After years of hand wringing and false starts, remote work finally has its moment. Suddenly, massive companies are being run from breakfast nooks and messy bedrooms. Those first Zoom1 meetings were pretty rickety–half the employees didn’t know how to log in, and the other half didn’t think they needed to wear shirts. It sometimes seemed that a huge percentage of the internet’s bandwidth must be carrying nothing but the sounds of crying children and the deafening feedback of people who cannot seem to learn that they can’t have two microphones on at the same time2.
This episode explores the research and realities of a fully remote world. In “Business-as-Usual” we review the limited existing research on remote work and find that many of our assumptions about distractions and productivity don’t hold true. “Measuring Remote Success” explores new metrics that take us from in-office productivity to employee growth. The research on remote work reveals how different personalities require different support; “People Are Different” identifies core remote worker profiles and what they need for success. But it’s not all about individual employees–“Controlling the Beyond-Control” begins exploring the vital role that companies have in supplying what their employees cannot. “Rebuilding Culture” reveals the cultural and organizational factors that predict success for distributed teams, and identifies the technologies native to distributed work. Finally, “Distributed Innovation” and "Diversity-Innovation Paradox" confront two of the biggest mysteries of remote work: innovation and inclusion. Without change, neither are possible, but these sections present a new framework that marries innovation and inclusion to move beyond the lazy limitations that have held us back.
Business as Usual
Outside a handful of quirky mid-sized tech companies, almost no one has run a completely distributed workforce before. It is quickly becoming clear that we have no idea what we are doing. One large-scale Harvard survey found that only 40% of companies felt they were “well prepared” for “flexible work”, much less a business-wide shift to remote work. Even the US Supreme Court has had to adapt to the new remote (it seems that not all of the “attendees” understood how to use mute). Unfortunately, there is very little research on universal remote work.
While we may not know much about what happens when an entire company shifts to working from home, there has been extensive research done in two domains which might help us: distributed computing and distributed cognition. Both deal with the problem of how to do something (search a database or make a judgment) across a collection of distinct nodes (computers or brains). Combining these domains not only helps us understand the flaws of remote work but points to something entirely new: distributed work.
The modern world could not function without distributed computing, but distributed computing is more than just a bunch of PCs operating in different locations. It is an entirely different approach to computing that is adapted to the “fragile narrow laggy asynchronous mismatched untrusted pipes'' connecting those CPUs together. Similarly, distributed work requires much more than a bunch of individual, pajama-clad employees working from their bathroo…bedrooms.
Distributed work requires different behaviors to be successful; using remote work as a patch without any adaptive changes doesn’t automatically translate for every company or employee. For example, the CEO of one fully distributed company, Automattic, observes that remote workers at most companies still expect “hyper-responsiveness” from their coworkers, who must continue to maintain regular hours and other maladapted habits inherited from traditional office work. He claims these bad practices “suppress the ability of knowledge workers to actually think”.
Research supports his claim. Hyper-responsiveness and multitasking across messaging and media tools reduces cognitive control and decreases cognitive performance by 60% in complex, creative work. Employees often work faster to compensate for increased interruptions but experience “more stress, higher frustration, time pressure and effort”.
During the early months of “remote work” average productivity increases, but over time most workers lose those gains and even fall below pre-remote levels. Only a small group remains hyper-productive and engaged. This drop in productivity has been hidden in much existing research as working from home has been “a privilege that you earn” at most companies; therefore, the studied workforce was self-selected to be self-motivated.
Many consultants in this space have argued the benefits of fewer hours wasted in meetings and more time is spent on “getting things done”; however, there is surprisingly little research on remote work supporting those claims. While some studies find employees self-report fewer interruptions, there is no objective confirmation. Do onsite interruptions outweigh the increase in digital and nonwork interruptions? As one researcher noted, “In person…the social cost of asking someone to take on a task is amplified… In a remote workplace, in which co-workers are reduced to abstract e-mail addresses or Slack handles, it’s easier for them to overload each other…” This dissonance in claims about remote work likely reflects the different types of work being observed (e.g., routine vs. creative). Organizations need objective measures of interruptions and their effect rather than guesswork and introspection.
Many business leaders have debated whether the current situation of Covid-19 lockdowns will accelerate the shift to digital and collaborative tools. It’s possible, but this acceleration may only be true for companies already in the process of transition. In this episode, we will explore how everyone can benefit from a transition to distributed work, from diverse individuals to whole companies, from daily routine to the cutting edge of innovation.
