Do Not Track

Turning Pulp into Prototype with Agile Documentary

Do Not Track is an interactive Web documentary series about internet privacy directed by Brett Gaylor and co-produced by the National Film Board of Canada, ARTE, Bayerischer Rundfunk, and Upian. It is made up of seven episodes, each roughly 7 minutes in length, which played on the Internet biweekly between April and June of 2015.

In the documentary’s original proposal, director Gaylor quotes Jeff Hammerbacher, a former Facebook employee: “The best minds of my generation are thinking about how to make people click ads.”1 Hammerbacher’s statement solidifies what Gaylor identifies as having gone awry with the Internet. Do Not Track addresses critical questions around this phenomenon: when and how did the Internet become home to targeted advertising, where a user’s browser history determines the advertisements she or he sees? How do data brokers collect and sell information about users to hundreds of companies? If we don’t want to be tracked, how do we fight back?

To expound on these questions, Do Not Track takes on the same production strategies used by many Web developers. It is a documentary for the Web, rather than a documentary on the Web. It optimizes the story for a Web experience by offering short, personalized video content that keeps the viewer focused, as well as basic interactivity that keeps the user constantly involved. In a sense, Do Not Track’s strategy is recursive: it makes use of the very same analytics‑based optimization techniques that it critiques in order to customize content and to engage the user.

Its storytelling method is not the only thing Do Not Track tailors to the Web. It also adopts the “agile” design production methods frequently used in tech and design fields. Documentaries could be considered agile when optimized through Web analytics and iterative, on-the-fly design. For example, depending on whether users have registered on the Do Not Track website and how they have answered a preliminary questionnaire, they are directed to different versions of a landing page, and they experience an episode with personalized content.

During the life cycle of the project, the Do Not Track filmmakers also developed communication and collaboration strategies to work efficiently with an international and interdisciplinary team. This case study focuses primarily on the methods Do Not Track creators used to develop a documentary in an agile framework, how they collaborated internationally, and how they composed a script under these conditions.


Agile documentary

Traditional filmmaking generally follows what Brett Gaylor refers to as “waterfall production methodology.”2 This means that research, filming, editing, and release follow each other in strict order, using a top-down approach. Waterfall development techniques put all the planning and research up front, followed by design, then finally by development and testing. If something unforeseen emerges during the testing phase of waterfall development, it is difficult to go back and modify the original design.

By contrast, Do Not Track borrows its production language from software development, where a host of development methodologies with names like agile, iterative, scrum, and others have replaced traditional waterfall methods. These newer methods focus on rapidly creating prototypes, cycling through the entire process of product development quickly, and changing approaches on the fly as new information emerges.

Gaylor encourages traditional documentary makers to adopt design and development principles from other Web-native disciplines like software development.3 He thinks that by considering agile Web production techniques as models, it may be possible to create a new production process for documentaries, accommodating a more rapid product cycle as well as Web-native interactivity and responsiveness. Gaylor explains that the waterfall method is still widely used because it fits into the funding and exhibition models (i.e. festival premieres) of the film and documentary worlds. Funders do not want to stray away from these conventional methods, because exhibition models are not quick enough to adapt to new, more agile filmmaking methods. “If you want your work seen at the festivals, it needs premieres. But what is a premiere on the Internet?” Gaylor asks.4

One way that an agile approach benefits the Web documentary format is that each episode can be changed based on the analytics of the prior episode. User analytics in this situation could refer to how long a user spends on a certain page, how a user arrived to that page, and what kind of device the user is using to access the page. Based on these metrics, a user can be directed to a personalized or a template episode. For example, a user experience flow described by the Do Not Track team illustrates this process as follows:

Figure 1.

For Do Not Track, interaction plays out on multiple levels, not just the “navigate your own way through a story” level.5 Do Not Track illustrates how unconscious and perhaps unintentional interaction takes place between a user and various digital environments. These invisible interactions are relevant to any Web-based text (for example, whether or not users open a linear story on the digital NYT; how long they stay there; and at what point they leave). But the advantage of Web interactives that explicitly use this invisible layer of interactivity is that they can apply this phenomenon to their needs, shapeshifting and modifying story content as needed to optimize the metric of viewer engagement.

In addition to the personalized aspect of Do Not Track, the documentary also changes based on aggregate analytics. Aggregate analytics refers to the statistical analysis of all user behavior, which helps to create changes that improve user retention—in other words, to keep eyes on the site. Consider how companies like BuzzFeed release multiple headlines for the same story, testing each to see which has the highest click-rate. Gaylor says that Web documentaries can also apply these techniques. By thinking of each episode and each two-week production cycle as a separate project, rather than as a single long project, it becomes easier to adapt to previous results and learn from them. For example, by using aggregate data, the Do Not Track team optimizes the design of the homepage and changes the landing page.

