On the Effects of Employee Monitoring

The classical workplace, where the history of modern management arguably begins, is the factory floor. Here, employees were monitored by a combination of direct supervision (such as a foreman) and some means of logging attendance and time of entry and exit (such as a punch card).

The advent of the knowledge economy introduced the open-plan office. Unlike the factory floor, the materials (data & information) were now intangible, and the work processes less easy to observe. Consequently, firms began to consider new ways to keep tabs on their employees.

The American Management Association reports during the period from 1997 to 2007, employee monitoring typically took the form of recording telephone calls and voicemails, and monitoring email messages, files, and internet browsing. It’s not clear how much of this data was actually used, as opposed to simply being kept for reference.

The proliferation of wireless internet during the late 2000s, and the constant connectivity it entailed, led to two significant developments.

It enabled more flexible working patterns - both in terms of location (remote working, digital labor, and the gig economy) and timing (working after hours and on the weekends), meaning that direct supervision was no longer as effective.

At the same time, wireless technology also opened up many more possibilities for employers to monitor the speech and movement of knowledge workers thanks to in-built microphones and sensors.

According to a survey of senior executives carried out by Accenture, six out of ten businesses were already “using new technologies and sources of workforce data extensively” in 2019. A study the following year revealed that the majority of companies were tracking employee internet usage, and nearly half tracked “time spent at the keyboard”.

The widespread lockdowns and work-from-home orders that accompanied the COVID-19 crisis propelled many organizations to hastily adjust their management practices to account for a remote employee base. Security firm Top10VPN reported that sales for employee monitoring software leaped by 87% in April 2020.

The pandemic consolidated and accelerated an existing trend that has now gathered substantial momentum, especially as working practices have permanently changed.

The Productivity Software industry as a whole is, therefore, more robust today than it has ever been. Higher demand has also led to greater competition and the need to differentiate.

Hence we see increasingly elaborate applications of artificial intelligence to not only record, but analyze and evaluate employee behavior, including:

• Monitoring chat messages between colleagues for trigger words
• Analyzing tone of voice and writing to assess mood
• Taking photographs of employees at random intervals to ensure attendance
• Tracking eye movements & facial muscles to ascertain concentration levels.

This is on top of AI-driven analysis of email, social media activity, keystrokes, and other in-office activities. The analysis is not only being used in some cases as part of employee evaluations, but also to take autonomous actions such as shutting down applications and blocking websites.

It is clear that not alltracking is bad, and that many targeted uses are justified. This is particularly true where liability to a 3rd party is concerned, and the inability to track or record information may make it impossible to get insurance or give comfort to a potential contractor.

There appears to be a danger when the remit becomes broader, and large amounts of data are collected without a specific hypothesis that is being tested, but with the general aim of, for example, ‘finding out how employees work best’.

When monitoring metastasizes, it becomes surveillance.

Without a tightly defined purpose beyond generally keeping tabs on everyone, it becomes harder to explain to employees why they are being monitored, beyond the fact that it is good for them. The employer becomes like the Orwellian ‘Big Brother’, and in time the monitoring can be perceived as primarily about control.

‍Numerous studies have shown that negative attitudes towards surveillance among employees can lead not only to reduced trust in management, but also anxiety and depression, and eventually absenteeism and higher turnover.

Four motivations typically apply when employers seek to gather data on their employees using technology. In terms of implementation, there is both potential for appropriate usage and a danger of over-reach.

The aim: Collecting certain data connected to adverse events should in theory allow companies to either prevent or at least diagnose the cause of those events and so reduce their risk of occurring.

‍The danger: “Adverse events” can be expanded to mean “anything an employer believes is bad”. For example, monitoring employee chats for statements critical of management could be defined as “mitigating risk of internal misalignment”, but could clearly lead to an increasingly oppressive environment where letting off steam becomes a firable offense.

Appropriate usage: Specific physical risks that no one debates are to be avoided are good candidates for tracking. An example of this is monitoring seat belt usage and driving patterns to ensure road safety.

The aim: In environments where sensitive, non-public information (belonging to 3rd parties, or the firm’s own intellectual property) is handled, firms have a duty to ensure such information is not shared outside the firm.

‍The danger: Once again the issue is one of “creep”. The only way to be absolutely sure no information is being shared by employees is to covertly track all their words - written and spoken - to the maximum possible extent, even in their own homes via their work devices. Without considering the impact on employees and the potential abuse, maximal surveillance runs the risk of violating regulations concerning employee privacy (see here for an EU perspective).

‍Appropriate usage: A case can be made for monitoring specific employee actions such as downloading certain kinds of data and scanning attachments sent externally. Ensuring such actions are justified, minimal, and transparent is key to preventing them from becoming unethical and illegal.

The aim: Clearly, productivity directly affects the bottom line of an organization, and also impacts the customer’s experience. Measuring productivity creates the opportunity to identify areas for improvement and determine if improvement measures are effective.