Creativity isn’t finding a solution to a given problem. It’s what happens when you stop waiting for someone to give you a problem and start exploring the unknown. Offices never forced us into passive, hierarchical work cultures, but they allowed them to persist even as economic reality has changed. Instead of adapting remote technology to traditional work practices, we need to adapt our practices and technology towards a true distributed work enterprise.
Measuring Remote Success
Writing in the New Yorker, Computer Science professor Cal Newport drew an analogy between the transition from centralized to remote work and the transition of factories from centralized steam engines to distributed electric motors. He argues, “…most companies that have tried to graft it onto their existing setups have found only mixed success,” in large part because mid-level management has been resistant to change. Electric motors demanded entirely new configurations of factories. Networked computing and distributed algorithms forced even more radical departures from the past. Abandoning the centralized workplace will require an even greater transformation, as if you’re launching a competitor, but one already native to distributed work.
One of the principal challenges in distributed computing is that it is fragile: a failure at any point in the network can block the entire system. Computers can only respond to contingencies for which they have been explicitly programmed3, and so every individual piece must be engineered to handle all possible failures. Because distributed computing might have many different components, identifying a point of failure in the moment can be nearly impossible and often requires sophisticated analytics to monitor everything.
By contrast, a point of failure on a colocated team (e.g. a sick day) is easily identifiable, and the entire rest of the team can quickly adapt with minimal preparation. A hierarchical team doesn’t need to be programmed to handle many contingencies because leadership can quickly redirect work. When working remotely, such a failure is much slower to reveal itself. People aren’t visibly absent, and their contributions will not be automatically absorbed by others. Leaders also cannot easily redirect the flow of work as people are simply not as responsive or available. This might suggest a need for deep intrusive work analytics to constantly monitor the productivity of every member of a team, but people are not computers and they do much more than execute an algorithm.
While remote work can be fragile4, we know from distributed cognition that people can adapt in ways computer systems cannot. As people work together to solve problems they develop communication shorthands, novel efficiencies, and even learn to model one another's thought processes. All of this allows the individual nodes of distributed work (i.e. employees) to adapt to hidden failures throughout the network, sometimes without even being aware of it. This power of distributed cognition is a huge advantage over distributed computing, and companies should explicitly train for this adaptability. Don’t treat human beings as dumb points of failure in a fragile system that need intrusive, ubiquitous monitoring. Use work analytics to understand how to increase the robustness and adaptability of your employees.
Traditional office work makes use of productivity metrics and KPIs such as net sales, profit margin, and order fulfillment time. The gig economy, which has invested heavily in remote work, has obscenely expanded these same metrics, logging individual keystrokes and time-on-task to the second. However, we already know that most employees begin to struggle with remote work at some point. Decreases in those numbers aren’t a measure of an employee failure so much as a diagnosis of a process failure. The transition to distributed work requires KPIs that measure adaptability rather than raw productivity, and even better, analytics that provide a map for how to get there. (As we will discuss in the next section, that map will be different for different people.)
Some researchers have begun exploring these new analytics. One finding reveals that “conspicuous monitoring”, transparently monitoring one specific task domain, “improves performance on task dimensions not being directly paid for.” Rather than logging every keystroke and tracking every dimension of work, simply visibly attending to employees promotes broad increases in productivity.
New Metrics for Growth
In our analysis of social tracking technologies for Covid-19, we discussed how analyzing social graphs can allow us to predictively isolate potential superspreaders. My inspiration for this project was my previous research on how work-based social interactions between employees affected productivity. The question is less, “How many times were you interrupted in a day?” and more, “Which interruptions increased productivity and which were harmful?” While few offices may be instrumented to measure such interactions, in distributed work this information is both readily available and crucial. A number of metrics can be starting points.
Remote interaction variables:
- e.g., manager, co-worker, direct report, inter-team
- e.g., meeting, planned, unplanned, one-on-one, group, time
- email, messaging, video
Post-interruption productivity variables:
- Individual level productivity
- Team level productivity
Possible models of the relationship between interruption and productivity:
- Simple correlation between productivity and types of interruption
- More advanced time-varying social graph clustering of factors causing changes in productivity5
Use these new (and old) variables to answer the following questions:
- Does distributed work actually come with more interruptions?
- Electronic interruptions?
- Non-work interruptions?
How does this relate to employees’ perceptions about interruptions?
- Which types of interruptions increase productivity?
- Frequency patterns
- Type of relationship
- Nature of work
- creative vs. routine
- Which types of interruptions decrease productivity?
