1 “Never again should it be possible to say, ‘we didn’t know’. No one should be invisible. This is the world we want — a world that counts.” (Independent Expert Advisory Group, 2014: 3)
2 This call for omniscience summarises a 2014 report by the Independent Expert Advisory Group convened by the UN Secretary General. The report urged the Secretary General to use the visibility afforded by data to implement the goals of the upcoming 2030 Agenda for Sustainable Development. Adopted in 2015 to eradicate poverty, the 2030 Agenda enabled various governments, international organisations, IT companies and NGOs to set in motion a “data revolution” to leverage information and communication technologies. The current explosion in the volume of data and advances in the capacity to analyse and circulate it presage the realisation of an age-old technocratic dream of harmonisation by numbers (Supiot, 2015). Planning interventions for the “vulnerable” would at last be objective, efficient and cheap. But the data craze was not limited to the development field. The 2018 Global Compact for Safe, Orderly and Regular Migration called on states, international organisations and NGOs to “optimise” migration flows by measuring them, addressing so-called data gaps and leveraging new data sources (UN, 2018: 8-9). The academic literature approaches this international enthusiasm for data in contrasting ways. Some authors argue that data would ensure better responses, as they would have an “empirical base” informed and used by the widest possible range of social actors (Kraly and Hovy, 2020: 11), while others raise questions about what such data would make visible and the connections that would be made with the target populations (Gabay and Ilcan, 2017: 468).
3 This article aims to analyse this dual effect of the data revolution project through the practices of data production, analysis and circulation of the International Organisation for Migration (IOM). Since the 2000s, IOM has insisted on the use of quantitative and qualitative data in the name of the migration management paradigm. IOM works to establish a “regulated openness” of migration flows to ensure their orderliness and predictability (Ghosh, 2000: 221). It recommends the use of data not only to rationalise national migration policies, but also to respond to the so-called unsafe mobilities of populations from the Global South affected by natural disasters and human-made crises. Although IOM has been specialised in the international transfer of refugees and migrants since 1951, over the last twenty years it has extended its interventions to the humanitarian field (Bradley, 2020). The organisation’s aim was to “stabilise” precarious populations that were likely to move suddenly and chaotically (IOM, 2006: 50). IOM formalised its entry into the humanitarian field following the 2005 Pakistan earthquake. At that time, it acquired the status of co-leader of the “Camp Coordination and Camp Management” sector, together with the United Nations High Commissioner for Refugees, within the cluster approach coordinating international humanitarian action.  From this central position, IOM developed its activities aimed at movement management in “emergency and post-crisis situations” which became its largest budget item on a long-term basis. 
4 In order to “systematise” its response to the crises of the 2010s in sub-Saharan Africa, North Africa and the Middle East, the organisation devised its Migration Crisis Operational Framework in 2012. This reorganised IOM’s activities into fifteen sectors of assistance on health, housing, transport, livelihoods and environment of displaced populations (IOM, 2012: 3). However, the framework also aimed to ensure their protection and “psychosocial” well-being by monitoring their movements. To this end, IOM undertook to “[refine its] data management systems and technological tools for displacement tracking and mapping”. (IOM, 2012: 4). This combination of a concern for the preservation of people’s lives and the monitoring of their movements through the technocratic production of data is part of “humanitarian border management” (IOM, 2012: 6-7). In response to crises and the unstable movements arising from them, IOM claims to seek to identify displaced populations in order to “facilitate” their movements “by making appropriate referrals” (IOM, 2012: 8). These referrals involve the successive mobilisation and immobilisation of these populations prior to their arrival at temporary or permanent places of safety. They include their passage through camps, “sites of institutionalised mobility” (Salter, 2007: 51) (airports, ports, bus and train stations, etc.), “smart” borders, and occasional stops to provide care and check their vulnerabilities (IOM, 2012: 4-5). In this way, it should be possible to ensure that the chaotic movements of these populations are not prevented by the erection of hermetic national barriers in the name of state sovereignty  (Deleuze and Guattari, 1980: 480‑481). Borders should thereby become humanitarian zones of passage for populations to be identified, monitored and acted upon in order to release them into a transnational space where mobility practices are orderly and predictable (Walters, 2010: 139).
