The Human Side of Programmatic Advertising

(1)

A practitioner ethnographic approach to discover the timespace of humans in programmatic advertising at Dentsu Aegis Network Netherlands.

By Daniël Landman

Supervisor: Dr. A. Gekker

Second reader: Dr. E Weltevreden

New Media and Digital Culture University of Amsterdam

(2)

(3)

1. INTRODUCTION

Every aspect of daily life seems mediated through the internet. We navigate our world – whether physically, via a maps app on our phone, or emotionally via a social channel – through the world wide web. These services seem free for users. However, the cost of this is paid for by someone else. Adhering to the maxim ‘If you are not paying it, you’re not the customer, you are the product being sold’. – Andrew Lewis (Quote Investigator). That being the case makes it worth knowing who is paying for these free services, what their motivations are, and how they are paying for it.

The business model of the internet heavily relies on advertising real estate, with the digital ad spend prospected to grow by 12% in 2018 to 254 billion US $ (Cooper). Users are paying by giving their attention, sharing their data and are segmented into groups, or target

audiences based on their behaviour. A monetary exchange still exists, only it is between an

advertiser and the publisher or app owner.

To raise the chance of getting a return on their investment, advertisers rely on

publishers and app owners to share data on their visitors in order to make informed decisions. This data lets them know if they want to be the one to pay for that particular user's attention and how much they are willing to pay for it. Besides user-data, publishers monetise their content through a tier-system based on the quality of the advertising real estate – or their

placements. (Münstermann & Würtenberger).

Advertisers are paying for users in the hope of getting to get a return on their investment by guiding them through the consumer funnel of Attention, Awareness, Consideration, Action, and Conversion: Vying for the attention of users so that they are aware of their brand, that they might consider their products and services, leading them to initiate action and ultimately covert by buying their products or service giving them a return on their investment. The price an advertiser pays is dependent on a lot of factors, ‘One hundred thousand people going to a site is worth something, but a site that only five people visit can be worth more if they are the right people.’ (Stephen Klein of I/Pro in Murphy 1996, in Hoffman and Novak 2000). Each step along the way has different best practices, goals and tactics.

Now think about how long it takes for a web page to load and an ad there to appear. That is not much time for contemplation, is it? However, in that time an auction has started

(7)

and ended, ads were served and in some cases with additional third-party measurement systems. Naturally, much of this process is not done by human hand but by programs and algorithms, leading to the term ‘programmatic advertising’, which will be a focal point for this thesis.

‘Programmatic advertising describes the automated serving of digital ads in real time based on individual ad impressions opportunities’ (Busch, 8). The exact workings of programmatic advertising systems are blackboxed, meaning that their inner workings are hidden to everyone, except to those who work in it. Research into it is limited to looking at its input and output (Winner). The most defining characteristics and the most significant selling points of programmatic advertising are granularity and automation. The granularity of data provides advertisers with highly detailed feedback to optimise their budget and automation offers the opportunity to extend decision-making to a minute level (Busch). Eric Bader, the Chief Marketing Officer of RadiumOne, an advertising tech company specialised in media buying automation, was quoted in a 2013 Adweek article saying “I know there are references to platforms and technology out there. But programmatic is about software that is productive in a way that humans can’t be” (Adweek,n.pag..). This selling point is repeated to the point where programmatic is conflated with automatic, meaning working self-sufficiently without human interaction.

Programmatic Advertising Landscape

The first ever display banner ad was one placed in 1994 on the HotWired website, currently named WIRED.com. The banner which is often referred to as the first ever, seen in figure 1, was part of a joint experiment for fourteen brands. The AT&T banner provoked users to click on the ad without even mentioning a brand or a product (Singel et al.).

Figure 1: First Display Banner (Singel et al, 2010)

At that time, advertisers would have to make direct deals with websites owners similarly to print advertisements. Buying ad space in the print manner is a labour-intensive

(8)

practice. The publisher would sell ad space or ad slots on the webpage called ‘inventory’ in the form of an impression. A defined as the moment where the ad had loaded on the site and at minimum, begin to render to be acknowledged as a valid ad impression (IAB, MRC Viewable Ad Impression Measurement Guidelines). In less technical terms, it is commonly understood as the moment when a user sees the advertisement. Later, companies called ad networks would come into the markets. These companies would group websites with similar demographics or content together, thereby simplifying the media buying process. However, this was also still in line with how print advertisements were sold. The true nature of

programmatic advertising came into being thanks to the continual innovation in computers and the internet. For programmatic advertising to come to fruition, five developments were needed:

1. the ability to store large amounts of data,

2. the ability to process that information as soon as it comes in,

3. the existence of a standardised structure for systems to communicate with one another,

4. fast data connections between systems,

5. a way to scrape the data generated by users (Busch, 4).

This evolution also introduced a legion of specialised tech companies to become involved leading to an increasingly fragmented and complex landscape (Brinker).

Three Stages of Programmatic Advertising

What separates programmatic advertising are five characteristics, which are built upon the five developments described above. The five characteristics are granularity, real-time trading, real-time information, real-time creation, and automation (Busch, 8).

The capability to scrape data generated by users combined with the ability to store large amounts of data gave the opportunity to higher degrees of granularity, which are used to create a higher level of customisation and targeting. The ability to process information as soon as it comes in with fast connection between the systems in a standardised way enables real-time functionalities such as real-time trading, real-time information, real-time creation. When these are combined, it allows for a level of automation which like Eric Bader from RadiumOne, said ‘is productive in a way that humans can’t be’.

(9)

Gertz and McGlashan differentiate between three stages of programmatic advertising: 1.0, 2.0 and 3.0. The first stage, Programmatic 1.0, was focused on reach. There were few options in terms of target audience segmentation, and these options were based on first-party data. First party data was based on the advertisers own website traffic.

In the next stage, Programmatic 2.0, second party data and third-party data also became available. Second-party data is website traffic from another advertiser such as a car rental agency buying the audience from an airline company, so they know, this audience of users will be in Athens in the foreseeable future, probably without a car. This second party data set lets car rental agency target that airline audience with ads offering their services in Athens. Third-party data, on the other hand, is from data owners such as a data company, a DSP, an SSP or a publisher. Programmatic 2.0 is also a stage where not only reach was important but also attention. How much of the ad did the target audience member see and did that person interact with it? This combination of reach and attention referred to as awareness inside the industry and with it comes a focus on quality. Quality is defined by the webpage it is located on and where on that page is the ad shown. Is it shown on a reputable news site or an illegal download site? Due to the focus on quality inventory premium publishers have adopted a three-tier pricing segment system. Where the top tier is represented by the homepage of the publishers, the mid-tier by the second best placement but where delivery and quality metrics such as viewability can be guaranteed, and third bottom tier is the excess amount of impression which does not fit in the first or second tier. Generally, users will not scroll down to the bottom of the website, which is why an ad slot on top of the page is more valuable than one on the bottom. This focus on quality led to the reintroduction of private deals.

