do otherwise under applicable law” (Corteva Agriscience, 2019). “Other information” in this case means data that does not reveal or suggest the user's identity—browser and device information or vehicle usage data, for instance, but also data about farmers’ “agricultural operation” (Corteva Agriscience, 2019). Likewise, in using the BASF xarvio Field Manager app, farmers give their consent to “the reproduction of this Licensed Data and to grant third party access without limitations (including, but not limited to, for the purpose of continually improving algorithms in the field of digital agriculture).” The company uses the term “Licensed Data” to mean information including “but not limited to data on crops and types of plants, the planting date, field boundaries, soil preparation measures, crop protection measures including the products used, doses and times, and contact data and statistics regarding the use of the app (‘Licensed Data’)” (BASF xarvio, n.d.).
The relevance of these legal agreements became obvious in a recent case where the sharing and use of farm data by third parties for price-setting was widely suspected. The “farm real estate” start-up Tillable entered a partnership with The Climate Corporation, a farm data analytics firm owned by Bayer. Tillable's online service brings together landowners and farmers to rent out land that previously was leased directly to farmers. This approach has caused concern among farmers, and things escalated publicly when CFV users disclosed having received from Tillable unsolicited offers to rent the land at a specific price creating the impression that Bayer and Tillable were misusing farmer data to set land prices (Janzen, 2020). After heavy public criticism, the two companies canceled their partnership (Charles, 2020). Even if Tillable has publicly denied having accessed the data from Bayer's platform, this case clearly shows the potential for farm data to be used for private gain in situations where doing so is not legally foreclosed. This again illustrates key mechanisms of data assetization in the agricultural sector, made possible, on the one hand, by farmers’ lack of rights in relation to their data (or data they help to collect) and, on the other, by firms establishing particular practices via use and license agreements that grant them exclusivity to big datasets, which is likely to eventually serve their assetization interests.
One peripheral industry in particular stands to benefit from legal agreements leaving open the possibility of aggregated datasets being shared or sold: insurance. The importance of managing risk in agricultural production is growing considerably. In the era of big agricultural data, insurance firms can not only look to historical weather data and a farmer's history—say, the average growth, in bushels, of wheat on a farm in a particular region—but also use big data and algorithms to predict the likelihood of losses almost in real time, using the information to validate present investments and drive future ones. Theoretically, “data-driven” predictions could enable these companies to make shrewder investments. Agribusinesses have formed strategic alliances with insurance firms such as Farmers Edge and Munich Re (Farmers Edge, 2022). Advertising aimed at investors boasts about such alliances and the uses of predictive analytics and big data; for example, IBM advertises its artificial intelligence, Watson, as participating in this predictive reinsurance industry (Bronson, 2022: 132). Furthermore, Farmers Edge Inc. recently launched a wholly owned subsidiary, DigiAg Risk Management, which will provide farmers across Canada with new parametric insurance products (Farmers Edge Inc., 2022). With satellite data, insurers can now more precisely analyze the condition of crops before and after they sustain damage. However, in certain cases, including countries where relevant data is either unavailable or not sufficiently transparent, traditional crop insurance with individual loss assessment is uncommon. Here, index-based insurance is offered as a solution. This insurance, based on digital technologies that support the modeling of crop yields using remote sensing, data analytics, and artificial intelligence, relies on a yield index to ensure crops “automatically pay out if the actual yield falls below an agreed percentage of the yield guarantee” (Munich Re, n.d.). One online advertisement for the insurance company Swiss Re reads, “Reinsurance and insurance companies will price their products better. Watson reads millions of pages of data, including unstructured data like discussion notes, contracts or tickets to help Swiss Re assess risk factors and make more informed decisions regarding price-risk accuracy. This helps Swiss Re reduce costs while increasing quality” (IBM, n.d.). Our analysis shows that it is very difficult to follow data from point of collection to management and use by private businesses; however, license agreements include clauses that would allow farm data, once aggregated, to be used for price-setting. Use agreements allow farmers access to their raw data only until it is “aggregated or anonymized.” Aggregated datasets and information generated from them are inaccessible, fully owned by the firms and thus open to any of the methods of assetization outlined in this article.
