4 Workforce Management: Scheduling Call-Center Workers

Students are positioned to consider the regulation of Workforce Management Systems. These systems draw upon predictive analytics— big data sets, machine learning, and optimization algorithms—to forecast worker demand and to schedule workers into shifts, subject to employment laws, business rules, worker preferences, and other constraints. Students explore how forecasting and scheduling algorithms can lead to unstable and unpredictable work schedules. Such schedules, while perhaps economically advantageous to business, might negatively impact the wellbeing of workers and their families with potentially major implications for the public interest.

INTRODUCTION

Timers, clocks, stop watches; the work hour, the shift, the schedule: For hundreds of years such tools have been used to guide, and often control, the rhythms of work and everyday social life.

Employers depend particularly on schedules. Schedules essentially forecast the amount of work needed to be done and organize workers to carry out the work at set times. But workers, quite obviously, have multi-faceted lives and identities, with roles outside of work—student, parent, caregiver, for example. A work schedule, especially one that changes frequently and unexpectedly, might make it difficult to fulfill such roles, with consequential ripple effects throughout society.

So called “crazy” schedules can wreak havoc on the personal lives of workers in high paying jobs. People on video game development teams, for example, often work 100-hour work weeks during “crunch time.” Investment bankers, and their staffs, often need to be continuously on call for several weeks when finalizing a business deal. The impact of problematic work schedules is likely to be even more acute for workers in low wage jobs, such as restaurant, retail, and call center workers, jobs often held by women and ethnic minorities.

Consider this advertisement, which promises worker input into the creation of work schedules:

Job: Retail Sales, *Flexible Scheduling Option!*, Part-Time


Opportunity:
This position uses a scheduling plan that allows an associate to participate in the creation of his/her work scheduling by managing availability and identifying a preferred work schedule. This position allows the maximum amount of scheduling flexibility.

Qualifications:
Ability to work a flexible schedule, including mornings, evening, and weekends, and busy events such as the day after Thanksgiving, special Big Event days, and the day after Christmas, based on department and store/company needs.


The key term is “flexibility,” that is, allowing the worker to create a schedule that meets his or her preferences. Things get murky, however, in the qualifications section of the ad, where we read that the employee should have the ability to “work a flexible schedule.” In short, even in this ad, we can see that the employee’s and the employer’s view of “flexibility” appear to be quite different. What is at stake with this apparent tension and how might it be addressed?

Workforce Management Systems make flexible scheduling on an hour by hour basis feasible. One key component of these systems is a forecasting algorithm. Based on machine learning techniques and large amounts of data, the forecasting algorithm enables these systems to precisely forecast the business’s staffing needs. The inputs to the forecasting algorithm might include historical data on sales, current events, weather forecasts, advertising campaigns, social trends, or for that matter, any kind of data. The forecasting algorithm is a learning algorithm because the algorithm learns how to weight particular features in the data sources, which in turn, allows the algorithm to improve its forecasts.

Consider how the forecasting algorithm might work if, hypothetically, it was deployed in a local coffee shop. At the beginning of a week, the algorithm might predict that the shop will be extremely busy on Thursday, between 6:45 – 9:00 P.M., because the evening weather is expected be cool but dry, the high-school football game will finish around 6:30 P.M., a sales coupon promotion will end on Friday, and since Thursday evenings are normally busy, even without taking into account the nice weather and the game.

The forecasting algorithm might look ahead 10 days, slicing staffing needs into 15 minutes intervals for all opening hours, and create a forecast. Further, seeking the best possible forecast, the algorithm scans for new data, updates worker attendance figures and, in turn, updates its forecast every hour.

While such forecasting is perhaps not needed for a single, small, independent shop, the economic benefits seem compelling if the coffee shop was one franchisee of many hundreds, if the software was managed centrally, if the software was designed to take local conditions and data sources into account, if available workers were distributed across a region with several such franchisees, and if the software was known to support worker’s conceptualization of “flexibility.” Clever and robust labor forecasting algorithms, in other words, might give a retail business with many stores competitive advantage.

Workforce Management Systems also schedule workers. The scheduling algorithm is a kind of optimization algorithm. Optimization algorithms seek to maximize an “objective function,” subject to a set of constraints. In this case the objective function might be profit, that is, sales minus costs per shift, and one key constraint is, of course, to meet the staffing need forecast. Other kinds of constraints might include:

  1. Employment laws, such that no part-time worker can work more than 20 hours in week
  2. Worker preferences, such as times when a worker is available or not available
  3. Business rules, such that a mix of both experienced and inexperienced workers are on every shift
  4. Unexpected contingencies such that a worker gets sick and will not make a shift or misses the bus and so will be tardy.

In addition to such constraints, the scheduling algorithm might use other data sources related to worker performance. For example, data related to the likelihood that the worker will show up to his or her shift, past sales performance, and customer satisfaction feedback. All of this data, and much more, might plausibly be used by the algorithm to obtain an optimal match of workers to shifts.