5 This was the justification for IOM gradually developing, generalising and refining the use of its “primary mechanism” (IOM, 2020: 12) for data collection: the Displacement Tracking Matrix. This technology systematically captures, processes and disseminates data on displaced populations during crises to monitor their movements and needs. It includes four “components” that can be deployed within a few days (IOM, 2017a): (1) tracking mobility of displaced persons within geographical areas; (2) monitoring flows of displaced persons at crossing points; (3) registration of displaced persons; (4) survey of displaced persons. IOM first used the Matrix in 2004 to monitor and assist internally displaced persons from the Iraq War. However, it was only from 2011 onwards that the organisation began to use it systematically for various natural and human-made disasters in the Global South, such as during the Libyan revolution in 2011. This was done with the help of the expertise of the “business intelligence” software developer SAS Institute from 2013 and the audit and consulting firm Deloitte in 2014. Following the so-called migration crisis on the Balkan route in 2015 and 2016, the Matrix was extended to monitor transnational migration flows (Münstermann and Van Der Vorst, 2017: 11). The scope of its monitoring was significantly enlarged: it encompassed seventy-one countries and 47.6 million people in 2018, compared to twenty-two countries and 14.4 million displaced persons in 2015 (IOM, 2019: 32). As the Matrix was no longer limited to internally displaced persons, it linked up with two other IOM technologies that targeted migrants: the Migration Information and Data Analysis System (MIDAS), which since 2009 has enabled several states in the Global South to monitor their borders through data; and the MigApp mobile application, which since 2017 has established a direct connection between IOM and so-called forced or irregular migrants.
6 I argue that these three digital technologies converge into a “surveillant assemblage” (Haggerty and Ericson, 2000: 608) that disciplines the uncertain movements of refugee, migrant and internally displaced populations in the Global South. An assemblage subsumes heterogeneous elements to make them interact in a functional whole (Haggerty and Ericson, 2000: 610; Patton, 1994: 158). Although it is composed of distinct technologies, the assemblage mobilises them to act on the populations it targets according to a biopolitical rationality and the technocratic production of digital data. Biopolitics works to “rationalise the problems posed to governmental practice by phenomena characteristic of a set of living beings forming a population” (Foucault, 2004a: 323). It identifies and monitors populations so that it can subject them to “comprehensive regulations” (Foucault, 1994: 179‑180). This is done through “security mechanisms” that focus on health, vaccination, education and the movement of populations to preserve their lives and prevent the spiralling of events in the crises that surround them (Foucault, 1997: 227 and 2004b: 22). These mechanisms contain populations in stabilised spaces of circulation where they indirectly adopt behaviours that reduce the randomness of flows of information, movement and disease that emanate from them (Foucault, 2004b: 49-50). However, by combining these mechanisms with the most recent technocratic practices that standardise, systematise and depersonalise the production of digital data, assemblage is not limited to populations in general. It establishes the socio-economic and biological profiles of the individuals that make up these populations in order to learn about them and monitor them more comprehensively. And it sets up digital connections that target them remotely.
7 Assemblage thus multiplies its points of contact with reality in order to harmonise the spaces of movement of populations as well as to precisely identify the unpredictable individuals among them who are likely to lose their lives. To this end, it delineates “cognitive spaces” (database, interactive map, analysis software, mobile application) and “physical spaces” (camps, sites of institutionalised mobility, points of entry and exit from territories) that capture and release displaced persons and their data in five stages of surveillance (Haggerty and Ericson, 2000: 608; Bogard, 2006: 100-101): (1) observation of displaced persons to assign them stable characteristics and tie them to monitored spaces; (2) standardisation of data to quantify the transnational population of displaced persons and isolate its regularities; (3) application of security mechanisms to camp populations; (4) disciplinary capture of individuals through biometrics; (5) empowering capture of individuals through survey and digital connection.
8 This article contributes to the debate on the data revolution project (Mayer-Schönberger and Cukier, 2013; Kitchin, 2014), by highlighting the way it combines biopolitics and technocratic practices to monitor precarious populations in the Global South. It begins by situating the data revolution within a process of technocratic digitalisation of responses targeting these populations. This process transforms biopolitical rationality to produce a more intense and capillary logic of surveillance. The paper then analyses the five stages of surveillance, drawing on the grey literature produced by IOM technologies: the Matrix methodological framework that “guides” its use (IOM, 2017a); data entry forms used by local IOM staff; Excel spreadsheets listing the data collected; and various fact sheets, brochures and promotional videos. The conclusion underlines the originality of the assemblage and the authority it bestows on IOM, despite uncertainty with regard to the utility of its data. Finally, it suggests that the humanitarian border management this system is said to embody is not intended to be applied to states in the Global North.