Product development and targeted marketing
In the analog era, companies predominantly used internal research and development to drive products; today, big data is essential to the process. For example, John Deere's See & Spray technology uses a machine-learning algorithm to identify weeds, allowing for site-specific herbicide spot spraying rather than broadcast spraying techniques. As with many machine-learning algorithms, the images produced by See & Spray's cameras are used to improve its weed identification models (John Deere, 2023). The collection of data for the sake of constantly improving algorithmic models is crucial to the ag-tech assetization business model. Another example of data use for product development and improvement is “predictive maintenance,” a service offered by machinery manufacturers such as John Deere and Claas. Most John Deere construction equipment is outfitted with telematics software that captures various machine metrics, such as location, usage hours, temperature, and fuel consumption. This data is transmitted to the cloud in real time and purportedly used by John Deere to improve the performance of its products by allowing analysis of data on driver behavior or engine load (Prime, 2022). Within the predictive maintenance framework, the company provides incentives for farmers to allow data about their farm operations and land machines to be tracked in exchange for the promise of a perfected machine. When problems are detected, “expert alerts” are activated (John Deere US, 2023a) and farmers can use the Service ADVISOR Remote tool to enable dealers to remotely access their machine through the company's Operations Center or the JDLink Dashboard (John Deere US, 2022). This predictive maintenance framework is based on contracts that offer, as a service to farmers, to monitor the condition of equipment in use and provide for timely maintenance. While the farmers who adopt it seem comfortable with this model, in effect, they pay for it at least twice, with one of these payments made in-kind, through the free provision of data via their machinery. Yet such data clearly helps the company improve its products and gain a competitive advantage in the market. The collection and analysis of agricultural data to generate data revenues from predictive services is a mechanism defined as key for assetization by scholars working in big data studies outside of agriculture.
For companies like John Deere, an additional benefit of predictive maintenance is the further strengthening of its relationships with its clients—or, put differently, of farmers’ dependence on the company's machinery and services. Crop and other agronomic data are also used to improve existing products and services, make established business models more profitable, and bring new data-based services to the market. Corroborating this is our analysis of John Deere's privacy statement, wherein the farmer agrees that the company can access and use machine agronomic data and machine-specific data in “anonymized and aggregated form for statistical purposes as well as to improve or enhance the services provided under this agreement, develop additional or new John Deere products and services, and/or identify new usage types of equipment” (John Deere EU, 2020). Similar language is common across legal agreements in the industry. AGCO states that farm data will be used in “on-going product development and improvement initiatives” (AGCO, 2019). Bayer's CFV privacy statement similarly illustrates the company's collection of personal data “and data on farm operations” for “research and development purposes, such as to improve Climate and Bayer's agronomic or scientific knowledge [or] develop and improve our products and services” (Bayer CFV, 2022).
Companies have historically sold their products to farmers via conventional mechanisms, like trade journals, newspapers, and broader (e.g. web-based) advertising, and, in a more targeted way, by sending agricultural advisors to offer farmers advice. In the digital farm era, companies use machinery and agronomic data to improve and market their products and services; analyzing license agreements, we found that in their EULAs, firms usually grant themselves the right to use clients’ farm data to do so. For instance, by using BASF xarvio Field Manager app, farmers agree that “data on crops and types of plants, the planting date, field boundaries, soil preparation measures, crop protection measures including the products used, doses and times, and contact data and statistics regarding the use of the app (‘Licensed Data’)” can be used by BASF for “the development, manufacturing, enhancement, and/or marketing of products and services“ (BASF xarvio, n.d.). Companies now have access to detailed data on clients’ farming operations, including their purchases and use of commodities such as seeds, chemicals, and other inputs. Theoretically, this means that companies are better positioned to sell products by making more precise and relevant recommendations. Raven Inc.'s EULA states that advertisements may be targeted to users based on data that has been uploaded to the Raven Industries platform (Raven Industries, n.d.). Again, in the EULA, data is legally configured as an asset, granting the corporation control over the data and assuring future revenue streams from users.