(10)

Figure 2: Visualisation of the programmatic ecosystem (IAB Spain, 2017).

In the third and current stage of the ecosystem, Programmatic 3.0, the focus is on the consumer. The buying side of the programmatic, seen on the left side of figure 2, is focused on reaching the ideal target audience as effectively as possible. Therefore, new technological features are more important on the buying side of the programmatic ecosystem (Schäfer & Weiss, 75). In this stage. “The Advertising organisation of the future is built on a unified view of the consumer, bringing data from all touch points together and building consumer segments based on their needs and values” (Gertz and McGlashan, 68). “They [Advertisers] combine as much data as possible into a holistic user profile”(Gertz and McGlashan, 57). In contrast to the dividual approach of segments, based on common demographics or interests. The focus in programmatic 3.0 is the individual, a segment of one.

This focus shift from quality to a segment of one is problematic for publishers since it makes their three-tier system mostly obsolete (Münstermann & Würtenberger, 32). An advertiser trying to retarget users in its most valuable audience pool will be less concerned as to where the next touch point might be. The advertiser has this user’s attention, they are aware of the advertisers brand, have considered the product range and made the first step towards a conversion leading to a return on investment. Looking back at the earlier quote from Stephen Klein, they are simply the right people. However, whether or not right users visit the publisher's site is something the publisher has no control over.

(11)

Münstermann and Würtenberger suggest that the way to combat this is to generate more first-party data that can be used to fuel granularity. This can be done by doing at least one of the following: profiling users, generating registration data, buying external data, buying external inventory or charge more for inventory with a high brand safety score (33-36). This means creating their own holistic user profiles and flagging their own content for

controversial topics, essentially creating more tier 2 inventory from tier 3 inventory. In practice, however, it is hard to match a user back unless you have that registration data and flagging their own content would be very labour intensive, which seems to go against the narrative of programmatic advertising, in which such processes should be easy because of the standardised framework and real-time functionalities.

Consolidation Into Walled Gardens

Another development in the programmatic landscape is the consolidation of brands and products into clearly defined software stacks. These proprietary software layers work well together with easy connections, which can enable new functions leading to a better product as a whole. However, they also do not work well with software layers of competitors. Leading to these ad tech stacks being called walled gardens. Which refers to a business practice where products are not compatible with competitors’ products (Rogers). One of the most notorious practitioners of this business practice is Apple. It has created its own ecosystem of products with proprietary chargers like the lightning cable for the iPhone, the MagSafe Power Adapter for its laptop range and the software function Airdrop to send files between these products. All these products work best together with other products from the same brand.

In programmatic advertising, Facebook and Google were long considered a

hegemonic duopoly of walled gardens, however Amazon, Verizon and AT&T are trying to establish more seats at the table by building and acquiring ad tech companies through

corporate takeovers in order to establish their own walled gardens as seen in figure 3. Making sure they have exclusive inventory, their own dedicated DSP, proprietary data and persistent, deterministic user identity.

(12)

Figure 3: Walled Gardens (Vidakovic, 2018)

This consolidation on ad tech is not just seen on a holding company level but also within the advertising technology stack of Google. In June of 2018, Google announced that it was going to rebrand and consolidate their ad stack by merging DoubleClick Digital

Marketing and the Google Analytics 360 Suite. The new stack is called Google Marketing Platform, and the ad tech products were renamed to remove the DoubleClick heritage brand.

This name change was not a simple rebranding. DoubleClick Bid Manager renamed to Display & Video 360, absorbed functionalities from Campaign Manager, Studio, and Audience Center; consolidating all their buying capabilities into a single product. The timeframe of when certain products will be consolidated into DV360 is still unknown. However, in the first trainingfrom Google at Dentsu Aegis Network following the announcement, the media agency was advised to stop selling DoubleClick Studio and

Audience Center to our clients as they would be integrated into Display and Video 360 in the coming months. This sort of announcement with multiple unknowns is endemic in the

programmatic advertising landscape, a reality which was mentioned throughout the interviews held in the research period. What happens when three specialists now have to work together in the same platform? How is their access defined and managed? Do all three specialisms still exist in the following year? Do media agencies consolidate their team to

(13)

mirror the consolidation of the platform? How does this consolidation effect the human roles within programmatic advertising?

Following Chapters

Ideally, a researcher would study at Google or Facebook as they are the makers of the most dominate programmatic advertising platforms. This thesis offers insight into the second-best sample, the power users. This thesis will provide insight into the human's role in

programmatic advertising by answering the main question: What is the role of human actors

within the programmatic advertising network surrounding an international client at DAN Netherlands?

This thesis studies the human role at the Amsterdam office of Dentsu Aegis Network, (DAN). DAN is the local office of the Japanese media holding company Dentsu Inc., which in 2016 was the fourth largest media holding company in the world in terms of revenue (Tan, n.pag.). Dentsu Inc is a conglomerate consisting out of more than 40.000 employees in 145 countries. These people are the ones advertisers employ to man the buttons of the duopolies’ products. They are the power users, the beta-testers and help shape the product by providing feedback. Dentsu Inc, the combination of the Dentsu Network in Japan and DAN outside of Japan self-reportedly was responsible for 38,5 % of global digital ad spend (Cooper, 6).

Chapter two provides a theoretical lens through which the discussion and ultimate conclusion are framed. It described how the office environment of programmatic advertising practitioner could be seen as a world in their own right through the work of Schatzki,

Takhteyev and Latour. Like a soldier transformed by the addition of their gear and their placement within the army. The DAN office is a transformative timespace with its own doings and sayings (Latour, Schatzki). Theodore R. Schatzki, Professor of Geography and Philosophy with a degree in Mathematics, calls these environments timespace since they are simultaneously bound by time and space. Within those timespaces, there lie common norms those norms are not set by the relationships between people within that collective common milieu but by the milieu itself. These timespaces are defined by four phenomena: action understandings, rules, teleological affective structure and general understandings (Schatzki). Through describing these four phenomena, we can begin to create an understanding of the human's role within this timespace. Latour defines this timespace as a network. Unlike

(14)

Schatzki he includes non-human entities as actors with their own agency. This agency is defined by the ability to influence other actors within the network. This agency is evident in frequent product updates and their changes to functionality. This structuring and

re-structuring of possible actions is called governance by Bruno Latour. Gilles Deleuze described a societal shift from Foucault’s societies of discipline to societies of discipline. Deleuze emphasised that, in societies of discipline, it is not the barriers that are important but the tracking of the person itself and how this affects the user. Within the programmatic advertising platforms change logs are kept of any change. How does it affect the human actors if all their mistakes are easily traced back to their user account? Do they hesitate to act? or could it be the opposite that they take pride in the work attributed to them, and what happens when a mistake is found?