Beyond raw productivity and interruptions, there’s also the problem of recognition. People that work remotely receive less recognition for their contributions and receive fewer promotions. Those telecommuters that do receive recognition have more face time with their managers. Remote work is a classic case of “out of sight, out of mind”, and we know that promotions have always been a function of proximity. Distributed work analytics must actively balance engagement by managers to prevent promotion based on the availability heuristic.
Companies can’t rely on existing productivity measures to define success in a distributed world. Develop new measures to help discover the paths to employee growth rather than snapshots of raw productivity. Don’t just change what you are measuring, but why you are measuring.
Some of the above analytics may seem complex while others are simple adjustments. In either case, we must shift from prioritizing old metrics and towards metrics that understand employee success in a distributed workforce. Focus on why something isn't working for certain individuals rather than punishing people for decreased productivity. Balance near-term productivity needs against building longer-term human capital capacity and consider long-term aims. Is what you’re doing now bettering your company post-lockdown?
People are Different
In distributed computing, one of the challenges is that the components of the system might be wildly mismatched. Cutting-edge processors interact with outdated hardware, and the latest operating systems receive packets from ancient versions. Distributed computing must be engineered to handle all of the different hardware and protocol versions in its network. Rules must be explicitly established to allow these systems to work together.
In distributed cognition, those rules often naturally evolve as individuals interact, developing into norms, shorthand, and shared culture. In fact, distributed cognition is more than just a metaphor; the brains of high-performing teams of humans literally sync up. In classrooms, for example, students whose brains show less synchrony with other students perform worse, and brain imaging of engineers reveals shared neural patterns in response to engineering-related scenarios. Whether computing or cognition, mismatch can throw a distributed system out of alignment.
Obviously, it doesn't take remote work for people with differing capabilities and work styles to become frustrated working together. Working remotely, however, transforms some of these experiences. On the negative side, remote work can exacerbate the effects of differing cultural norms. On the upside, an unexpected benefit from early surveys in Los Angeles suggest that public employees working remotely engage in fewer downward comparisons. In other words, they spend less time complaining about their co-workers’ laziness.
One of the few core findings we already have concerning remote work is that some people will be successful working remotely and some won’t. Most studies of companies’ remote work practices haven't revealed this because remote work has largely been a small, earned privilege. Now that many companies have gone entirely remote, differing populations will need different support in order to be successful and productive in the context of distributed work. We simply can’t assume that standard onsite methods for supporting employees are still viable.
One challenge for understanding individual differences is the measurement itself. For example, more neurotic individuals (as measured on the Big Five personality inventory) report a preference for working from home but are in fact less successful. It is emotionally stable individuals that are resilient to the challenges of remote work. Similarly, conscientious individuals prefer the idea of working from home, possibly in response to the popular notion that fewer distractions will allow them to get more done. Despite those expectations, people that have actual experience with the unique challenges of remote work generally haven’t preferred it. While actual data on distractions is hard to come by, this preference against remote work by people who have done it suggests hidden obstacles.
So, the research indicates that we need to find synchrony between mismatched coworkers, but it also reveals that we understand much less about remote work than we assumed. If we wish to take full advantage of distributed work, we must start by understanding the mystery of why a minority of employees are robustly successful while most struggle, and develop targeted practices that support each group. This turns out to be a question of balance versus synergy6.
Balancers vs. Synergists
The majority of employees rely on the structure, explicit and implicit, of a traditional work day. Much of how they work flows from the daily rhythms of arriving at an office, participating in regular meetings and interactions, and simply being present in the work environment. In fact, the majority of workers desire a clear distinction between their work life and their personal life with unambiguous boundaries. For them, the biggest challenge is establishing a work-life balance. These “balancers” are that majority group that has struggled with remote work.
For the rest, work is part of their core identity, and stripping away the barriers between home and office allows an even greater synergy between their work and life. This group has thrived working from home (as they often thrive in any work environment). The biggest threat to these “synergists” isn’t a lack of structure but, ironically7, too much.
Previous research has identified this split between balancers and synergists in traditional work environments, exploring topics such as boundary management and micro-transitions. The contrast between synergists and balancers isn’t “good” versus “bad” employees. Traditional workplaces scaffold most employees' self-management, providing support for what is commonly a weakness. Yet even in traditional workplaces, our research finds that a broad array of cognitive, affective, and social skills–we call this Meta-Learning, learning how to learn–are more predictive of quality of work than education level or work skills. Now that the scaffolding of the workplace has suddenly been removed, it is meta-learning, not elite university degrees, that drives those hyper-productive synergists. Ignore intuitions that those with education or seniority will all be synergists, and hide your surprise when some lower-level employees find success in a distributed world.
The following are a set of contexts in which balancers and synergists will experience differing challenges and how to respond.