Data Revolution, Technocratic Practices and Biopolitical Rationality
9 Some authors argue that data plays an effective role in solving the “problems” faced by government practice (Verhulst et al., 2019: 1). However, others locate its role within a process of digitising responses targeting precarious populations in the Global South following a number of failures. Richmond and Tellidis point to the post-9/11 crisis of an “analogue” model of international relations that has been undermined by the acceleration of trade, mobility and technological progress. State sovereignty, geopolitical rationality, territorial citizenship and the multilateral bureaucratic system that characterise it, are now powerless to stabilise the international order that emerged from the twentieth century (Richmond and Tellidis, 2020: 936). Despite efforts aimed at so-called peacebuilding, reconstruction of failed states and sustainable development, the crises of the 2000s and 2010s (in Afghanistan, Iraq, Libya, Syria, Haiti, etc.) dragged on without any permanent resolution, while at the same time producing sudden and chaotic movements of populations. Analogue interventions are limited to managing millions of “undesirable” displaced people (Agier, 2008: 13) while ignoring local specificities, knowledge and needs. Their inadequacies also stem from the physical disconnection they involve between humanitarian actors and precarious populations (Broome and Seabrooke, 2012: 6; Duffield, 2010: 471).
10 These failures prompted attempts at reform (Richmond and Tellidis, 2020: 940). Duffield (2001: 310 and 2016: 148) highlights the gradual emergence of a new type of response that is less concerned with states in the Global South and more focused on influencing the behaviour of their populations. Thus, the data revolution project aims to “better” access these populations through the supposed efficiency of technocratic practices of digital data production. This project implies that interventions should no longer be concerned with imposing hierarchical control, respecting legal or geopolitical considerations, and following rigid bureaucratic procedures (Ginty, 2012: 289-290). Instead, they should define the “problems” that emerge from reality in technical terms. Then they are supposed to solve them by creating information systems that establish, in a depersonalised and distanced way, networked connections between humanitarian actors and precarious populations rendered visible (Mulder et al., 2016: 2-3; Jacobsen and Sandvik, 2018: 1509; Gabay and Ilcan, 2017: 477). However, the data revolution should not be reduced to its technocratic practices. Its concern for preserving the lives of populations affected by crises places it within a biopolitical rationality. Yet, by relying on the most recent advances in the systematised and cheap digital capture of reality, this rationality is transformed.
11 Indeed, biopolitics cannot be analysed independently of the information production practices by which it determines the characteristics of populations and the risks they run (Foucault, 1994: 178-179). Foucault noted that the biopolitics that emerged from the eighteenth century onwards was based on statistical estimates drawn from analogue data concerning the mortality of a population within a national territory to be developed (Foucault, 1997: 227). It thus treated this population as a macroscopic phenomenon to be inserted in its entirety into a series of mathematically probable events (Foucault, 2004b: 64). However, the technocratic practices of digital data production are now said to make it possible to find out about a transnational population that is not confined to a space delimited by national citizenship (Reubi, 2018: 89-90), as well as to accurately capture among this population individuals at risk of losing their lives. Biopolitics is therefore no longer reduced to the population. It comprehensively encompasses the individual who previously escaped security mechanisms (Macmillan, 2010: 49-50). Biopolitics is thus accompanied, particularly in more restricted spaces, by disciplinary forms of action on individuals (Barry, 2019: 369).
12 Hence, biopolitical rationality and technocratic practices of data production together produce a logic of more intense and capillary surveillance of the displaced populations of the Global South. IOM technologies illustrate this; they converge in an assemblage that delineates cognitive and physical spaces to capture and release populations and individuals in five stages of surveillance.
Observation of Displaced Persons to Assign them Stable Characteristics and Tie them to Monitored Spaces
13 In the first stage, the assemblage does not shape the displaced persons that pre-exist its surveillance (Haggerty and Ericson, 2000: 608). It merely captures the multiple flows that emanate from them in space and time by producing voluminous data that are supposedly “commensurate with the diversity of reality” (Rouvroy and Berns, 2013: 172). This sustained harvesting of data is presumed to remedy the shortcomings of analogue interventions that are disconnected from local populations and knowledge. However, surveillance does more than simply observe, describe and count the displaced in order to learn more about them and disaggregate them into different types of data (Haggerty and Ericson, 2000: 612). It introduces cognitive brakes on the uncertainty associated with them. It ties displaced persons to an a priori chaotic space of circulation, which it striates with bounded geographical areas where “in and out movement is routinely captured” (IOM, 2017a: 4). It assigns them a set of socio-demographic characteristics that are stable over time and reduce the heterogeneity of their experiences. This is the dual operation performed by the first two components of the Matrix: “mobility tracking” and “flow monitoring”.