David Lyon, Zygmunt Bauman and Rob Kitchin discuss that there is a participatory aspect to surveillance. To the point where Kitchin even argues that the master metaphor of surveillance is no longer Foucault's panopticon but the confession. Where people give their data willingly and await judgement. This judgment does not have to be to an external power; it can also be to oneself. The Quantified Self movement sees self-surveillance as a tool for effective self-improvement. At DAN the human actors do this by logging their hours. They confess how they spend their working hours to a system. They detail on which clients they spend their time and in some cases writing down exactly what they did where. But how detailed can they be? And can it ever fully describe what was done and why?

This is a critique Ulises Mejias provides on what he calls the oppressive nature of networks since they can never completely represent the work and that these digital networks can never hold their shape long enough to solidify frames of references. For a digital network to work, data must be defined and structured in order to be easily processed. The superposing of technological models and episteme onto social structure while leaving everything out which is not easily processed is what Mejias called nodocentrism and what falls outside the network Mejias refers to as paranodal.

By looking at these theoretical frames, the following questions arise: • What happens when you enter the fold?

• How do these phenomena present themselves?

(15)

• What type of surveillance and governance is present in the research subjects

timespace?

• What types of data are ‘confessed’ in the research subjects timespace? • What forms of nodocentrism can be seen in the research subject and what

paranodal data is missed?

Chapter three describes the methods which lead to the findings in chapter three and four. The first method used was a mapping based on the Actor-Network Theory of Bruno Latour, characterised by setting human and non-human parts of the network on equal footing. The mapping details not only the who and what within the network but also the parts which interact with each other. The second method used was a workplace practitioner ethnography. Through thick description, an account is made of the doings and sayings within the network by a researcher who was already a part of the culture. In order to reduce the number of nodes within the network as well as trust and an open line of communication, metonymic names were used instead of actual names. Metonymy works by referring to a salient characteristic of the domain to represent the entire domain. (Forceville). The use of metonymy as a

pseudonym for the actor provides a highlighting and evaluative aspect. In-depth interview were used to add depth and verification to the ethnography account. These interviews also allowed the insight into the role of the actors within the network, that was inaccessible for the ethnographical account of one researcher.

Chapter four provides a knowledgeable foundation of the network; the human and non-human actors in the network and how they work together to create a programmatic advertising campaign. After this knowledgeable foundation has been laid out, the experiences and observations from within the network are laid out. Thick description was made from the first day to the not long after the first conflict, to recognition within the timespace and failure of action understandings. The ethnographic account also describes situations where rules are broken, and action understandings have not yet been materialised, which highlight the need for them in the first place. Through the case study this goes into further detail, describing how a programmatic campaign is set up, human role to human role, how these human actors interact with non-human actors and how that relationship between human and non-human shapes the action understandings and rules between human roles within the network. Through

(16)

metonyms, their role is further highlighted within the network. In so doing, an understanding of the doings and sayings of humans within the timespace is created.

Chapter five brings the findings and theory together in order to provide answers to the research sub-questions stated on the previous page. It reiterates the different roles within the media agency in regards to a programmatic advertising campaign. In it, the connection is made between the liquid modality of liquid surveillance (Lyon & Bauman) and the liquid modality of working with every changing and unforgiving advertising technology platforms. After which, the practice of logging of hours is looked at through the frame of

self-surveillance and Kitchin’s metaphor of the confessional. Human actors log their hours and can these hours be lumped into two categories billables, meaning a client will pay for them and non-billable meaning the media agency will get no direct compensation for that time. Lastly, the cost of agency within the network is discussed with a focus on translation (Latour). What contributes to the cost of translation, possible ways to lower the cost and potential hurtle that stands in the way of those possible solutions.

(17)

2. THEORY

To understand the human role in programmatic advertising at DAN Netherlands, one must understand their environment: the network of actors, their agency through affordances and translation, the dynamics of power. To Bruno Latour, human and non-humans roles are equal parts within a whole. Latour proposed a thought experiment to address and subsequently sidestep the culture- or technology determinism debate. Imagine putting the people and the equipment on either side of a battlefield. Could you point to the naked men on one side, or could you point to the guns, tanks and uniforms on the other side and say, there stands the army (Latour 77 in Verbeek, 157). You would not be able to. Likewise, advertising one should not be separate the equipment of advertising without looking at the men and women behind it. Their capabilities drive their decision making, and their wants and needs drive innovations as well as the capabilities of the competitor.

A Transformative Enclosure

Latour’s thought experiment highlights the transformative nature of the whole and the performative action needed to exist inside that environment. Inside its network exists a collection of shared knowledge about how things work, a collection way of doing and an encompassing system of doings and sayings. Which Takhteyev calls worlds of practice in his ethnographic account (21).

It is not overly zealous to name these environments worlds in their own right. In fact, it is quite common. Just think of the colloquial phrase ‘the business world’ or ‘the academic world’ (Schatzki). Theodore R. Schatzki, professor of Geography and Philosophy with a degree in Mathematics, calls these worlds timespace since they are simultaneously bound by time and space. Within those timespaces, there lie common norms and to him, those norms are not set by the relationships between people within that collective common milieu but by the milieu itself. The doings and saying that compose a social practice are organised by phenomena of four types, best explained by Schatzki himself.

(1) action understandings, which combine knowing how to perform an action that helps compose the practice, knowing how to recognise this action, and knowing how to respond to it; (2) rules, by which I mean formulated directives, admonishments, orders, and instructions to perform or leave off

(18)

certain actions; (3) a teleological-affective structure, which embraces a range of ends, projects, actions, combinations thereof, and emotions that

participants should or acceptably pursue or exhibit, and (4) general

understandings of matters germane to the practice involved. (Schatzki, The

Timespace of Human Activity, 140).

Let us apply the four types of phenomena within the timespace of the subject, DAN Netherlands’ office and the human-centred business practice of programmatic advertising.

