Case Study 28-2: The RWDSU and Amazon Bessemer — Organizing Against the Algorithm

Overview

The 2021 union election at Amazon's BHM1 fulfillment center in Bessemer, Alabama — organized by the Retail, Wholesale and Department Store Union (RWDSU) — was one of the most high-profile labor campaigns in recent American history. The campaign ultimately failed (workers voted against the union 1,798 to 738, with the NLRB ordering a re-run election in 2022 that also failed), but it generated an extraordinary volume of worker testimony, legal proceedings, and journalistic investigation that illuminated the specific experience of algorithmic management from workers' perspectives — and documented Amazon's use of surveillance in opposing the organizing campaign.

This case study examines Bessemer not as a story of union defeat but as a case study in how workers experience algorithmic management and how employers use surveillance infrastructure to contest worker organizing.

Worker Testimony: Life Under the Algorithm

In the months before and during the Bessemer election, dozens of workers gave detailed accounts to journalists, to NLRB investigators, and in public testimony of what working under Amazon's algorithmic management system felt like.

The rate and the anxiety:

Jennifer Bates, a BHM1 worker who became a prominent spokesperson for the organizing campaign, testified before the Senate Budget Committee in March 2021. She described the rate system's psychological effects: "Amazon's system tracks every minute of our shifts, and we're constantly aware of our time off task, our rate, and our standing. If you don't keep up, you get warnings. If you don't respond to the warnings, you get fired. The system never stops. It doesn't know if you're sick. It doesn't know if you need a bathroom break. It doesn't care."

Bates's testimony illustrates a specific form of psychological burden that the algorithmic management literature calls "monitoring anxiety" — the continuous background awareness of surveillance that makes it impossible to not think about performance, even during moments that are not formally monitored.

The bathroom break reality:

Multiple Bessemer workers described the practical reality of bathroom breaks under the rate system. The facility's bathrooms were at the facility's edges; walking to a bathroom and back from certain bays could take 6–8 minutes. Workers reported making tactical decisions about bathroom timing — waiting until they were in a "good" bay close to a bathroom, or holding off bathroom breaks until they had sufficient buffer above rate to absorb the idle time without a warning.

Workers with medical conditions — bladder conditions, digestive conditions, conditions requiring regular medication — described the rate system as particularly hostile to their management of those conditions.

The task assignment opacity:

Several workers described frustration not just with the rate requirements but with the opacity of task assignment. Workers who were consistently assigned to the most physically demanding bays wondered whether this was random, intentional, or a function of their performance history — but had no way to find out. One worker described asking his supervisor why certain workers got "the bad bays" and receiving the response: "That's what the system assigns. I can't see why."

Amazon's Anti-Union Campaign and Surveillance

Amazon's response to the Bessemer organizing campaign was intensive, well-funded, and documented in considerable detail through NLRB proceedings and journalism.

Mandatory captive audience meetings:

Amazon held mandatory daily meetings at which managers presented anti-union messaging. Workers who did not attend were noted absent. The NLRB subsequently found that some of these meetings may have involved unlawful coercion, though the election outcome was not overturned on this basis.

The surveillance signal: The mandatory attendance at anti-union meetings was itself tracked through the badge system — the same infrastructure used for productivity monitoring now used to ensure compliance with the anti-union messaging program. Workers who frequently left meetings early were noted.

Digital surveillance of organizing:

Amazon's approach to digital surveillance at Bessemer closely paralleled the patterns documented at JFK8. Workers reported that known union supporters were more closely scrutinized on the production floor — receiving supervisory attention when their rate dipped, receiving TOT warnings at what felt like lower thresholds, and being called into supervisory conversations about performance issues shortly after visible organizing activity.

The RWDSU filed multiple unfair labor practice charges with the NLRB alleging that Amazon had: monitored protected concerted activity (including organizing communications) through its electronic monitoring systems; conducted surveillance of worker organizing meetings held near the facility; and applied performance discipline selectively to workers identified as union supporters.