Bayer explicitly pitches this targeted marketing strategy as part of its new assetization business model—one based on “predictive seed selection and placement” (Burwood-Taylor, 2021: 2)—which seems advantageous to farmers, who get more targeted advice. A few years ago, the then head of research at Monsanto described in a positive light the digital synergies that would emerge once Bayer and Monsanto's disparate product portfolios were combined as part of the 2018 merger, saying: “In a few years, we want to send farmers a satellite image of their field once or twice a day to track crop growth and give an early warning if there's a problem somewhere with fertilizer, water, disease, or insects. And then they can find a solution with Bayer products” (CGB, 2019: n.p., own translation from German).
One specific example of a targeted recommendation uncovered by our analysis is Bayer's Seed Advisor, a digital tool purportedly developed to select—from Bayer's own product ecosystem—the hybrid corn most suitable for farms’ specific locations (Bayer, 2018). Note that, according to its own press release, Bayer offers exclusively hybrid varieties that, unlike seed-stable plants, farmers cannot regrow but instead must purchase anew every year (Bayer 2019). Seed Advisor is primarily aimed at seed dealers and retailers, who pass on input recommendations to farmers. After the Bayer/Monsanto merger, The Climate Corporation's Seed Advisor now provides farmers with recommendations for “best-performing” hybrid seeds based on Bayer's seed genetics library and “regional seed performance data” (Claver, 2019: n.p.; CGB, 2019). According to the company, benefits for farmers are significant and could lead to considerable yield increases (Maurin, 2019).
Despite its benefits to farmers, critics see such targeted marketing (a form of data assetization) and precision recommendations as further locking farmers into a state of dependency and into companies’ product ecosystems. One criticism is that this targeted advice steers farmers only toward a particular company's proprietary seed varieties, limiting their options and potentially resulting in a lack of crop diversity. Critics also highlight the potential for agrochemical companies to use the data they collect to target farmers with offers and information which is misleading, inaccurate, or more beneficial to companies than to farmers. Overall, there is concern that companies with large amounts of data may significantly influence farmers’ decisions: “The greater the amount of data that an agrochemical company has at its disposal, the more targeted the bait offers and bogus information on the farmers’ screens” (CGB, 2019: 10–11; Maurin, 2019). According to the Open Markets Institute, “Farmer data collection and digital agriculture programs owned by agrochemical manufacturers result in farmers getting management recommendations from corporations with a vested interest in selling more agrichemicals” (Kelloway, 2022: n.p.). Notably, companies’ vested interest in advice promoting their own products is arguably less visible to farmers in the digital era: the advice (or, in CFV's case, “prescriptions”) of decision support platforms is putatively “data-driven” (i.e. algorithmically derived) and therefore has the connotation of sitting outside of human and economic interests. The distinction between a decision platform and a retailer or extension agent giving farmers advice on which inputs to use is the former's air of objectivity (Bronson, 2022). Put differently, many farmers are likely to trust in the decision platform's neutrality and follow its advice—in contrast to the skepticism they might feel toward any human farm adviser a company might send.
Discussion
Through the lens of assetization, we have examined corporate strategies for transforming agricultural data into value. Our analysis contributes empirically to the data studies literature by identifying three main agricultural data assetization strategies: securing relationships and dependence, price-setting and data sharing, and product development and targeted marketing.