Action understandings consist of knowing how to perform the tasks inherent to your assigned role within the company. Often a briefing comes in the form of an email with an excel attachment detailing what is required in an as standardized form as possible. In order to remain in that timespace, you need to know how to read that briefing, know what to do with it, know when it is incomplete or unclear, and know how to turn to in order to ask for help or clarification. In order to read a briefing or engage in a meaningful way during a meeting one must know and be able to differentiate various industry-specific terms, like programmatic guaranteed, and the numerous acronyms such as CTR, VTR, CPV, CPL, CPA, CPO, OMP, PMP. There are so many that in the beginning, it can appear as a separate language.

Rules, as it related to the practices, are exemplified by the Service Level Agreement, SLA for short. It cements what can be expected. When is a briefing complete, how much time does the actor have to complete the task, when does the timer start, how picks it up when the actor is not there, whom to escalate towards when these agreements are not met. Besides this, there are also rules of the larger industry, like knowing how to work within the boundaries of the new General Data Protection Regulation, known as GDPR.

The teleological-affective structure is the most elusive to pin down as it consists of combining the other three to reach the accepted or prescribed goals and even performing the proper emotions and moods. It relates to how one can and should navigate the performance within the workplace. Practices such as pitching for new clients, learning more about one's discipline or the disciplines adjacent to one's own in order to communicate and think more effectively within the world of practice.

Lastly, general understandings transcend the particular world of practice.

Mannerisms regarded as professional and polite, ones such as holding doors open for people, not bad-mouthing people and being able to write a formal email.

(19)

Through advancements of technology, the milieu of time and space has become increasingly distant, moving beyond Schatzki’s original definition. The introduction of information and communication technology, overwise specified as non-human actors. Within the global network of DAN, users contribute to the network across vast distances and asynchronously. What has not changed is the transformative character of entering these the world of practices and the performative actions needed to exist within them.

Agency within a Society of Control

Power is predominately discussed concerning bodily harm and corrupt humans whose power seems absolute, but it is hard to critique power when mechanisms and politics are more subtle and increasingly convoluted (Foucault). Foucault equates power to an ensemble of actions. It does not consist or exist in some concentrated or diffused form but fundamentally is a

relationship where one's actions influence the actions of another and which are recognised by the other. Ergo in Foucauldian terms, power needs freedom. It needs a choice. (Foucault). This influence connects to what Latour characterises as agency within the network. The ability to act with and make other actors act within the network. By structuring possible actions, an actor is able to have a governing role within the network. By standardising possible actions, the cost of agency is lowered within the network (Latour), not just for the individual actor but for the network as a whole too.

Foucault described the transition of societies evolving from societies of sovereignty to

societies of discipline. Where people move from enclosure to enclosure and are subject to the conditions of whichever enclosure they are present. Deleuze described a second evolution to societies of control. In a control society, individuals have evolved from a static being to a more fluid form, one he named dividuals. In the former people are both individuals as a numeration within a mass; in the latter fluid form, they are both a dividual and a passcode within ‘banks’. In societies of control capitalism forms, the power structure of social control and marketing is its instrument. Not to enclose people but to keep people in debt (Deleuze).

The conceptual manifestation of societies of control is Felix Guattari’s thought experiment of a city where a person would be given access or restricted throughout the city

(20)

based on his or her key card. The city landscape continuously shapes and shifts based on the data on the citizen’s key card. Deleuze emphasised that it is not the barriers that are important but the tracking of the person and how it affects the user (Guattari qtd in Deleuze, 7). This surveillance is built into the ecosystem. Guattari’s experiment would not work without constant validation. The web is in effect this city. A video may or may not be available to a user due to distribution rights, or a site might be blocked by the government that regulates the physical space in which the user resides, or one ad may be shown instead of another based on a users navigation history. This society of control is fluid and contextual. Since most

programmatic advertising platforms such as Display and Video 360 are web-based and change logs are kept of every change made by any user; it shares more than a few similarities with Guattari’s city. Though the access modifications and surveillance are not as noticeable as in the concept city, activity is tracked in change logs, access levels can change at any time, and there are tools and access levels only privy to the employees of Google, but the

experience of the landscape is relatively static. The experience is static since there is no active feedback loop based on the tracking, meaning your access level will not change based on a coded trigger. It will be done by a different user, not the system itself.

Participatory Surveillance

Within the discussions around the ethics of surveillance, there seems to be a tendency towards what sociologist and philosopher of modernity Zygmunt Bauman calls

adiaphorization “in which systems and processes become split off from any consideration of

morality” (Bauman, 13). Adiaphorization comes in two different types.

The first type consist of people distancing themselves from the issue. They say ‘that is not my department’ or pushing morality to a different department saying ‘I just collect the data’ or ‘I am not the one collecting this data, I am just using it’ or ‘I am just doing what I am told to do’.

The second type consists of distancing the people from the data, which represents them. By converting a person into an anonymised ID with all their data attached, this ‘data double’ will serve as their proxy without the ethical implications. An example of this in programmatic advertising is the infamous cookie-IDs saved by a user’s web browser.

(21)

This notion of fluidity in surveillance is also taken up by David Lyon and Zygmunt Bauman in Liquid Surveillance. They discuss the actuality and influence of surveillance; also stressing that there is a participatory aspect to surveillance in liquid modernity. Liquid surveillance has two characteristics; the first is that “all social forms melt faster than new ones can be cast.“ (9); the second is “power and politics are splitting apart.” (9). Reasoning that currently power manifests itself on a global level and politics is still locally based.

The participatory aspect of liquid surveillance is so prevalent in modern society that Rob Kitchin argues for a new master metaphor. One to replace the dominant master metaphor of Foucault's panopticon. The prison concept where the guard might not be watching but where there is the constant possibility is they are. No, to Kitchin the new master metaphor is the confession, where people willingly put forth their data and await judgement. Within the network, this done by logging hours, with a clear distinction between billable hours and non-billable hours. Billable hours are hours where clients will pay for that hour. Another form of confession is were the trader role hands a report of the campaign result to the client. It is a confession of performance, where a trader can be prideful of ashamed based on its contents.

This confessional does not only need to be a confession to an authority or higher power. It can also be to oneself. If governance is the structuring of possible behaviour, self-monitoring allows users the self-governance to make informed decisions on easily obtained data points. Take the previous two examples of logging hours and reporting campaign results. If an actor gets proficient at a task or finds another way to doing it, they can look at previous confessions and assess the worth of that now honed skill or of that procedural change. The concept of self-surveillance is not exclusive to the business world. The Quantified Self movement, started by Gary Wolf and Kevin Kelly sees self-surveillance as a tool for structured self-observations. Through these observations, one can come to more effective self-improvement (Whitson).