The forecasting and scheduling algorithms, at least in one view, turn the coffee shop into a kind of a demand and response machine. As forecasted demand for workers comes and goes, the machine responds, scheduling workers into shifts. The algorithm might send messages to workers about last minute shift openings, incentivizing workers with extra pay. Or, workers might bid on shifts through auctions. Or, workers might be compelled to take shifts to avoid penalties such as fewer hours in the future.

In a different vein, note that some crucial information might only be collected by surveilling employees or at least by invasive monitoring; for example, what workers say and do on the job, where workers live and how they travel to work, and so on. Perhaps data on workers’ sleep patterns and indicators of psychological well-being are decisive, substantially improving the forecasting and scheduling algorithms. There appear to be few, or perhaps no, technical limits on what data might be sensed and collected. But as a matter of human dignity or of the public interest, perhaps some data should not be collected and used by algorithms.

Returning to the value of “flexibility,” Workforce Management Systems are likely to create more profitable schedules if all workers are always “on call” and available to work. This is so because, in general, with fewer constraints, the scheduling algorithm has more options for finding the optimal mix of workers for any given shift. So, from a business or corporate point of view, the definition of “flexibility” is one related to the availability of workers. But, for workers work becomes unpredictable and unstable. In the last two decades major retailers changed their worker status, from about 75% full-time workers to about 25% full-time.

The trouble with this view of “flexibility” is that the public interest is pushed to the background. Workers have duties, aspirations, and lives outside of work. But, week- to-week schedules have serious ripple effects for workers. The New York Times, for example, reports on the challenges of a single mother working part-time. In the absence of a stable work schedule, organizing childcare and attending classes become extremely difficult.

Addressing the problem, Congress proposed legislation, known as the Schedules That Work Act. The opening paragraphs of the bill summary (H.R.5159 — 113th Congress, 2013-2014) are as follows:

Schedules That Work Act – Grants an employee the right to request that his or her employer change the terms and conditions of employment relating to:

  1. the number of hours the employee is required to work or be on call for work;
  2. the times when the employee is required to work or be on call for work;
  3. the location where the employee is required to work;
  4. the amount of notification the employee receives of work schedule assignments; and
  5. minimizing fluctuations in the number of hours the employee is scheduled to work on a daily, weekly, or monthly basis.

The Schedules That Work Act …

  • Requires the employer, if the request is made, to engage in a timely, good faith interactive process with the employee that includes a discussion of potential schedule changes that would meet his or her needs.
  • Outlines the process for either granting or denying a change.
  • Requires the employer to grant a request, unless there is a bonafide business reason for denying it, if the request is made because of the employee’s serious health condition, his or her responsibilities as a caregiver, or enrollment in a career-related educational or training program, or if a part-time employee requests such a change for a reason related to a second job.

Design Activity

Design Setting

Assume that you are part of a technology team. Your team has extensive expertise in predictive analytics—big data systems, machine learning, and optimization algorithms—with specialized knowledge for labor forecasting and workforce scheduling.

You have been hired to develop a Workforce Management System for scheduling call-center workers. The Workforce Management System has two main modules: (1) A forecasting algorithm; and (2) A scheduling algorithm.

Key features of the design setting include:

  • Workers at the call-center help callers with their questions about a range of products – different workers have different areas and levels of expertise
  • The call center is open from 5 A.M. Pacific to 10 P.M. Pacific
  • Workers at the call-center work out of several different offices located in different regions and time zones of the U.S.

Design Prompt

Your goal is to outline a policy document for regulating the Workforce Management System. You’ll need to address both the forecasting and scheduling algorithms. The policy document should comprise a list of requirements, focused on:

  • Allowed and prohibited data sources
  • Worker and employer constraints that are allowed be considered
  • How conflicts between workers and employers will be handled

Design Process

To develop the requirements, you should follow the following process:

  1. Value scenario. Consider the life circumstances of a call-center worker. Concretely, write a 150-word story, focused on a single worker. The story, called a value scenario, should focus on what it might be like to be a call-center worker who balances call center work and other life responsibilities and aspirations. Consider especially aspects of time and scheduling outside of work. For inspiration you might consider a student who is attending college, a retired, elderly man or woman, or a high school student on summer break. The person of focus could be anyone. You choose. (Optional: For background, and time permitting, read Kantor, 2014).
  2. Direct and indirect stakeholder analysis. Based on your value scenario, and reflections about the call worker and their social context (i.e., their family, friends, and other relationships), identify the direct and indirect stakeholders and briefly discuss their point of views. A direct stakeholder directly interacts with a technology (e.g., entering scheduling preferences, viewing a work schedule, etc.) whereas an indirect stakeholder is impacted by a technology but does not directly interact with it (e.g., a young child might be impacted by a change in his parent’s work schedule).
  3. Policy design. Given your value scenario and direct and indirect stakeholder analysis, propose a set of requirements for the forecasting and scheduling algorithms. The requirements should be written as follows:
    a.         The forecasting algorithm may, or may not, consider data source X1, X2, … Xn.b.        The scheduling algorithm should consider the following worker constraints, in order of importance: X1, X2, … Xn.

    c.         The scheduling algorithm should consider the following employer constrains, in order of importance:  X1, X2, … Xn.

    d.        When conflicts between workers and employer constraints occur, they will be handled as follows: ….

