Remote Work Revisited

Some people — not least in engineering — have been advocates of remote work long before the pandemic. Then we all were to experience working from home in full, with most employees liking the idea. Now the world is mostly back to normal. Knowing what we know today, should we change our work model again? Or more precisely: who should, and how?

This is going to be a long one. So I’ll start by sharing conclusions and recommendations, adding the reasons why further down below.

What should companies do about their remote work policies?

Start by reflecting on how your engineers, developers and experts contribute most critically to your business value. Think about how, where, and why that works best. At the same time don’t forget what your workforce wants.

• If you think that you gain most of your value from non-repetitive, challenging, unique, ambiguous, innovative work, then I strongly believe you should benefit from regular in-person collaboration. I’m speaking of the type of work that needs intense collaboration between fluid groups of team members and stakeholders, fast iterations, sometimes a sudden course correction, brainstorming and a lot of ongoing knowledge exchange.
• At the same time, remember it’s always a trade-off, since most of your team will prefer fully remote options. Also, not all organizations will benefit equally from in-person collaboration depending on structure and work styles.
• If the work you value most from your engineers are repetitive routine tasks or can be described as mainly straight-forward, well-specified and delivering based on stable plans with limited collaboration needed, then fully remote work is perfect for you. Embrace it and use it as a tool to recruit and retain top talent.

And what’s the best approach for individual engineers?

Well, research tells us you’re likely to prefer 100% remote work, or something close to that.

If instead you prefer regular days in the office, then just pick such a job where everyone is regularly going. Easy.

And if, in fact, you do want to work fully remotely, that's also fine. But I would challenge you to reflect where this preference of yours really comes from. And whether you are fully aware of the downsides, or might be overlooking something. Then weigh your options. Here's what I would ask:

• What are the real and significant benefit for you, or is it more like just a convenience? I get a feeling that many people overestimate this.
• Are you accounting for long-term risks to your well-being? Commuting several days a week is not only a burden, it's also low-intensity time to prepare and switch. A change in scenery is a brain stimulus. "Forced" routines can help setting the right boundaries and reduce stress. Fewer social interactions may lead to a decline of relationship quality at work.
• Are you really making as much impact when fully remote in the long run? Think of coaching, learning from others, questioning their logic, giving ad-hoc help, reflecting on ideas and adjusting... Over time these things matter much more than burning down backlog items. And your long-term impact will influence your career inside and beyond your current role and company.

So what's the reasoning behind this?

To avoid any confusion, as always, what I’m saying here is about the IT industry in general and about organizations who design and build software in particular. Also, today I will discuss with one particular type of engineering work in mind: the non-repetitive work of solving challenging, unique, and often ambiguous problems in creative and innovative ways usually requiring intense collaboration. Why? All software engineers typically face a percentage of simple “coding monkey” tasks. However, I hold the firm belief that pure coding monkey roles are slowly but surely on their way out of the door. AI will take over most of these tasks. So what matters for the future of work is the other portion of engineering duties, the part that feels a bit more chaotic as described above.

With that being said, let’s take a look at the facts. If the pandemic has shown one thing, then that fully remote work is generally possible. There’s also been a bunch of older and newer research, both proper academic papers and simple business surveys. What research mainly tells us is that people tend to like remote work and that they speak favorably of it in all kind of ways.

But what interests should an engineering organization have at heart when evolving its working models? Companies need to aim for teams that perform reliably and increasingly well over time. And their team’s appreciation of a certain work model is only one of many factors at play.

And why do people even say they like remote work so much? Is this an entirely wise choice for most of them? After all, we’re known to be fairly bad at predicting what really makes us happy [1].

These two key questions — long-term performance impact of remote work, and whether it’s really as good for individuals as we seem to think — have been poorly discussed and considered in both serious research and in the public discussion.

Public discussion, especially here on LinkedIn, is heavily biased as it’s mostly driven by passionate advocates of remote work. Of course there are clear benefits. But the discussion appears incredibly one-sided, and has therefore probably already lead to repetition bias [2] among the rest of us.

At the same time, all serious publications that I have found*) share a set of limitations that render their findings all but useless to answer the question of long-term performance impact.