Limitations of the Network

In the same way that it is hard to critique power when mechanisms and politics are subtle and convoluted (Foucault). It is hard to critique the oppression of networks when people join it willingly, bartering privacy for convenience and social participation. The networks default mode is inclusion and participation in it is rewarding in its own way, according to Associate

(22)

Professor Communication Studies and internet critic Ulises Ali Mejias. His critique of digital networks is not in that inclusion and participation in networks is without merit or to the determent on its inhabitants. The base of his critique is that networks will never be able to capture the world to its full extent and the misconception that it can is what is problematic.

The reason why networks cannot and will never fully capture the worlds is because these digital networks cannot hold their shape long enough to solidify frames of reference, mirroring the first characteristic of liquid surveillance. Furthermore, activities must be defined and structured in a way that is easily processed in order to work. If these activities and nodes are not easily processed, they will fall outside the network and into what Mejias calls the paranodal.

This superposing of the technological models and episteme onto social structures while neglecting everything that is not indexable is what Mejias calls nodocentrism. Nothing outside the network matters as these paranodal activities or influences are impossible to index. Nodocentrism provides an incomplete representation of reality. It is not incorrect per se, but it still incomplete. An example of this, are search engines which work through web crawlers indexing the web, but it is impossible to return a result for a website that is not indexable (Mejias ). Likewise, an ad that - upon clicking - redirects the user to an for Google unindexable landing page will not be allowed on Google Ad Exchange. So, when examining the mapping in chapter four, it is worth remembering that inclusion and exclusion of what is “node” in the hybrid human/non-human network of the agency greatly affects the mode of analysis.

Networks only see what in the same network and govern through structuring possible actions (Latour). With possibilities enabled by the network also come constraints within the network. By looking at those possibilities and restraints, the politics of the network are revealed (Gillespie). These politics can be set by the network owners such as Google or Facebook or when it is fully decentralised by the network’s logic itself.

Networks have gone from being metaphors to describe society, to a technological model to organise society, to what now seems to be the dominant lens through which we, as a society, understand our world. Hereby, networks have transcended their initial technology, becoming a knowledge structure. This internalisation combined with the inequality between network owner and network member is Mejias main grievance towards networks.

(23)

Conclusion

Participants in modern society are willingly and unwillingly under surveillance; publicly celebrating our accomplishments while atoning for our sins or lack of accomplishments in isolation, as per Kitchin’s revised view. Carefully tending to our data-double in a way that is indexable within networks since our presence within the networked society is all that matters.

After entering a network, one learns the doings and saying within this new world of practices. These are performed until they feel natural. Imprinted on the human to become a soldier or in this case, a media professional. If not adapted effectively enough, the soldier is removed for the network and loses its soldier identity. To understand the human role within programmatic advertising at DAN Netherlands, we will uncover these four types of

phenomena: action understandings, rules, teleological affective structure and general understandings, within the timespace of DAN Netherlands in order to create an understanding.

Media professionals are in a unique situation where you are a data subject yourself but also a data controller. Meaning that they collect and manage data for ad campaigns in order to meet the clients business goals. Putting them in the unique position where, unlike the general public, they know intimately how the data is collected, which data is collected and which is not. How does adiaphorization take shape within this sample? Within the company network, what information is indexable and collected on the employee and what stays in the paranodal. In the next chapter, we will examine the methods used to answer these questions.

(24)

3. METHOD

In order to show the human role within programmatic advertising at DAN Netherlands, we must set the stage by mapping the enclosure, the network, the world of practice. This mapping was done based on Actor-Network Theory, describing both human roles as non- human roles.

Ethnographic methods of workplace practitioner participation (Barton), thick description (Geertz) and in-depth interviews, we were also able to describe the unindexable part of the network, the paranodal (Meijas). By entering and participating in the network, the four timespace phenomena: action understanding, rules, teleological-affective structure and general understandings were identified through ethnographic notetaking and thick

description. By experiencing the network, In-depth interviews were conducted in order to illuminate further parts of the network that were engaged with but not participated in. These interviews were also useful in getting a second opinion on the findings.

Actor-Network Theory / Mapping

Despite the name, it is not a theory. Latour famously stated: “there are four things that do not work with actor-network theory, the word actor, the word theory and the hyphen (On

recalling ANT, 15). As Annemarie Mol gave that quote a little more meat explaining: “ANT is not a theory; there is no coherence to it. No overall scheme, no stable grid that becomes more and more solid as it gets more and more refined, the art is to move - to generate, to transform, to translate. To enrich. And to betray.”(257). The terms acquire their meaning relationally, through contrast and relation. Mol borrows from De Saussure’s version of semiotics here where words do not point directly to the referent but are part of a network of words. She uses as an example the word fish. The word gets its meaning from its contrast to meat and by its association with gills or scales.

In programmatic advertising, there is a strong focus on measurement and data. In order to compare apples to apples, these dimensions and metrics must have solid definitions. This is also part of the standardized structure described as one of the five developments needed for the existence of programmatic advertising in chapter 1.1 (Busch, 4). Dimensions are labels such as advertiser, campaign, ad type and creative. These dimensions are used to

(25)

filter and compare smaller parts of the whole campaigns, such as comparing the performance of one creative versus the other. Metrics are values such as impressions, clicks, conversions, view-through rates, view-completes. These metrics are used as key performance indicators to measure the success of a programmatic advertising campaign. They are also used as the basis for payment. It should therefore not come to a surprise that there can be a fierce debate about these definitions as it can affect the financial bottom line and can make the difference

between the success or failure of a key performance indicator.

Figure 4: Difference between impressions and viewable impressions (screenshot by author).

The metric ‘impression’ is a good example of this. It is a word to describe when a user

sees an advertisement but also has the association that it did left something with the person

that saw it. That by seeing the advertisement, it has left information, influenced their consideration of the product or brand, or going further drove them towards action. This metric is especially important since it is one of the most common is a basis for payment between clients and media agencies and tech vendors. So the definition and how it is measured is important to standardise. Impressions are measured by the ad loading on the webpage. Which means, it appeared somewhere on the page, and the user had the opportunity to see it. So in 2014, the Interactive Advertising Bureau a trade group comprising out of 650 leading media and technology companies, which fields critical research on interactive advertising and together with IAB Tech Lab develops technical standards and guidelines (IAB, Our Story) in conjunction with the Media Rating Council and the Mobile Marketing Association came with a new metric, viewable ad impressions (IAB, MRC Viewable Ad Impression Measurement Guidelines). The term viewable ad impressions seem pleonastic, but it has to be since the initial use of the term was inaccurate, as can be seen in figure 4. Viewable impressions get their relational meaning from its contrast to non-viewable

(26)

impressions and its association to the now formalised requirements that at least 50% of the ad's pixels have to be visible for two continuous seconds.