14 Mobility tracking produces quantitative estimates of the presence of a group of displaced persons who are immobilised in a defined geographical area (village, building, neighbourhood, camp, national administrative area) (IOM, 2017a: 3). To this end, 4,000 local IOM officers, assisted by 200 technicians and a support team based in Geneva (IOM, 2018b), collect data through: direct or telephone interviews with key informants (government officials, leaders, humanitarian actors, hauliers, displaced individuals); focus group discussions; and direct observation in the field (IOM, 2017a: 3). These officers fill in digital forms using touch screen tablets that record a profile of the observed geographical area on a daily, weekly or (bi)monthly basis (see Figure 1). This profile assesses the number of displaced individuals and households living and circulating in the pre-defined, mapped and geo-referenced space (IOM, 2017a: 5). The estimate also identifies their ages and genders, as well as the most vulnerable among them (pregnant women, unaccompanied minors, disabled and elderly people) (IOM, 2015a: 2). It also specifies the origins of the displaced, the reasons for and dates of their movements, the location, type and accessibility of their shelters, their access to humanitarian assistance, and their needs in terms of drinking water, food, healthcare, education and security (IOM, 2017a: 4-5). When the displaced persons observed are instead in a state of sudden and uncertain movement, the Matrix flow monitoring is triggered to generate quantitative estimates at more frequent intervals (hourly or daily) (IOM, 2017a: 3). IOM does this by establishing “monitoring points” in the territory, where its local officers contact key informants or displaced persons, or conduct direct observation to establish: the type of flow and its direction; the number of individuals and households on the move; their nationalities, ages and genders; the type of and reasons for their mobilities; their destinations; their means of transport; and their “intentions” once they reach their destination (IOM, 2017a: 6-7).
Figure 1: Extract from the Matrix Data Collection Form
Figure 1: Extract from the Matrix Data Collection Form
15 Through its two component parts, the Matrix simultaneously and repeatedly covers thousands of geographical areas across several countries. For example, the quantification of IDPs and returnees in Libya in the period July-August 2018, required the collection of information in 759 territories to track the movements of an estimated 575,000 individuals (IOM, 2018a). Nevertheless, the raw data produced at this stage are not intended to guide any interventions on displaced populations. Stemming from heterogeneous local contexts and human observation, which is said to be imprecise when the territory monitored is large (IOM, 2017a: 4), these data require processing to make them intelligible and supposedly objective.
Standardisation of Data to Quantify the Transnational Displaced Population and Isolate its Regularities
16 After observing displaced persons, in a second stage surveillance standardises the capture of data which emanates from them and transforms such data into information through various centres of calculation (Haggerty and Ericson, 2000: 613). The displaced are then abstracted from their territorial locations so that their data can be released into digital spaces where standardisation can play its full role in making them homogeneous and comparable (Haggerty and Ericson, 2000: 608). This implies the systematic production of lists that compile, cross-reference and link the data produced in the previous stage. In particular, this standardisation should suggest that the singular, imperfect and heterogeneous observations of distinct and unpredictable individuals actually relate to a single social fact: the transnational population of displaced persons. Despite its daunting complexity, standardised surveillance is said to be capable of approximating its magnitude from a distanced, encompassing and impersonal perspective that sudden crises cannot disrupt (Desrosières, 1993: 92). The data from this supposedly homogeneous population can then be scrutinised to determine intervention strategies according to a “statistical determinism” that ignores the ambiguity of reality (Desrosières, 2014: 166; Haggerty and Ericson, 2000: 613). Thus, the standardisation of data confers on surveillance an “inescapable inertia” (Bowker and Star, 1999: 117) when confronted with the uncertainty of displaced persons’ movements.