‍The danger: Living in a world of productivity metrics, management can begin to think of employees as data on a page or machines in a factory, leading to obtuse and oppressive rules (e.g. “If the average number of bathroom breaks decreases by 0.25 we could make $2.5m additional revenues, so let’s set a target to reduce those breaks”). Stifled employees may begin to look for ways to dodge surveillance or, worse still, create the appearance of productive work (machines can sometimes be easier to fool than human managers).

‍Appropriate usage: Analyzing GPS data to optimize routes and come up with policies such as ‘no left turns’ (as UPS did) has little to no effect on the employee experience (except reducing the stress of left turns), and translates to substantial savings in both dollars and minutes.

The aim: Politics, favoritism, and (in the remote age) simple lack of information can lead to unfair evaluations and promotions, and low morale. Analyzing what an employee has actually done (e.g. billable hours) is an objective measure that can be applied consistently across peers.

‍The danger: Creating targets and adding incentives can be powerful, but can also have unintended consequences if a) the metric chosen is a poor one (e.g. number of emails written = level of dedication) and/or b) focus on this metric ends up drawing focus away from equally important behaviors that happen to be more difficult to measure.

‍Appropriate usage: Some KPIs naturally lend themselves to quantification (e.g. sales, leads generated). The trick may be to identify what cannot be measured automatically (i.e. requires human judgment) and rather than trying to find a quantifiable proxy, acknowledge the importance of the variable and give it equal weighting in the computation of an employee’s overall performance grade.

The phenomenon of surveillance creep has culminated in many notable PR disasters for well-known brands.

The banking giant was investigated by a UK regulator after it came to light that it had in 2017 installed a surveillance software (called OccupEye) beneath employee desks that tracked heat and motion to record how long staff were present at their workstations. Barclays defended it as an exercise in ‘assessing office space usage’.

Whatever the justification, it was reported that many employees took the sensors off and disposed of them, implying a general disgust with the practice. A few years later, a whistleblower reported that a new software (Sapience) had been installed that assessed and scolded employees for perceived lapses in concentration levels.

The data gathered by this software was originally anonymised, but this was later changed to allow managers to link the data to individuals - which proved a breaking point for the whistleblower.

The fulfillment centers of Amazon have - rightly or wrongly - become associated with strenuous and joyless working conditions.

The company’s value proposition is highly linked to the speed and efficiency with which it delivers, and hence it is understandable that productivity KPIs are a key area of focus in its management of employees.The company’s employee tracking systems are famous for not only measuring every second when an employee is not working, but also for issuing warnings and terminationswithout any input from supervisors.

‍While this may be efficient, one has to question if it - or the underlying numbers-focused attitude - is beneficial in either the short or long-term for either the workplace atmosphere or the external brand of Amazon.

The boom in videoconferencing that accompanied the beginning of the pandemic led to rapid growth at Zoom, and also put its features under greater scrutiny. In a matter of weeks, the company felt compelled to remove an ‘attendee attention tracker’ that alerted the host when a participant did not have a Zoom meeting screen active for more than 30 seconds while the presenter was sharing.

The official reason for removing the feature was to ‘protect user privacy and security’, but the true reason appears to have been a combination of negative feedback (the feature felt like snooping) and a host of creative ways to bypass the feature available to those who genuinely did want to tune out. This illustrates one of the biggest challenges with surveillance - the internet gets wise (and as we know, angry) fast.

In what is perhaps the most invasive example of employee surveillance, Bank of America was reported as experimenting as far back as 2013 with employee badges that contained a range of sensors, including a microphone to record conversations. The aim was to measure interactions between employees, including in the cafeteria (but not the bathrooms) with a view to - yes, that’s right - improve productivity.

This is part of a broader set of studies that claim to be able to predict when an employee is going to leave a company or score a promotion. While the intention of such devices is typically phrased positively, the coverage in the press has been at best ambivalent (“Big Brother or Big Data?”, asks the Wall Street Journal).

Particularly in a politically-charged environment, the idea that one is being listened to all the time is unlikely to make for a relaxed, creative environment.

As we can see, the danger to avoid in implementing any kind of tracking technology is that the second-order effects (reduced trust, increase anxiety levels) more than cancel out the desired benefits.

The following rules, in our view, will ensure that a tracking scheme does not turn into an act of runaway cultural self-sabotage:

• Broad, imprecise aims are a red flag: schemes that are designed to ‘gather data’ or ‘investigate ways to improve X’ are not only unlikely to generate useful insights, but can metastasize into ever-expanding, police-state surveillance systems.
• Employees should be consulted: any new scheme should be generally acceptable to employees. A consultation, trial, and feedback cycle should precede the rollout of any scheme to ascertain unexpected consequences, and potential flaws.
• Data should belong to the employees: this should ensure that management can only view individual employee data with their explicit consent.

At My Hours, we do not believe in delegating judgment to technology, and we are highly skeptical about the claims of AI-driven employee monitoring solution providers, mainly because it is difficult to trust what you do not understand and cannot control.

Lastly, the pandemic has solidified a trend that has been building since we entered the internet age: namely that the boundaries of personal and work life are increasingly blurred.

We are as like to take a work call at home as we are to perform a personal errand in the office. As humans, we intuitively understand this blurring of boundaries, but machines cannot be expected to do so.