3.1.1 Actors

As the name suggests, actors act and - in ANT - actors are acted upon. Fundamental to ANT is to look at the effects of such an interaction. Within the network, actors are enacted, enabled and adapted by their mere presence in the network and their interaction with other actants within the network. To go back to the army example, when a human joins the army and receives his uniform and equipment, he is transformed into a soldier, and in order to remain in the army, they must adapt to that environment. The ability to enact upon others in the system is called agency. In this interaction, a process works through translation; in the army, an order may be given, and these orders go down the ranks. An officer might instruct a soldier to fire in turn, the soldier will instruct the gun to fire by pulling the trigger, in order to take the target down. In programmatic advertising, an advertiser might provide the media agency with a briefing, and the media agency will instruct a Demand Side Platform, to buy ad space by pressing a virtual button in order to buy ad space to get the attention of their target audience. The DSP is a platform used on the buy side of the programmatic ecosystem seen in figure 2 and 5, to buy ad space on the ad exchange. Opposite of it is the Sell Side Platform or SSP used by publishers to put their inventory up for auction. A short form for this chain of command and show of agency would be officer>soldier>gun=fire=take down the target and conversely, advertiser>media agency>DSP=buy ad space=get the attention from their target audience.

(27)

Figure 5: Simplified visualization of the Programmatic Ecosystem shown in Figure 2. (ProTableau, n.pag..)

The agency of the human and non-human roles are defined by how they can influence the other actors within the network. How orders are interpreted is called translation. In order to be translated properly, an order must be actionable; therefore one's agency is also defined by how well he can get the order across the network and for that the one who gives the order must understand the network and their capabilities. To understand the capabilities and

language of the network, it is also important to communicate constantly, not just down but in all directions. In order to guide and facilitate these orders, media agencies use briefs, debriefs and reports. The formalise the language and translation. A brief defines an order. A debrief details how it was translated before the confirmation is given to act and the report provides feedback as to what was the result of that action and what was learned from that action. These written documents also serve as an anchor, cementing which action was agreed upon, for reference and accountability. It is important to cement past actions in a network which is so dynamic, where innovation is constant and ‘perpetual beta’ the norm. What was capable three months ago might not be doable now, nor might it be desirable because of the introduction of a new member to the network; be it the addition of a data scientist, the deprecation of an ad format or the introduction of a privacy law such as the General Data Protection Regulation.

(28)

3.1.2 Modes of Ordering

DAN Netherlands, by dissecting its name one can already learn some key elements from it. It is the national branch of an international company, DAN. This global label was founded in 2014 (AdAge), two years after the Japanese company Dentsu bought the British company Aegis Media for a 3.2 billion pounds in 2012 (Sweney). The Network part of the names alludes to it being a holding group of multiple agencies, as can be seen in figure 6, in multiple countries. At the time of the DAN brand conception, it operated in 110 countries with 22.000 people working within the company. Now four years later, this has grown to 40.000

employees in 145 countries (DAN, ‘Who We Are’,n.pag..).

Figure 6: Labels within DAN (About DAN Deck, 2018)

This makes it a complex object of study but one that is representative of the industry as a whole. There are five major global networks; the other four are WPP, Omnicom Media Group, Publicis, and IPG (Advertising Age issue 72).

To settle on the scope of this thesis, we need to look at what can be fleshed out with the practitioner ethnography described in the next section. By participating in this company, it will not be possible to make any statements of the work practices in Tokyo or how the London-office interacts with the Brussels-office. However, it is possible to describe the relations between the labels, teams and actors at the Amsterdam office working to service the international client from the case study, see table one. The case study is unique in its

interconnectedness with individual actors in different countries but still representative of the action understandings and rules within the network. Likewise, the case study also operates on the same underlying structure as seen in the programmatic ecosystem visualisation seen in

(29)

figure 2 and 5. The case study will showcase what happens on the left side of the ad exchange in those figures, the buying side. Furthermore, most of the actors in the case study mapping work solely on this client, leading to a stronger teleological affective structure within this part of the timespace. This mode or ordering will be used throughout this thesis in order to

formalize some of the relation and levels.

Table 1: Modes of Ordering - DAN

# human actors Level Example

40.000+ Global Network Dentsu Inc.

400+ Local Network DAN Netherlands

50 Label Amplifi

10 Team Client Hub

1 Actors AdOps Consultant / Ad Server

x Client Philips

In chapter four, a case study mapping is laid out surrounding the actors involved with one international client. For this client, an inter-label group of actors has been formalised as a client hub. Client hub span countries also, which within the network are also called markets. A hub client will have centralised client services for multiple markets such as the six in the case study. Within the EMEA-region, the United Kingdom and the Netherlands are the common physical location of those hub. With local actors and hub actors working together. At the start of this research project, DAN Netherlands had three client hubs.

The client hub in the case study mapping shows the labels, teams and actor types which are part of this clients’ hub network. In the Amsterdam office, there are ten human actors as a part of the client hub. Four display traders, three social traders, a manager, a team lead and one ad operations consultant. They use one ad server, two demand-side platforms, one verification partner, one data management platform, one project management tools, one messaging application as well as multiple standard office applications less relevant to the case study. The case studies client hub facilitates the display and social programmatic campaign for six countries: France, Germany, Italy, Spain, The Netherlands and the United Kingdom. The hub has both client contact as well as agency contacts in all six countries. However, these were not included in the case study mapping. They represent the fringe of the hub’s network but are not part of the local hub-team itself.

(30)

Ethnography

In ethnography, researchers do three things: they observe, they record and they analyse (Geertz, 19). Ethnographic description is distinguished by four characteristics: it is

interpretative, it is interpretative of social discourse; it tries to simultaneously conserve text and context, its details are microscopic (Geertz, 20). This is what Geertz coined thick description’, not only is the flow of social discourse documented. When that flow is documented in microscopic detail; thick description captures it for others to perceive, appreciate, and contemplate further. This ability to re-viewing and re-analysing is what Bechky calls an ethnographic account of the work practices of knowledge workers.

Knowledge work is characterised by transforming “knowledge”, something that in the form of computer files is highly mobile, allowing for workers from all around the world to work on one product (Takhteyev). In the case study, the operational team of traders were in constant communication with other DAN and hub clients employees

When creating an ethnographic account, the researcher sorts out the structures of signification (Geertz, 8); mapping out the frames of interpretation to understands a people’s culture,

exposing their normalness without reducing their particularity (Geertz 13). These social structures form its own languages (Takhteyev) and are learned through social interaction (Bechky). One could liken it to enter a foreign country. A researcher could learn its history and its language, but when entering a social gathering, the locals would be quick to note them not following the social norm or pick up on their accent. These structures, therefore, are comprised of more than mere words. Interviewing the subjects alone would not be enough to create understanding. For that, one must actively participate (Takhteyev) by documenting these structures; it turns from a fleeting moment to something that can be re-examined.