17 First, standardisation applies to the forms used by IOM officers to populate the Matrix. They are produced automatically and uniformly thanks to “knowledge management” experts seconded by Deloitte on a pro bono basis in 2014  (IOM, 2013). The data collected then flow back from the various territories where the Matrix is deployed to a Central Data Warehouse (CDW). This database is the deterritorialised but structured place where the shapeless accumulation of data on flows emanating from heterogeneous populations is released to be harmonised and simplified. The experts at Deloitte designed the CDW dictionary to address any ambiguities arising from reality (Quesada, 2016). It is meant to align the understandings and practices of all those who populate and use the Matrix. It does this by succinctly defining the categories of the Matrix (IDP, returnee, migrant, refugee, gender, age, etc.), the ways in which they are measured and classified (Anderson, 2018; IOM, 2017a). The Matrix support team then carries out its “data processing and analysis activities”, cleaning and refining the data from the field through automatic and manual checks for discrepancies, redundancies or outliers using the dictionary (IOM, 2015c: 2 and 2018b: 2). This leads to the production of so-called raw datasets that, for each round of observation, convert the forms of local IOM officers in a given country into standardised Excel tables with a dictionary to guide their interpretation. The thousands of rows in these tables list the territories observed (each identified by a code, latitude and longitude), while the columns present the various characteristics of the individuals monitored using (non-)numerical values (see Figure 2).
Figure 2: List of IDPs Monitored by the Matrix in Iraq
Figure 2: List of IDPs Monitored by the Matrix in Iraq
18 The datasets are then transferred to the Matrix website homepage (IOM, 2017b). This includes an interactive map of the world, from the US “geographic information systems” publisher ESRI, on which the numbers, locations and causes of displacement movements are recorded. This map, which the user can enlarge or minimise, thus establishing a linear and univocal continuity between the precision of the particular and the generality of the whole (Latour, 2007: 261), is accompanied by an eight-digit decimal counter. It counts the entirety of what is presented as the transnational population of displaced persons that the Matrix is said to track, rather than simply estimate, during the current year. However, the datasets are not only supposed to represent this population. They are intended to allow for the definition of systematic strategies to act on this population. SAS Visual Analytics software processes the datasets to produce operational and strategic knowledge destined to guide IOM’s interventions. SAS Institute made its software available to IOM on a pro bono basis following Typhoon Haiyan in the Philippines in 2013.  It analyses the Matrix data using descriptive statistics methods and without prior assumptions to supposedly avoid any biases (Rouvroy and Berns, 2013: 170). The software then supplies the Matrix support team with various distributions, averages, variations and percentages displayed in tables, graphs and interactive maps (Business Wire, 2014). The team then cognitively, remotely and quickly identifies the areas that are most in need of clean water, food and healthcare, or that are overcrowded and composed of particularly fragile individuals (SAS Institute, 2015b: 45s). The software also scans datasets over several years and countries to isolate typical crisis correlations such as between irregular food distribution, tensions between displaced people and “host communities”, and cases of rape and attempted rape. This “historical” processing of the datasets is said to capture regularities in the population of displaced persons and to enable the Matrix to “quickly spot unusual trends and therefore predict issues before they emerge” (Hsieh, 2015).
Application of Security Mechanisms to Camp Populations
19 This predictive identification of irregularities is intended, among other things, to enable the Matrix to correct the “problematic” characteristics of “spontaneous” or “organised” gatherings of displaced persons as provided for in the Migration Crisis Operational Framework (IOM, 2012: 3-4). The challenge of the third stage of surveillance is to capture displaced persons who are present in these semi-enclosed locations to preserve their lives and address the causes of their disorderly movements. But at this stage, surveillance does not yet act directly on the displaced persons to discipline them and codify the behaviours prohibited in a camp that is fully enclosed and policed. Rather, it focuses on transforming the physical reality of the living and circulation space of these displaced populations through security mechanisms (Foucault, 2004b: 48-49). The camp must be equipped with the necessary resources to preserve them, and it must accommodate the fluid collection and circulation of data to anticipate their needs (Rouvroy and Berns, 2013: 172; Haggerty and Ericson, 2000: 613). Surveillance therefore establishes a digital network encompassing the camps so that remote but systematic action can trigger the adequate security mechanism to correct a given risk to the lives of the displaced.