There isn’t a formal onsite office and that absence is particularly problematic for balancers. A vast research literature has documented how brains naturally link spatial context (e.g., work office, bedroom, kitchen, coffee shop, etc.) to memory, affect, emotional intelligence, and more. In the specific context of work, for example, spatial context influences procrastination; if you are trying to work in a room normally meant for play, the work suffers. And the reverse is true: many employees find their homes become work-associated, causing them to neglect personal and family issues. These issues have minimal impact on synergists who already integrate work and personal life, but for balancers, establishing a distinct work environment in the home is crucial. Ideally, everyone would have a dedicated home office, but that is rarely true. So, create a unique workspace that provides a multimodal signal to your brain that defines “working”. For example, in a studio apartment, set up a workspace in one corner looking into the room, creating a unique view of your room that you only experience while working. (Don’t get lazy and shoot off emails from the futon!)
There is no formalized onsight work schedule anymore, requiring distinct strategies for synergists and balancers. The daily processes of traveling to and from work provided an enormous amount of implicit structure to employees' work lives. For balancers, it gave an unambiguous start and end to the workday. They can largely set aside work responsibilities while away from the office. For synergists, this has always been annoying–all of those emails they send out during the weekend don’t get answered. Balancers are less inclined to blur personal and work and inevitably struggle as structures previously embedded in the office environment are stripped away. They need external boundaries in work for successful boundary management at home. They also need the freedom to establish regular availability that mimics a normal workday without social pressures to deviate from it. Without this additional structure they pull back and invest less in their efforts.
Synergists are more adept at micro-transitions between work and personal and don’t require the temporal and physical structure offered by a traditional work day. Working remotely can be a good thing for them. Some strategies like conspicuous monitoring may be useful for balancers, but are likely to have negative impacts on synergists. Their problem is that they need to be given more control over their own schedule and flexibility in their engagement or they will burn out. Without change, they are simultaneously answering emails on demand during the traditional workday and engaging extensively outside of work hours. They need the autonomy to set their own flexible schedules, just as balancers need an established structure that defines when they are expected to engage.
Rather than demand all employees behave the same, establish norms that respect different needs.
Working from home comes with unique distractions in the form of leisure activities and familial demands. When employees are given more control over how and when they work it inevitably leads to a greater integration of work and life. For balancers this means more interruptions in both directions. For synergists, this may not be a change and the “interruptions” are part of a natural daily flow. Autonomy for them allows dynamic shifts in priority to deal with tasks, work or life, as they occur.
Generally speaking, more interaction between supervisors and employees decreases family and leisure interruptions of work productivity. Much like conspicuous monitoring, increasing interaction with supervisors drives an implicit demand on employees to remain focused on work. This relationship largely disappears when comparing proactive to more passive personalities. Only more passive employees, those waiting for direction from supervisors, benefit from increased interactions. For balancers, frequent interactions are desirable as long as they are during prescribed periods of time. Without strong meta-learning skills, particularly meta-cognition and emotional intelligence, frequent touch points can be essential for remaining productive. As employees become more proactive in creating work opportunities for themselves, they more naturally integrate work and home together.
All of this has particular implications for working parents. Research shows that working from home increases the number of family interruptions of work as well as the number of work interruptions on family. If you’re a balancer, it’s the worst of both worlds; if you’re a synergist, it’s finally the freedom to set your own priorities. And so, just as with distributed work’s other challenges, different types of working parents will need different support.
Of course one major difference for many parents is the different working experience of mothers and fathers8. Research long before Covid-19 has shown that women do substantially more housework and childcare even when controlling for work hours, but the total lockdown has revealed new frictions for both mothers and fathers. For example, remote work creates greater work-to-family conflict for men; in other words, without clear boundaries, most men increase their contribution to family care less than their wives. Women experience a similar phenomenon, but rather than being driven simply by an increase in remote work hours, it is related to aspects of meta-learning, specifically the inability to disengage from work. Balancer moms (or more generally, all balancers with high conscientiousness) need a work culture that actively supports clear boundaries.
(Covid-19 has created a unique remote work experience for families around the world. Not only are parents working from home, but the children are at home as well, taking classes or simply an unscheduled vacation. This has become the principal source of distraction during the pandemic lockdown. It’s not an inherent feature of distributed work that we also need to be full time caregivers for our children. Even for synergists, multiple 80-hour-a-week jobs at the same time has been too much.)