One way to review the interpretative nature of thick descriptions and its analysis used in this thesis was to supplement the findings by asking the human actors in the network about the flow of social discourse in their world of practice. During the interviews, participants were asked what the workflow is from start to finish. Their descriptions showed not only what they add to the process in more details than a sole ethnographer could hope to describe. It also showed their knowledge and understanding of what their co-workers in another department do. In other words, the salient description of what the other actors in the network

(31)

contribute.

Workplace Practitioner Ethnography

Most ethnography start with selecting a culture and finding ways in. For example in their book ‘Doing Ethnographies’ Mike Crang and Ian Cook provide a list of suggestions on how to ‘cast your net’, this includes talking to friends, family, fellow student or faculty members, contacting the relevant organisations and the authors of relevant material (18). In workplace ethnography, researchers seek to spend time within a company (Takhteyev). This is exactly what Scott Rosenberg (2007) and Takhteyev did (2012). Rosenberg spend three years working at a software company where he experienced the whole process of making a software product. Takhteyev joined an open software project. Joining those organisations allowed them to write thick descriptions to describe what Takhteyev calls worlds of practice, “systems of activities comprised of people, ideas and material objects, linked simultaneously by shared meanings and joint projects”(2). Notice that he adds material objects to structures of significance.

This thesis, where the researcher is already an existing member of the sampled culture, is what Thomas Barton would call a practitioner ethnography (11). What Rosenberg and Takhteyev did Barton would then further categorise as a specialist ethnography. The specialist enters an ocean of new information for the first time, whereas the practitioner already knows how to swim. This is contrasted with a traditionalist way of doing ethnography, where the researcher is merely an observes from the beach-side.

The three activities used in any ethnography are prolonged engagement, persistent observation and triangulation of multiple measure- and data sources. By their very nature, practitioners fulfil the prolonged engagement criteria. Additionally, the practitioner tends to be more direct and specific to practical problems and practical relevance. However,

practitioner ethnographers must be aware of and acknowledge the connection between their intimate understanding of the research aim and their relationship with the research sample and must consider how this may affect the outcome (Barton). This is why reflecting on the culture and the work processes through the lens of theory is so important. It serves as an anchor for the practitioner, not to swim too far out as he will lose sight of the shoreline.

(32)

Highlighting Through Metonyms

In the process of ethnographic note taking names were documented; however in this thesis, metonyms were used to refer to the people in the observed workplace. There is a certain irony in the dehumanizing aspect of removed the individual aspect in a thesis aimed to highlight the human component of programmatic advertising. However, it was chosen for two reasons. Firstly, it simplifies the work field, allowing to project a framework of relevant actors. Secondly, it harbours their privacy and trust. Which is needed to collect the data and minimise the research bias so common to behaviour research in a laboratory setting.

There has been some debate on what metonymy are in relation to metaphors. Is the

metonymy a subclass of metaphors or are they two different things? Raymond Gibbs is in the latter camp, stating that while metaphors are based on similarity, where one conceptual domain is understood in terms of another; metonymy are understood within the same conceptual framework to express contiguous relation between objects. The models for these tropes are ‘A is B’ for a metaphor and ‘B for A’ in a metonym. The example utterance “Marc is a pig” is a metaphor, communicating that Marc shares a salient characteristic with a pig, maybe Marc eats a lot, which makes him similar to a pig. An example of a metonym would be “London says ‘no’ to Brexit-deal”. Here London refers to the United Kingdom

government, which is the physical place of where the government resided, making it their seat of power.

Metonyms work through a referring function where a salient characteristic of one domain represents the entire domain, “the essence of metonymy is highlighting” (Warren, 123 in Forceville). Charles Forceville expands on this writing that a communicator always has a reason to use a metonym and likewise has a reason to use that particular salient characteristic instead of another. This means that metonymy does not only have a highlighting aspect but also have an evaluative aspect. The limiting aspect of choosing a salient characteristic is that it must be conventionally associated with each other (Gibbs). However, this limitation is not absolute, as the chosen characteristic can also be contextual. Such as the example utterance “The ham sandwich is getting impatient for his check” (Forceville). It is not conventional to refer to someone as a ham sandwich. However, in the context of an eating establishment, it is easily deduced that the ham sandwich refers to a customer. It is also more efficient to

(33)

description of roles, there is also the process of selection and filtering. In this thesis, this was done by first mapping the people involved, describing a metonym to each role involved and then interviewing people representing these roles to get a deeper understanding of what sets them apart. It is more important to know what is required of the person to fulfil the role than it is to know than it is to know the banalest detail which separates the person from his or her co-workers. Within the industry and even the company in the case study, there are multiple job titles for a very similar role. For example, the job titles Campaign Manager,

Programmatic Trader, Programmatic Consultant, Paid Social Executive Paid Media Executive, and Digital Performance Specialist could all be categorised as traders.

The planner metonym, for instance, highlights their part in the strategic creation of the media plan. While the server metonym, highlights their part in the technical aspect of the ad serving process and the trader metonym highlights their part in trading media budget using a wide array of buying tactics to fulfil the goals set in the media plan.

In-depth Interviews

There are different conceptions of how interviews produce knowledge. Two contrasting metaphors discussed by Steinar Kvale are those of ‘the miner’ versus ‘the traveller’. The miner uncovers a truth that already exists. The miner does not shape it; the miner only

collects that which is ‘given’. This uncovering is in line with positivist and empiricist schools of thought. The ‘Traveller’ on the other hand, wanders together with the subject. The

knowledge is not given but produced with the subject. Through conversation, the interviewer leads the subject to new insights and the subject leads the interviewer to new questions: there is a transformative element to the journey. This metaphor is in line with the postmodernist and constructivist (Kvale, Doing Interviews, 19-21). It is necessary to listen to their explicit descriptions, the implication and meaning behind those descriptions, and what is said between the lines to get insights into the point of view of the respondents. (Kvale, Doing Interviews, 11). This is where the synergy of the ethnography and in-depth interviews comes to fruition as the researcher already has established a rapport with the subject, and reading between the lines is facilitated when the researchers as actively participate in their world of practices for an extended period of time.