20 For this reason, SAS Visual Analytics “findings” are systematically transferred to the tablets of experts in IOM’s assistance sectors and to other humanitarian actors in the camps (SAS Institute, 2015b: 1min30s). This downward and depersonalised transmission remotely guides the application of security mechanisms. For example, the presence of an unusual proportion of young children should lead to the distribution of appropriate food and hygiene products by the UNICEF-led “Nutrition” sector (Goodnight, 2015). Whereas the lack of sheet metal shelters as the rainy season approaches should prompt IOM to import the necessary building materials from the nearest international market (SAS Institute, 2015a: 2min17s). Finally, the prevalence of symptoms of fever, diarrhoea and skin diseases should be reported to the World Health Organisation’s “Health” sector in order for medication to be prescribed (Business Wire, 2014). The timely activation of these security mechanisms therefore depends on IOM’s ability to “flag issues to other agencies” (SAS Institute, 2015b: 1min6s). In order to maintain its position as co-leader of the Camp Coordination and Camp Management sector and to encompass the various actors in the camps, in 2014 IOM again called on the standardisation expertise of Deloitte (2013). It applied the business management method of “process mapping” which diagrammed the Matrix information system within the camps to identify overlaps and gaps in data exchange between humanitarian actors (Deloitte, 2015). The resulting “bird’s eye view” was intended to optimise the flow of data to capture the camp space more quickly, establish responsibilities, and avoid gaps in data availability (Pluto and Hirshorn, 2003: 1). For example, in response to sexual violence, Deloitte’s mapping highlighted the need for IOM to pass on data on the number of HIV post-exposure treatment kits available in the camps to NGOs assisting victims (Deloitte, 2015).
Disciplinary Capture of Individuals through Biometrics
21 However, surveillance is not limited to such remote and simplified actions on displaced populations. Its fourth stage seeks to directly capture individuals in order to discipline them and condition their sustainable release into a wider space of orderly circulation (Haggerty and Ericson, 2000: 613). If surveillance aims to make the macroscopic object of the displaced population intelligible and predictable, it also seeks to translate the microscopic level of the individual into “pure information” through biometrics (Haggerty and Ericson, 2000: 613). This measurement of individual life is necessary because the statistical work of quantification in surveillance faces a tension. The regularities of the population that it isolates subsume the unpredictability of the displaced individual without completely cancelling it out (Desrosières, 1993: 95). In order to monitor individuals from whom the unexpected may emerge and fix them to unique, unalterable and permanent physiological and behavioural traits (Ajana, 2013: 3), the multiplicity of displaced populations must be disaggregated into “data doubles” of real individuals (Haggerty and Ericson, 2000: 613). This allows surveillance to make and verify specific discriminations among displaced populations (Haggerty and Ericson, 2000: 614). The data doubles become passwords that regularly condition and control the practices of real individuals (Deleuze, 1990: 244). Surveillance then determines which individuals are eligible for the resources reserved for those in precarious situations in a camp. But above all, it controls the access of displaced persons to a transnational space of orderly circulation. Indeed, the surveillance of data doubles ultimately leads to the “direct physical relocation” (Haggerty and Ericson, 2000: 613) of their real counterparts.
22 It is through the third component of the Matrix, “registration of displaced persons”, that surveillance directly captures displaced persons present in the camps. IOM justifies the use of biometric registration by the need to combat double registration fraud (IOM, 2015b). Registration thus consists of an in-depth individual interview in which respondents are required to tell the truth about themselves in accordance with a stringent disciplinary procedure. They must give their name, age, gender, nationality, ethnic origin and religion, their point of departure, the date and reason for their movement, their education and employment, their identity documents and their telephone number. The registration then captures their fingerprints and a photograph of their face and stores them in a dedicated database. This registration covers hundreds of thousands of individuals worldwide, including in South Sudan where it extended to more than 700,000 people in 2018 (IOM, 2018d), and assigns them “registration cards” to benefit from security mechanisms. To further circumscribe them in the camps, “exercises” subsequently regularly check their identities and latest locations, whether they have their cards and whether they are still eligible for assistance (IOM, 2018d).
23 Nevertheless, the use of biometrics is not limited to the restricted space of the camp that IOM must eventually abolish. It extends to the so-called relocation and return operations of displaced persons envisaged by the Matrix (IOM, 2017a: 8). These operations draw on IOM’s expertise in “transport assistance” to release individuals captured in one crisis space into another more stable one. They evacuate, relocate or repatriate displaced persons, within a country or across borders, by air, sea and land to temporary or permanent places of safety (IOM, 2012: 4-5). These operations subject displaced persons to rites of passage that discipline their movements: presentation of a valid travel document, submission to biometric identity verification and collection of travel data. They thus move through the sites of institutionalised mobility where the surveillance of their data doubles conditions, records and filters their physical entries into an orderly transnational space of circulation. This biometric conditioning of individual movements does not take place directly through the Matrix. IOM has been selling its MIDAS to national border control institutions in the Global South since 2009 (approximately twenty in total in 2018 including in Mali, South Sudan and Somalia) (IOM, 2018c: 1). This technology monitors entry and exit points, in real time, on land and airport borders that are a priori porous and disrupted by crises. After examining travel documents under “white, ultraviolet and infrared light”, it captures individual biometric data and compares them with its national database and the Interpol alert list (IOM, 2018c: 2). It also tracks individuals once they cross an entry point. It distributes and records their residence status and stores their data for the purpose of identifying them during future movements (IOM, 2018c: 2).