The USPS mailbox controversy:

The most documented and legally consequential Amazon surveillance action at Bessemer was the installation of a U.S. Postal Service drop box in the BHM1 facility parking lot. The NLRB election was conducted by mail ballot — workers received ballots at home and could return them by mail or to collection boxes. Amazon reportedly lobbied for (and was granted) the installation of a USPS drop box in the facility parking lot, which the RWDSU and the NLRB found problematic.

The NLRB found that the parking lot mailbox interfered with the election because workers who used it might be observed by supervisors, creating a chilling effect on the privacy of the ballot process. The NLRB's initial decision ordered a re-run election on this basis.

The mailbox is a surveillance artifact in precisely the sense this textbook uses: its placement — in a space controlled by Amazon, observable by supervisors, adjacent to badge readers that tracked who was entering and exiting the facility at what times — transformed the act of mailing a ballot into a visible, potentially attributable act. Whether anyone was actually observed using the mailbox is less important than the fact that the architectural possibility of observation existed.

The Algorithm as Anti-Union Technology

The Bessemer case illustrates a structural feature of algorithmic management that is important for understanding its labor relations significance: the performance management algorithm is, by its nature, an asymmetric information system that advantages management in labor disputes.

Management has real-time data about individual worker performance, patterns of deviation from standard behavior, and changes in activity patterns that might correlate with organizing involvement. Workers have no equivalent data about management. Workers cannot see the algorithm's assessment of their organizing risk; they can only observe its consequences in increased scrutiny, unexpected discipline, or adverse scheduling.

This asymmetry does not require that Amazon explicitly use its algorithmic system to target organizers. The system creates the conditions for targeting without explicit intent: any manager who believes an organizing worker might be more carefully scrutinized has the data infrastructure to do so without creating a clear evidentiary record of discriminatory intent.

What Failed and What Succeeded

The Bessemer campaign failed to win recognition in both the initial vote and the re-run election. Understanding why is important for understanding the limits of organizing against algorithmic management.

What worked: - Generating national attention to the specific conditions of algorithmic management - Building a worker-leader cohort who testified effectively in public forums - Creating NLRB proceedings that documented specific Amazon surveillance and coercion practices - Inspiring subsequent organizing campaigns, including the successful JFK8 campaign

What was difficult: - Amazon's scale of anti-union spending and organization - The mandatory captive audience meetings, which Amazon (legally) used intensively - The monitoring infrastructure's chilling effects on organizing conversations within the facility - The race and class dynamics of Bessemer, Alabama — a majority-Black workforce in a region with limited alternative employment options, subject to specific economic pressures that made the cost of potential retaliation more acute

The structural lesson:

The Bessemer campaign suggests that organizing against algorithmic management requires strategies that are adapted to the specific conditions of algorithmic employment: encrypted communication channels for organizing conversations, organizing structures that don't depend on facility-internal communication that may be monitored, and legal strategies focused on the documented unfair labor practices of algorithmic surveillance systems.

Discussion Questions

1. Jennifer Bates testified that Amazon's system "never stops" and "doesn't care." What does this description reveal about the specific psychological burden of algorithmic management, as distinct from intensive human supervision?

2. The USPS mailbox installation was found to interfere with election integrity because of the possibility of surveillance, not documented surveillance. The principle is: potential observation has the same chilling effect as actual observation. How does this principle extend to the algorithmic management systems described in the chapter?

3. The Bessemer campaign failed in both election rounds but generated significant legal proceedings and inspired subsequent organizing campaigns. How should we measure the success of organizing campaigns against algorithmic management? Is winning union recognition the only relevant outcome?

4. Workers described receiving discipline for performance issues shortly after visible organizing activity, but could not prove that the discipline was retaliatory rather than legitimate. How should labor law address the epistemological problem created by algorithmic management systems that produce discipline automatically — making discriminatory enforcement difficult to distinguish from legitimate enforcement?

5. Amazon's anti-union campaign was documented extensively, yet the campaign succeeded. What does this tell us about the relationship between documented unfair labor practices and actual outcomes in labor organizing? What factors beyond legal violations determine organizing outcomes?