Firstly and most crucially, unless publications speak about some kind of repetitive, routine work (like [3] does), these published papers all base their findings on performance evidence from asking employees to self-assess. Repeat. A positive impact on performance is being measured by asking people how they feel about their performance. People who mostly love remote work. It’s a well-known fact of life that we unintentionally judge things more favorably if we like them [4, 5]. In newer research, recency bias might play an additional role. How much of a bias do we get from all that? Your guess is as good as mine, but it would be a scientific miracle if there wasn’t any bias at all. And I haven’t found a single paper that effectively controls for these factors.

I don’t blame the authors of these papers for using the only numbers they could realistically acquire. But what that means is that their data points don’t prove much regarding performance. Especially if numbers are fairly close (which is often the case).

Secondly, research that considers the type of work that we’re interested in always has to use proxy metrics. For instance, to predict someone’s ability to perform on complex tasks they might ask how well they can focus in different environments. As an input to how well people can innovate they might ask how well they collaborate in different situations. All this is fine and necessary in the absence of better and more direct metrics. But it’s a classic case of when data and insights are only distant neighbors, and the path from one to the other is full of pitfalls. Hello there, McNamara fallacy [6]! It all becomes a question of how you interpret the numbers in context and what narrative you develop around them.

Thirdly, most publications don’t assume a long-term perspective. Obviously, long-term studies are hard, take time and are expensive. But in their absence, we have to accept that the data we see only observes short-term effects. And we should mostly care about the long-term impact if we adopt a new policy. Making this worse, many of the potential risks around working remotely are unlikely to have much short-term impact but would materialize only in the long run. For instance, imagine junior colleagues learn relatively less efficiently from their peers given less informal collaboration. We would only feel the effects of that gradually over the course of years. At the end of the day the impact would likely be significant, though. Or what if great teams feel increasingly less attachment towards their colleagues as a result of fewer social interactions? This attachment and play factor have often been named as strengthening teams. So a decrease in either should have a detrimental effect over time.

Now, are such potential adverse long-term effects of remote work ever going to materialize? We don’t know for certain, but there is a strong narrative to support the idea. And, so far, available research isn’t telling us anything material that contradicts this logic.

All in all, I haven’t found credible data to support the idea that working remotely full-time is helping the performance of engineering teams conducting non-repetitive, innovative work when interpreted in context. There is no quantitative evidence that it’s going to hurt performance either, but plausible logic suggests it might in the long run.

What’s more, there’s another avenue of research. It’s non-academic, non-public, internal surveys conducted by a slew of different companies and organizations in the industry. When I try to summarize what I’ve learned from numerous conversations with people in different companies,?I see three common themes. In line with public research, engineers like working from home. But more specifically, they agree that some types of work are benefiting from being done remotely (focusing on hard problems), while others generally work better in the office (creative collaboration). And, again like seen in publicly available surveys, they self-assess to work more effectively from home. None of the people I spoke with, however, were able to confirm this assumed better performance from other internal metrics. Instead, one such company has checked against their long-standing metrics for innovation success. From the onset of working remotely in 2020 to mid-2022, far from seeing a “wfh dividend”, they found their rate to remain steady in 2020, and slowly declining since. From another (different) source I’ve heard broader performance went down specifically for junior engineers.

Of course all this is only my summary of other peoples’ summaries, and you have to take my word for it. So take it with a grain of salt. Still, this is most interesting and fits the wider picture surprisingly well.

Also, here’s another bit of anecdotal evidence that I’ve observed. I have also discussed with startups, i.e., places where you would expect the innovation pace to be about the highest. And I mainly heard two types of stories: those who are embracing a fully remote culture, use it for employer branding and cite it as a key reason for their hiring success. And those that tried to re-claim as much of an office based experience as quickly as possible as the pandemic was fading out. The reasons why the first group is claiming they do so well on fully remote work has nothing to do with how well their teams actually perform remotely — it’s indirect benefits. And when I look at who are the most outstandingly successful ones among this (non-representative) group of startups, then three out of three are team (hybrid) office, not team fully remote.

All considered: with the data and experience we have so far, I have decidedly mixed feeling about working fully remotely. Quite likely, most innovative companies would benefit from significant in-person collaboration. For individual engineers, the picture is more nuanced. Additional data will certainly become available over time. But as long as we’re waiting for this to happen, I would say it’s wise to be skeptical of limited data from research, observe closely, consider the long-term impact, and iterate on policies and working models accordingly.

*) If you know of something that I’ve missed, sharing some pointers would be much appreciated!