Empowering Capture of Individuals through Survey and Digital Connection
24 However, surveillance is faced with the difficulty of directly capturing displaced persons (especially irregular migrants) who are not fixed in camps and whom smart borders cannot identify. These individuals move as if they are not part of the collective object that is the supposedly homogeneous population of displaced persons, and thus risk upsetting its predictability (Foucault, 2004b: 45). As individuals at risk, who are unidentified and inaccessible to security mechanisms, surveillance seeks to reach them and empower them in a fifth stage through the use of surveys and the remote establishment of digital connections.
25 Thus, the fourth component of the Matrix, “surveying displaced persons”, aims to find out about these individuals at the “micro-level” (Lanfranchi, 2018) by integrating their motivations and intentions into surveillance. In 2016, the Matrix support team designed the “Comprehensive Migration Flows Survey Model” to reach these “hidden” individuals. This is a questionnaire for irregular migrants that addresses eight themes to obtain in-depth knowledge about them: socio-demographic profile, migration routes, resourcing the journey, role of intermediaries, vulnerability factors, decision-making factors, role of the diaspora, perceptions towards Europe (Münstermann and Van Der Vorst, 2017: 12). They were broken down into forty-three questions that were asked in a survey in February 2017 to a snowball  sample of 7,248 Afghans and Pakistanis with diverse (im)mobilities: residents in the country of origin; potential migrants; migrants travelling through the Balkan route; migrants who have reached destination countries; families of migrants remaining in the country of origin; and returnees (Münstermann and Van Der Vorst, 2017: 12). The survey results showed that deprivation, violence by smugglers and traffickers, and deportations threatened the lives of irregular migrants. Given these dangers, “most” respondents would not repeat the migration experience and would not advise their friends and family to migrate (Münstermann and Van Der Vorst, 2017: 13). However, the survey ignored the structures that underlie the dangers faced by these precarious individuals in exceptional circumstances (Mbembe, 2006: 29). It assumed that migration is an individual decision and that it depends “on the individual’s resources, aspirations and capabilities” (Münstermann and Van Der Vorst, 2017: 13). Hence, many questions in the survey focused on the preconceptions and information held by respondents prior to their irregular migration and on whether these proved to be correct in retrospect (Van Der Vorst, 2017: 13):
“Did you make the decision to migrate yourself? (if not, [who] did); Did you discuss your potential migration with others? (if yes, with whom); Where, what, from whom [did you have] information about Europe?; Did friends and/or family provide information on Europe?; What type of information was provided by friends and/or family in Europe?; Do you know and can you explain what an asylum procedure is?; Where did you obtain information regarding asylum procedures?; Did you decide upon a destination before departure?; Expectations before arriving in final destination country?; Would you advise others to migrate? (why or why not); Would you migrate again?”
27 These questions suggested that only an informed individual choice would avoid dangerous movements. The Matrix took up the codes of medical surveillance surveys that seek to identify health risks for individuals who are alone responsible for their health (Reubi, 2018: 91). The survey thus constitutes a tool for cognitive capture and empowerment of a type of individual who is not accessible to surveillance. It encourages such people to join the transnational population of displaced persons, which is predictable and subject to life-saving interventions.
28 With the hope of integrating one million of these hidden displaced persons into empowering surveillance, in 2017 IOM acquired the MigApp mobile application from the IT company Kony (IOM, 2017d). This digital space establishes a direct digital connection that is intended to capture individuals in the midst of uncertain movement to put the choice and responsibility of making themselves visible and supported “at their finger tips” (IOM, 2017d: 1). To use the application, they have to provide genuine personal information (surname, first name, gender, age, nationality, current place of residence, phone number, email address, Skype or Facebook IDs) and their geolocations.  Users’ data doubles then access various “services” that supposedly capture real individuals and correct their mobilities: information on risks and conditions of legal movement; testimonies of former irregular migrants; consultations for medical examinations; anti-trafficking hotline; assistance in applying for visas and obtaining discounted air tickets; registration of travel documents; IOM’s so-called voluntary return assistance programmes. The application thus feeds the organisation with “anonymised registration data” which are analysed to identify patterns among these hidden individuals (IOM, 2017d: 1). And if they are in danger of death, they are transmitted to the authorities and NGOs to physically capture them. 