Assetization in agriculture is largely based on gathering as much data as possible from different sources as a source of (future) value—and on securing the ability to do so. Scholarly work in data studies has outlined three ways to extract value from personal data collected online: data holders can access big data; data specialists can “mine” it for insights that strengthen their competitive position in the data market; and data strategists—that is, firms powerful enough to do so—can flexibly orient their corporate practices by extending data insights to as-yet-undefined novel applications (Mayer-Schönberger and Cukier, 2013). Our analysis shows that similar processes are at work in the agricultural sector, where already powerful input supply companies now have the means to assetize and profit from farm-level data. Big agricultural data is an intangible resource, an income-generating asset for those who control quantities of data and are able to further process it. Accordingly, agricultural companies, like other Big Tech firms, currently collect as much data as possible to develop products and services such as predictive seeding. To ensure the flow of data, they develop agricultural platforms that stream farm data directly from machines. They create incentives and lock-ins that make the full functionality of data-based decision support tools like Bayer's CFV contingent on farmers uploading their data, or that, like John Deere tractors, foreclose any other option. Monopolization, according to Birch and Muniesa (2020), is a key feature in the assetization process. Our analysis shows that in agriculture, just as in other spheres, platforms have become increasingly important to help establish relationships of monopoly, which partly explains why almost every agricultural firm offers its own platform. Furthermore, our analysis shows that key features of data assetization are present when agricultural data collection and valorization are made possible through policy and regulatory frameworks that protect trade secrets, allowing for obfuscatory practices and for companies to “protect” data as their own intellectual property (Birch, 2020). Companies are not required to disclose how they use the farm data they collect after it has been anonymized and aggregated. And license agreements are used as legal tools to obstruct access to data and, thereby, the ability of external observers (including big data studies researchers) to test and validate algorithms. Artificial intelligence and decision platforms are “pernicious black box[es]” (Pasquale, 2015) protected from critical oversight by intellectual property rights, copyright, and trade secrecy law—key elements of assetization dynamics (Birch, 2020: 4). Ultimately, it is impossible to test the validity of these platforms, the accuracy of AI-generated recommendations, or whether these recommendations work as claimed or are biased toward incommensurate corporate gain (i.e. toward recommending in-house products). Moreover, the publicly available license agreements, terms of use, and privacy statements we analyzed for this article were uniformly very difficult to understand. We believe that this obfuscation is a key element enabling assetization—complex language, inconsistent terminology, and references to different types of data in the legal statements likely make it difficult for farmers to comprehend the use of farm data.
We conclude that the three strategies of agricultural data assetization overlap and partly reinforce each other's effects. For instance, Bayer's “outcome-based” pricing program not only allows the firm to profit from adding such digital services to its portfolio but also helps to secure existing dependencies and relationships with farmers, whose data helps feed this new business model. Furthermore, we observed that often, the three forms of assetization are employed along with “traditional” forms of commodification. Bayer's predictive seed selection and placement, for example, is not just a new, assetization-based business model that sells predictions along with the company's seeds. It also secures Bayer's relationships with farmers and appears to help lock them into its product ecosystems, future data collection, and a state of dependency. Here, we see that the assetization of data complements rather than replaces commodity production and the commodification of tangible goods such as seeds, thus working to reinforce long-standing historical patterns of corporate control over farmers and, ultimately, the entire food system trajectories (Clapp 2021).
Data assetization in agriculture furthers both continuity—building on the existing dependencies and power asymmetries between agribusinesses and farmers—and change in food system power relations. Most important here are the numerous platforms, all with subscription models and monthly fees as an important part of their business strategy and key to corporate revenues (e.g. Farmers Edge Inc., 2022). Additionally, although the strategies we have identified relate to novel digital technologies in a variety of forms (apps, platforms, and hardware), they all work toward servicing a high degree of concentration in not only big agricultural data and centralized data infrastructure but also market power in the food system (Prause et al., 2021). Smaller start-ups being bought off by larger corporations are common in the ag-tech industry—a prime example is Monsanto's purchase of The Climate Corporation in 2014. The concentration of power related to these technologies is a long-standing and well-known pattern in agriculture. Compared to Big Tech firms like Amazon and Google, powerful agribusiness and ag-tech companies are uniquely positioned to benefit from data assetization. Agribusinesses have unique capabilities because for almost a hundred years, they have shaped the development of technology to control farmers and allow capital to infiltrate into farming—what political economists of agriculture call “appropriationism” (Goodman and Redclift, 1994; Stone, 2022b). These long-standing relations of dependence have also fostered the uptake of proprietary products that have further tied farmers to technologies and companies. In contrast to other fields, where data is transformed into value via assetization strategies such as personal data collection, assetization strategies in agriculture build on historical relationships and patterns of power that are by now incredibly solidified (Bronson and Sengers, 2022).