(34)

Key features of the in-depth interview according to Legard, Keegan & Ward are: being flexible, interactive in nature, using a range of probes and other techniques and being generative. The flexible structure allows for conversation to flow more naturally and personalised in the sense that some topics may be explored more closely based on the

interaction between the researcher and respondent. The researcher uses a range of probes and techniques to keep conversations going and guide the conversation towards new findings. (Leegard, Keegan & Ward 141-142).

In in-depth interviews, it is important that the respondent can convey their answer in the most natural form with a much specificity as possible (Legard, Keegan & Ward, 142). This natural form is essential for the discovery of both commonalities as diversity. When the interview is done face to face at the same time and place as the subject, it allows the

researcher to better read and respond to social cues. These cues range from a shift in the tone of voice, body language to a long pause, adding to richness to the interactivity feature. The importance of this is less prevalent in the case of expert interviews, when the expert is

interviewed on things or people that have nothing to do with the interviewee (Op den Akker). However, the interviews conducted in this thesis were not just about the mechanics of

programmatic advertising but also about their part in it and the researcher is a fellow-practitioner within the same team, making reading social cues more important. The social component already became prevalent in the first interview. After which the adjustment was made to ask the subjects to refer to the people involved by their job title. This also showed that the importance of reading social cues where the respondent might presuppose hostility (Legard, Keegan & Ward) or perhaps tailoring the timing of the interview to a more relaxed stage in the ebb and flow of work pressure.

Positionality of Researcher and Subject

The active participation of the researcher in the company helps to establish a comfortable connection. However, in the roles between co-workers are different from that of interviewer and respondent. Especially regarding the power relation which Kvale describes in

‘Dominance Through Interviews and Dialogue’. Since the researcher has two roles within the company, it was important to signal what goes ‘on the record’ and what is said in confidence. Often permission was asked after the fact. It is good to note that because no individual names

(35)

were used in this thesis, the answer was always ‘yes, you can use it’. Asking for permission protected the relationship build. This shows how useful role relevant metonymy is helpful when documenting workplace interactions and conducting interviews. The subject can discuss work events and practises without feeling guilty for bad-mouthing their co-workers. The exact identity is also not relevant to research. Therefor documenting exact identities would violate Gibbs maxim of relevance

In qualitative interviews, social scientists attempt to understand the subject's point of view and to unfold the meaning of their lived world. This thesis is no different as it tries to understand the point of view of human professionals in the programmatic advertising landscape and unfold the social structures of signification. According to Kvale social scientists often refer to this interview as a dialogue which suggests mutuality and

egalitarianism (Kvale, 481). However, this is a misnomer as an interview is a ‘conversation with a purpose’, serving the purpose of the interviewer and the interviewer exercises extensive control over the conversation. It is the researcher that initiates the conversation, poses the question often only listening to the answers without providing his own opinion. It is even considered poor taste when the respondent starts to question the interviewer.

Furthermore, the researcher controls the length of the conversation and may steer the questions and answers to fulfil a hidden agenda. Finally and perhaps most importantly the interviewer has the exclusive privilege to interpret and report what the respondent meant between the lines and what that means with the theoretical framework the researcher works from (Kvale, 485).

The interactive process of in-depth interviewing has five stages. These are arrival,

introduction, beginning, during, ending and after. During these stages, it is the responsibility of the researcher to remove any source of anxiety and to establish a comfortable connection (Legard, Keegan & Ward).It is near impossible or in the least inefficient to hold a true egalitarian interview, however, to mitigate the asymmetrical power relations, the following steps were taken. The respondents were allowed to name the time and place of the interview. Letting the participant have control over the environment helps put them at ease and also serves to lower the threshold to participate. At arrival, small-talk about other subjects was used to help establish a comfortable connection. Before beginning, the researcher reiterated

(36)

the nature and purpose of the research topic, approximate time length and asked permission to record. This helped manage expectations, reinforces the sense of control and reassures them; there will be no unexpected turns. Furthermore, the respondents were allowed to review the transcripts and allowed them to clarify and even redact their answers. However, only one interview redacted one paragraph. The first respondent felt that in retrospect that their statements were too personal and hyperbolic to the point where it felt untrue.

(37)

4. ENTERING THE TIMESPACE

The following section details the findings as a result of the practitioner ethnography. It will showcase a coming into being as a human in the programmatic world. In 5.4.1, my

experience in entering the company is described. The welcome I received and the label-signalling within the onboarding process. In 5.4.2, the first team meeting in which I immediately received clients from my new team members, how it struck me how casually they gave them away and how proud I was to receive the well-known brands.

Onboarding – Entering the Fold

Figure 7: Welcome e-mail (screenshot by author).

The first thing I noticed after signing my contract was how structured the entry process is. Every month a group of new employees start and are greeted by the welcome email shown in

(38)

figure 7. After that, the company fills their first week with meetings to get them up to speed to the more general elements of being part of the company. I was in a group of fourteen new hires, most of them joining iProspect. It was easy enough to identify. Everyone had some piece of brand merchandise, be it a notebook, a pen or a keychain. Everyone showed their colours. Green for iProspect, Blue for Carat, Yellow for Vizeum. I had a grey notebook signalling being part of a specialist label. People grouped together according to colour. `

I became onboarding buddies with a young woman from another specialist label. She started at Amnet, the programmatic buying branch of the network. This was her first job in advertising, and she was eager to learn, jotted down vigorously every meeting. Joining the programmatic landscape is somewhat akin to being dropped in a foreign country and having to learn the language. Impression, Clicks, CTR, CPM, CPC, CPA, pixels, Open Marketplace, Private Marketplace, the list goes on and on. After her first contract was up, she moved to a publisher. As a very organized person herself, she had a lot of difficulty handling the chaos of overly tight deadlines.

At a general business update held by the CEO and COO about the status of the company and its future, it was suggested that the annual employee turnover is almost a quarter of the company.

Joining the AdOps Team

The AdOps team, which I joined on the second of October 2017, consisted of four people. They used to be six people strong but two members left after being approached by IPG Mediabrands, a competing media agency network. Sufficed to say, my new team members were eager to re-divide the workload, and upon my arrival, they wasted no time assigning me my own clients.

The clients which we service are organized in an online spreadsheet, going through it almost all the brands on it were brands I heard of before. Coming from a small mobile DSP, mostly catering to small business’, this list was a little intimidating and exciting at the same time. It gave me a strong sense of ‘welcome to the big boys’. What added to that feeling was that the team meeting was held in the conference room meant for the managing board. It had a massive dark wooden table, a giant TV for presentations and a wall to wall window looking

The Human Side of Programmatic Advertising

Table of Contents

1. INTRODUCTION

2. THEORY

3. METHOD

4. ENTERING THE TIMESPACE