29 This article examined the assemblage through which IOM monitors the displaced populations of the Global South. Its originality does not lie in the systems (IT, logistics, health), practices (humanitarian, business management, police border control, sociological survey), technologies (database, biometric scanners, mobile application) and actors (non-state and state) that comprise it. Rather, it lies in the convergence it seeks to establish between these distinct and heterogeneous elements to make them operate as a whole made functional by biopolitical rationality and technocratic production of digital data. However, the complexity of the assemblage leads to a production of data that seems to overwhelm IOM and which is of uncertain utility. This was suggested by the IOM Coordinator of Camp Coordination and Camp Management and future Coordinator of the Matrix, Nuno Nunes, when he presented the Matrix to a UN audience in 2016: “We have lots of data in most circumstances and we need to figure out what to do out of that data” (UN Web TV, 2016: 1h59min57s). So why is IOM still investing in this assemblage?
30 First, it allows the organisation, which some describe as opportunistic and entrepreneurial (Bradley, 2020: 8; Pécoud, 2020: 11), to access external technical resources such as those of SAS Institute and Deloitte at no cost. Yet, IOM is first and foremost a bureaucracy whose authority depends on its ability to perform tasks expertly (Barnett and Finnemore, 2004: 24). The organisation can rely on the a priori technical, neutral and effective assemblage to legitimise its interventions in a humanitarian field that does not historically fall within its remit. The assemblage not only strengthens IOM’s position as one of the main producers of data on displaced populations,  it enables IOM to assume the central role of systematically determining where, when and what actions other humanitarian actors (particularly those in the UN clusters) should implement. IOM thus avoids ambiguous negotiations, which would limit its authority, with the doubts, experiences, interpretations and diverging objectives of these actors.
31 The assemblage also confers moral authority on IOM. In situations where the movement of displaced populations is a matter of life and death, the organisation can present its interventions as acts of protective charity (Walters, 2010: 145). By appropriating humanitarian practices, IOM can be said to depoliticise them and limit them to certain peripheral and crisis territories of globalisation in order to contain the population movements they produce (Heller and Pécoud, 2017: 81; Walters, 2010: 146). Thus, the assemblage focuses only on the populations and states of the Global South and assigns them a subordinate position as targets of surveillance or as receptacles of turnkey technologies. The humanitarian management of borders that the assemblage supposedly brings about in order to facilitate the movement of displaced persons is not intended to apply to the states of the Global North. It does not call into question their border security policies, in which IOM participates (for example, by returning migrants they deem undesirable) and which may lead to displaced persons risking their lives (Heller and Pécoud, 2017: 64).
This UN approach consists of eleven clusters (health, food security, nutrition, education, etc.) each (co-)led by an international organisation.
These activities accounted for USD 473 million of a total of USD 956 million in 2018, the majority of which was funded by states in the Global North (IOM, 2017c: 24).
For example, IOM was proud to have “helped” the Tunisian authorities to maintain their borders open in 2012 with civil war-torn Libya, by aiding them “in the identification of those fleeing the crisis” (IOM, 2012: 8).
The technical resources provided by Deloitte to IOM are part of its Humanitarian Innovation Program, which has been supporting humanitarian actors in protecting crisis-affected populations since 2012 (Deloitte, 2021).
SAS Institute’s technical assistance to “modernise” IOM was part of its philanthropic activities in which it uses its expertise in data analysis to address humanitarian issues (SAS Institute, 2018).
Under this method, respondents direct the interviewer to other potential participants in the survey. This makes it possible to target a hard-to-reach population, but without constituting a representative sample.
Use of the application by the author, November 2018.
This potential transmission of data to authorities and NGOs is a condition that is communicated to the user on first access to the application in a document entitled “Terms and Conditions”.
For example, IOM used its self-proclaimed status as the main “provider” and “holder” of data on IDPs to influence a 2021 UN report on this population (IOM, 2021). The report called for action on IDPs through data, citing the Matrix as an example. IOM considered that the report reinforced its authority.