Gaylor compares the phase of releasing a new episode every other week to a festival run of a film, in which creators get useful feedback and press attention, then have time to make changes before officially releasing the film. However, as noted in the case study of A Short History of the Highrise, agendas and processes as well as values and traditions of documentary and interactive Web content are sometimes at odds. The textual flexibility of analytics-driven work presents a dilemma: personalization and reader retention do not always concur with journalistic attitudes about the stable documentation of knowledge. Many journalists still view their work as producing “texts of record,” which remain stable over time, whereas personalized content provides for a more dynamic experience that changes for every audience. Although fully documented storyboards could serve as texts of record, the execution of documentaries that rely on personalization, like Do Not Track, opens up new possibilities and thus unexplored conventions for indexing or archiving this type of work.


Remote project management and the RACI model

One remarkable feature of Do Not Track is its international and interdisciplinary production team, which collaborated remotely for the majority of production. The team used what Gaylor calls the “RACI” model to collaborate (see Figure 2).6

Figure 2. RACI model. Source: Brett Gaylor/

As shown in the figure, the model assigns each team member to categories entitled Responsible, Accountable, Consulted, or Informed. This approach indicates the role of each member upfront and clearly. People who are “responsible,” “accountable,” and “consulted” all contribute creatively to the project, but in the end, the final decision is the responsibility of the “accountable” person. The “informed” team members receive updates about the project, but they do not collaborate creatively. This coherent distribution of responsibility prevents micromanaging and gives people enough room to experiment, says Gaylor, who firmly supports the method.7 “That’s how I think you get good results when something is distributed like this. If you are in there micromanaging every step and nobody is having any fun, you’re going to just fail, you’re going to have ultimate burnout,” he says.8

The team behind Do Not Track was organized into separate core groups for story, design, development, montage, project management, and conversation management. Figure 3 illustrates how the RACI model was employed across these groups: the script team created the story, the UX team designed all the elements that went into the project, the development team coded the interactive documentary, and the montage team edited the film.

Figure 3. Organization of the “Do Not Track” team according to the RACI model. Source: Brett Gaylor/

Gaylor explains that on an interdisciplinary team, crosstalk between developers, designers, and project managers is a necessity.9 A project manager, for example, might not need to be able to code, but should be able to ask the right questions of coders. Gaylor says, “A good product manager is going to ask the right sort of naïve questions like, ‘Did you think about it this way?’ Or, ‘Oh okay I understand, it takes too long to create this because of this service, have we considered writing this one ourselves, what would that take?’”10

There were two organizational teams in Do Not Track. The project management team was responsible for the production of the whole project, from filming, to Web design, to development. The conversation management team handled outreach, marketing, and user experience outside of the Do Not Track interactive documentary itself (i.e. blog posts about Web security, Tweets, emails to each registered user, etc.). Figure 4 illustrates how the conversation team worked.

Figure 4. The conversation team methodology. Source: Brett Gaylor.

But how do you communicate with a team dispersed around the globe? Collaborators on the project were from Quebec, British Columbia, France, and Germany, which made communication a challenge. Do Not Track began with several large, in-person meetings in Paris, where team members developed a style guide and a production schedule. After this, the team relied heavily on real-time collaboration tools like Google Docs, Basecamp, and Slack, as well as a weekly online action meeting. For Gaylor, communication is synonymous with project management. So in order to collaboratively brainstorm and make decisions, team members discussed the script through Google comments, as shown in Figure 5.

Figure 5. Script discussion via Google comments. Source: Brett Gaylor.


Storyboarding for interactive documentary

Writing for interactive documentary is a challenge: multiple types of media, interactivity, and personalization all have to be captured by the script. How does a team organize all this information? How does it create a script format? These are complex questions, and the storyboard of Do Not Track reveals some unique solutions.

First, interactivity is communicated through state-based storyboards, where a user does not merely arrive at a timestamp, but also has several variables attached to that timestamp based on decisions she or he made previously as well as on aspects of personalization. Second, “fail‑states” on the storyboard convey scenarios when things do not go as planned (e.g., a viewer’s Facebook page does not reveal enough data about the person to proceed with personalization). The script also has additional labels to mark different types of media. These labels offer a convenient vocabulary for writing interactive documentaries. Figures 6-9 depict the storyboarding process along with labels used for some of the states and types of media.

Figure 6. Video label. This label marks the video content shot specifically for “Do Not Track.” Note all the different languages for which the production team planned. The video automatically plays in the language appropriate to the user’s location. Source: Brett Gaylor.