Presentation

Prepare a 5-minute presentation where you:

  1. Read your value scenario
  2. Identify and briefly discuss one indirect stakeholder
  3. Introduce 1-2 requirements, focusing on the policy goals and how the requirements are intended to shape the algorithms.

Reflective writing prompts and exercises

  1. Your design process. Write a 500-word reflective statement on your design process. You might focus, for example, on how your value scenario framed your thinking or how considering indirect stakeholders lead to certain kinds of requirements.
  2. Policy proposal. Based on your group work, write up a policy proposal. In your write-up, (a) summarize your design process; (b) present your value scenario and stakeholder analysis; (c) present your requirements; and (d) present your next steps. In your proposal pay special attention to the public interest.  How, for example, might indirect stakeholders help in an analysis of the public interest?
  3. Policy investigation. Investigate the “Schedules That Work Act” (H.R.5159 – 113th Congress, 2013-14). (a) Read the act and write a summary, focused on stakeholders, value tensions among stakeholders, and technological elements. (b) Investigate the business community’s perspective on this act. (c) Investigate how this act has influenced laws and regulations in States and local communities (e.g., Massachusetts, City of Seattle, etc.)
  4. Technological investigation. Beginning with the patent by Schwartz, & Desai (2017), investigate the state of the art of workforce management systems. Seek to summarize the capabilities of these systems.
  5. Understanding of time.  Consider how your day to day use of technology, in all its forms, influences your understanding of time.  How do you characterize time?  How does technology influence your characterization? Reflecting on the past—How, if at all, has “work time” encroached upon your “personal time” during your life? How has technology, if subtlety, been implicated?  Speculating on the future—What might the future hold for your understanding of time? Now consider how various social economic classes might experience time in similar and different ways.

Notes and further reading

Introduction

  1. Snyder (2016) offers a fascinating account of time and capitalism, and shows how the patterns of work can impact human experience in subtle and not so subtle ways.
  2. O’Neil (2016, chap. 7) discusses the challenge that retail workers experience in their everyday personal lives as they respond to the demands of last minute scheduling.
  3. Schreier (2017) notes that the working hours for video game developers can be extreme, leading to increase risks to physical and psychological wellbeing.
  4. This advertisement is based on an example given by Snyder (2016, p. 209) and modified slightly.
  5. This description of Workforce Management Systems is based on an examination of several patents for such systems, including O’Brien (2003) and Schwartz & Desai, (2017).
  6. The statistics on the change in part-time versus full-time workers employed at major retailers comes from Greenhouse (2012).
  7. See Kantor (2014) for insight into the challenge of daily life when depending on unstable work schedules.

References

DeLauro, R., L. Schedules That Work Act, Pub. L. No. H.R.2942 (2017). Retrieved from https://www.congress.gov/bill/115th-congress/house-bill/2942

Friedman, B., Hendry, D. G., and Borning, A. (2017). A survey of value sensitive design methods. Foundations and Trends in Human-Computer Interaction, 11 (23), 63-125.

Greenhouse, S. (2012, October 27). A Part-Time Life, as Hours Shrink and Shift for American Workers – The New York Times. Retrieved November 2, 2017, from http://www.nytimes.com/2012/10/28/business/a-part-time-life-as-hours-shrink-and-shift-for-american-workers.html

Kantor, J. (2014, August 13). Working Anything but 9 to 5 – The New York Times. Retrieved November 2, 2017, from https://www.nytimes.com/interactive/2014/08/13/us/starbucks-workers-scheduling-hours.html

O’Brien, K. (2003). System and method for online scheduling and shift management. Google Patents. Retrieved from https://www.google.com/patents/US6587831

O’Neil, C. (2016). Weapons of Math Destruction. New York: Broadway Books

Schreier, J. (2017, October 25). Video Games Are Destroying the People Who Make Them – The New York Times. Retrieved November 2, 2017, from https://www.nytimes.com/2017/10/25/opinion/work-culture-video-games-crunch.html

Schwartz, L. S., & Desai, M. D. (2017). Method and apparatus for real time automated intelligent self-scheduling. Google Patents. Retrieved from https://www.google.com/patents/US9679265

Snyder, B. H. (2016). The disrupted workplace: time and the moral order of flexible capitalism. Oxford ; New York: Oxford University Press.

License

Icon for the Creative Commons Attribution-ShareAlike 4.0 International License

Designing Tech Policy by David Hendry is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License, except where otherwise noted.

Share This Book