However, we also observe a fundamental transformation in agriculture, with new actors entering the field. For instance, Google's parent company, Alphabet Inc., has recently invested in Mineral, a project that, according to news reports, has already collected data on 10% of the world's farmland (Marston and Burwood-Taylor, 2023). Furthermore, agribusinesses are now partnering with classic tech companies to expand their digital capabilities and datasets and, thus, the reach of their assetization strategies. New partnerships between Big Tech players and agribusiness firms can be assumed to generate revenue from assets of different types and to be mutually beneficial, like the cooperation between Microsoft and Bayer (two firms that control large amounts of agricultural data) on Microsoft's Azure Data Manager (Marston and Burwood-Taylor, 2023). Thus, while assetization is a broader phenomenon described by data studies scholars, assetization in agriculture, given the sector's historical political economy, shows unique features when examined empirically. As such, this empirical examination of the assetization of farm data provides to big data studies broader insights on assetization processes, reinforcing the importance of looking at big data practices in ways that are attuned to potential specificity across forms of data and across sectors.
Conclusion
We have empirically studied and systematically described three strategies with which input supply corporations create value from agricultural data; all three fit within the larger pattern of big data's assetization. The strategies we have identified have socio-ecological implications; they affect social justice, food sovereignty, and sustainability, the latter of which does not always receive due attention in critical data studies (c.f. Gabrys, 2016; Goldstein and Nost, 2022). Our results indicate the reproduction of asymmetrical power relations in the agri-food system favoring corporations and the continuation of long-standing dynamics of inequalities. We can infer that the big data-based predictions agribusinesses sell to farmers are directed toward a productivist model of “surveillance agriculture” (Stone, 2022a) that reinforces existing patterns of unsustainable agro-industrial farming and renders other routes, such as agroecology, peasant farming, and organic farming less legitimate and possible.
Further work is needed, including research into effective regulation, which could ensure that the benefits of data assetization be equitably distributed between farmers, consumers, citizens, and those who control the data. Our analysis suggests that, even as farmers who adopt big data tools are likely to gain a competitive advantage over their neighbors via these tools’ precision analytics, corporate actors currently stand to benefit incommensurately. Currently, it is input supply companies who collect and control the bulk of farm data and who stand to gain from future data uses, not farmers, because the governance context surrounding agricultural data grants these companies control. Despite differences among international jurisdictions, in most countries where the firms analyzed in this article have their headquarters, we observe a lack of state-level regulation of farm data. In the absence of effective regulation that specifically considers the unique relations surrounding agricultural big data, service and technology providers are setting their own rules for data use via internal policies and enforcing them through license and user agreements.
Furthermore, antitrust laws in the United States and the European Union have generally failed to prevent increasing corporate concentration in agriculture (e.g. allowing Bayer to acquire Monsanto). Yet in Europe, leadership has been shown in the context of data governance, an example of which is the General Data Protection Regulation. Thus, policy leaders could use the novel technological domain of digital agriculture as a starting point for legal reform and a broader restriction of oligopoly in food and farming. What is needed is a reshaping of not just how corporations harvest value from data in digital agriculture, but how they operate in relation to food sovereignty in the food system more broadly.
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Bundesministerium für Bildung und Forschung (grant number FKZ 031B0750).