Requests for information are a consistent source of distraction for all types of employees. “Do you have the latest numbers…?” “What is the status on…?” “Where can I find the…?” Whether you are a balancer or a synergist, organizations with poor access to information suffer from more interruptions. One of the easiest ways to improve productivity for both groups is to ensure that people don’t need to ask for information trapped in another person’s head. Any organization engaged in distributed work must make all information readily available through collaborative knowledge repositories online. (We’ll discuss asynchronous collaboration in “Rebuilding Culture” and “Distributed Innovation”.)
Complex creative work is suffering in a remote world. Chinese search giant Baidu conducted one of the first analyses of software developer productivity during the pandemic lockdown. They found that productivity increased for simple, modular tasks that could be done by a single developer. Other research has also shown that employees doing complex but non-collaborative tasks were more productive when remote. These findings support the intuition expressed by many that remote work means fewer distractions and greater productivity. But these intuitions are wrong where it counts the most. Baidu’s report shows gains only for routine or solo work–as task scale, complexity, and innovation increased, productivity decreased dramatically.
This early research has strong implications for different types of workers. Balancers will shift towards oversimplified work, steering away from the messiness and blurry boundaries of complex projects. Contrary to intuition, this shift will be greater in more conscientious employees as traditional productivity measures incentivize employees to shift away from complex, collaborative tasks that suffer in remote. For managers, these shifts in balancer productivity will be exacerbated by issues of time, space, and distraction identified above.
Synergists are less likely to shift their work from more complex tasks, as they tend to be more goal-oriented rather than process-oriented. Instead, they will work faster in order to compensate for the inefficiencies of remote work. This might sound great in the short term, but if they feel under constant pressure to deliver and accelerate their work in response, research has shown that creativity decreases, while error rate and chance of burnout increase. While managers should give synergists increased autonomy, this doesn’t mean just throwing problems at them as though they are a boundless source of productivity. More generally, distributed work needs new tools to facilitate complex creative work, particularly asynchronous and semi-asynchronous technologies. We will return to this last point from a company-wide perspective in “Rebuilding Culture”.
For the individual employee, we already know what works in the long-term: meta-learning. Importantly, individual meta-learning “skills” are changeable, and organizations can invest in programs that foster meta-learning throughout their workforce. Over the long run this will reduce mismatch in distributed work, just as we might address hardware and software mismatch in distributed computing by investing in infrastructure. Actively lifting meta-learning skills and increasing the number of synergists has the wonderful side benefit of increasing job, life, and family satisfaction.
Unfortunately, shifting a workforce towards synergists is an effortful, multi-year process, and it won’t include everyone even in the best circumstances. In the short term, companies must deploy differentiated support policies for different types of workers: autonomy for some, conspicuous monitoring for others; explicit boundaries for some, self-regulation for others; and, project complexities that optimally match individual and team capacity. People are different; successful distributed work recognizes this simple truth.
Most of the existing research on remote work has assumed that all employees respond the same to remote work and that one set of policies can support everyone. These assumptions emerged because pre-Covid-19 researchers studied largely homogenous populations, either those that earned the right to work remotely or low-autonomy gig workers. The differences we describe above don’t become apparent until everyone is forced out of the office.
So, be honest with yourself and others about what you need. Managers should learn to recognize the clues that something is going wrong, but with self-reflection, you can identify them in yourself as well. Different types of fragility require different types of intervention. For example, the following set of questions from “6 Tips For Managing Remote Employees” can be a helpful framework if you reject the idea that there is one answer for every employee.
“What are the normal working hours for the team?”
Individuals in different time zones can quickly feel isolated and frozen out of distributed work teams if work hours are defined by the convenience of headquarters. Further, defined work hours are really only beneficial to balancers. When answering this question, be flexible.
“How long will it take to get back to each other?"
Predefined expectations can be incredibly helpful for balancers because it provides unambiguous structure to their workday. For synergists, having a set expectation is an attack on their autonomy that undermines their ability to set their own priorities. Everyone needs to be explicit with their answer to this question and also comfortable accepting everyone else’s answer.
“How will we notify each other when we will be unavailable and unable to meet these expectations (e.g., out at a doctor’s appointment)?”
This is just another distracting request for information and it has the same solution: asynchronous tools.
“Establish a Video-First Culture.”
As the author notes, video is powerful because of its ability to convey non-verbal cues. Unfortunately, existing technology is a rather poor medium for non-verbal communication. At Socos Labs, we have been exploring the possibility of artificial intelligence to enhance non-verbal cues for both individuals and teams, such as conveying the laughter or the boredom of an audience when you can’t see every face. Video can only be one part of a solution, however, as it is a real-time (synchronous) tool and research from distributed cognition and distance learning shows that asynchronous tools are essential.