Figure 7. Text input label. This label refers to instances in which users must input information about themselves to personalize their experience. Note that the video is in “cinemagraph style,” which means it is a moving photograph. It can loop infinitely until the user enters personal information. Source: Brett Gaylor.

Figure 8. Real-time label. This label marks instances when content is changing according to input in real time. Users see information that is dependent on data they previously provided, as well as on data that the “Do Not Track” application programming interface (API) collects from their IP address, such as where the user lives, which computer she or he uses, etc. In the screenshot, this is noted in the left column as “Realtime User Dynamic Information.” Source: Brett Gaylor.

Figure 9. Fail-state label. What if the user doesn’t provide any information, or the information provided is not adequate? For those scenarios there are alternative scenes, which are marked as “Fail States.” Source: Brett Gaylor.

Other labels include “Audio Only,” for cases in which the user hears audio without an image, “Archive,” for clips from famous movies, Web images, etc., and “Animation,” for cases when animated images are used.


Walking the tightrope

Do Not Track presents a highly innovative production model that borrows from Web and technology development as well as documentary filmmaking in order to find the best process for the form. It brings techniques from Web development into documentary production that include agile methods, long‑distance collaboration, and state-based storyboards (meaning several variables based on personalization and prior choices are attached to a user, not just a timestamp). Furthermore, in order to illustrate its subject, Do Not Track uses the same tools as advertising companies: personalization and social media logins. It does not preach against the use of analytics or algorithmically customized content; rather, it relies on these tools to demonstrate their capabilities and to show how different parties are tracking users online, thereby sparking contemplation of the functions and uses of the tools. By revealing how these tools work, the documentary seeks to inform its audience about these technologies and what implications they have for everyday life, while also inducing the audience to take action steps by leading them to the Do Not Track blog, which contains articles and links about Web security that can help ensure their privacy online.

While agile documentary filmmaking fits naturally into the distribution system of the Web by making content immediately available and modifying it on the fly, it still faces the challenge of balancing old with new. In the case of Do Not Track, the unfamiliarity of the production method caused several challenges; having an extremely tight deadline, working with a team dispersed in different time zones, and adapting the interface and the content of the project into four different languages forced the team to constantly reinvent their process. Furthermore, Do Not Track needed to address the central tension between the documentary and technology worlds: the idea of authorship. In documentary, the author’s responsibility is to present a cohesive and complete vision to the audience, whereas in the technology world, the product is boss. It doesn’t necessarily matter what the author thinks if users don’t find value. That’s why there is such an emphasis on user-centric design, A/B testing (running multiple variations of an interface or text for each user and then choosing the right one), user testing (sitting down with real users and observing their behavior), market research, and segmentation. These are the methods that allow people to make products that users want.

Do Not Track attempted to bring these distinct approaches to authorship together by building a platform that made people reflect on issues about privacy, while also telling a compelling story. Gaylor says that his role as the author was “to set the guidelines and the parameters—to design the system—which includes the story.”11 The challenges of trailblazing a new production method resulted in insights for future projects as well. For example, Gaylor learned that it would help immensely to have a break in the middle of the project to allow a “retrospective,” where team members reflect on the process to date and share what could be done better; this would enhance the iterative nature of the production, he says.12 More frequent and insistent user testing would also, much earlier in the process, address issues about the interface and content that confuse users.

Integrating a new mode of production with established funding and distribution models for traditional film, documentary, and journalistic storytelling is still a challenge to be met by future filmmakers who venture into this new territory. Do Not Track demonstrates how new analytic tools for understanding user behavior can represent a powerful source of knowledge about the audience to filmmakers, and how they can help to identify problems or miscommunication with users. As we have shown, these tools are not without risk; the rapid prototyping cycle can be reductive, meaning user-centered filmmaking risks leading to flat content. However, by combining new tools with precise production management models, Do Not Track successfully walked on a tightrope—balancing the flashy with the meaningful, the brief with the comprehensive—and could henceforth serve as a framework for other interactive documentary production cycles.


1. Ashlee Vance, “This Tech Bubble is Different,” Bloomberg Business, 14 April 2011 [http://www.bloomberg.com/bw/magazine/content/11_17/b4225060960537.htm]

2. Skype interview with Brett Gaylor, Cambridge, MA, 29 April 2015.

3. Ibid.

4. Ibid.

5. Email correspondence with Gregory Trowbridge, 10 June 2015.

6. Gaylor, 29 April 2015.

7. “Agreeing on Roles and Responsibilities: Summary of RACI” (2015) [http://www.valuebasedmanagement.net/methods_raci.html].

8. Gaylor, 29 April 2015.

9. Ibid.

10. Ibid.

11. Ibid.

12. Gaylor, 29 April 2015.

13. Ibid.