They Stopped Talking To Each Other

This is the last of the substantive articles in this series. The hundredth will look back across all of them. It seemed right to finish on a decision almost every firm makes, where somebody went and measured what actually happened.

The Verdict, Stated First

Five claims, in descending order of confidence.

One. Face-to-face interaction fell by about 70 percent in both organisations studied, which is stated flatly in the abstract and is the reason this article exists.

Two. It was measured rather than asked about, using wearable devices and communication servers, which places it in a different evidential class from most workplace research.

Three. Electronic interaction rose, so the exchanges moved rather than disappeared, and the authors describe the pattern as social withdrawal.

Four. On our own arithmetic the floor-space saving is not close to covering the loss, at any plausible valuation of the interaction.

Five. And on our own arithmetic the substitution is not neutral, trading one interruption for several and three minutes of elapsed time for something closer to forty.

The first two are the article, ours. A finding this contrary to expectation is only worth anything because of how it was measured, and the instrument is the reason to believe it.

Two of the five come from the abstract and three are ours, ours, and we have kept them separable deliberately. A reader who accepts only the first two still has enough to stop making the collaboration argument.

Our Grades For These Claims

Applying the scheme from the first article in this series.

Grade A for the abstract, obtained verbatim from the journal and confirmed identically against five independent records.

Grade A for the measurement approach, which the abstract describes explicitly and which is the study's principal contribution.

Grade A for our own arithmetic, though every business figure in it is invented and only the 70 percent comes from the paper.

Grade C for generalisation, since this is two corporate headquarters and we did not survey the wider literature.

Grade D for mechanism. The authors offer an interpretation and we did not obtain the analyses that would support or qualify it.

That last grade is the one we would most want a reader to notice, ours. Knowing that interaction fell is enough to change a decision; knowing why it fell would be needed to design a remedy, and only the first is available here.

A Note On Method

Everything here is verified to August 2026.

We obtained the abstract verbatim from the journal's own record[1] and confirmed it identically against a national medical index[2], an open archive[3], an academic index[4], and two further accounts[5][6].

We did not obtain the paper's full text, so we cannot describe the two organisations, the sample sizes, the observation periods, the statistics, or how the badge data were processed.

We therefore have no absolute baseline for how much face-to-face interaction there was before the change, which shapes how we present the cost arithmetic.

We surveyed no other literature on office design, which is a real limitation given how much of it exists.

All arithmetic is ours. Only the 70 percent comes from the paper; the firm, the rents, the wages, the message latencies and every conclusion drawn from them are our own constructions.

We have been stricter than usual about labelling that here, ours, because a business article full of dollar figures reads as though the dollars came from somewhere, and in this case only the percentage did.

This article discusses research on workplace interaction and is not advice on office design, property or employment.

The Claim As Used

What the belief does before we look at what supports it.

The abstract states it: "Organizations' pursuit of increased workplace collaboration has led managers to transform traditional office spaces into 'open', transparency-enhancing architectures with fewer walls, doors and other spatial boundaries."[1]

Four observations, ours.

Note that the abstract identifies collaboration as the stated motive, not cost. That distinction runs through the whole of this article.

The reasoning is intuitive and almost never examined. Remove the barrier and the behaviour it blocked will occur, which sounds obviously right and is a claim about people rather than about architecture.

Stated that way the claim has a hidden premise, ours. It assumes the wall was the reason people were not talking, rather than one of several reasons, and that removing it leaves the others unchanged.

And the abstract's next clause is the gap this study fills: "yet there is scant direct empirical research on how human interaction patterns change as a result of these architectural changes."[1]

Which is remarkable given the sums involved. A decision made at enormous cost, across most of the developed world, on a proposition nobody had measured directly.

Two things that combination usually indicates, ours. The belief was intuitive enough that measuring it felt unnecessary, which is the condition under which large mistakes survive longest.

And the decision had a second benefit nobody needed to argue for, so the stated reason was never load-bearing and was never tested.

The Study

The source.

Bernstein, E. S., and Turban, S. (2018), The impact of the 'open' workspace on human collaboration, Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1753), 20170239, DOI 10.1098/rstb.2017.0239, PMID 29967303[1][2].

The abstract describes the design: "In two intervention-based field studies of corporate headquarters transitioning to more open office spaces, we empirically examined, using digital data from advanced wearable devices and from electronic communication servers, the effect of open office architectures on employees' face-to-face, email and instant messaging (IM) interaction patterns."[1]

Four observations, ours.

Intervention-based is the phrase that matters. These are the same organisations before and after, rather than a comparison between different firms that happen to have different layouts.

That design removes an obvious confound. Firms choosing open plan differ from firms that do not, in industry, age and culture, and a comparison across firms cannot separate the layout from the sort of firm that adopts it.

That confound is the reason most office research says little, ours. The firms with open floors are younger, and younger firms differ in a hundred ways, so a cross-sectional comparison measures the era as much as the architecture.

The keywords listed include interaction, transparency, collaboration, communication, spatial boundaries and collective intelligence[3], which locates the work in organisational behaviour rather than in facilities management.

And we did not obtain the paper, so we cannot say how large the two organisations were, how long observation ran, or what else was changing at the time.

That last gap is the one that would most change our reading, ours. An office move is rarely the only thing happening to a firm, and a reorganisation or a headcount change occurring alongside it would be an alternative explanation we cannot rule out.

Sensors, Not Surveys

The reason to take this seriously, and the direct answer to the previous article's problem. Ours.

The abstract specifies "digital data from advanced wearable devices and from electronic communication servers"[1], and an index records the instrument as a sociometric badge[2].

Four observations.

The ninety-eighth article was about a literature contaminated by single-informant self-reports, where the people being studied classified themselves. This study asks nobody anything.

That matters enormously here, because self-report on this topic would be worthless. An employee asked whether the new office improved collaboration is answering a question about their own adaptation, their manager's expectations and their memory of a year ago.

And the direction of the likely bias is knowable in advance. People who have just been moved into a space they were told would help them collaborate are being asked whether it did, which is close to the worst conditions for an honest answer.

We would note that badges have their own limitations, which we cannot assess. They measure proximity and speech, not conversation quality, and we did not obtain the methods that would explain how the authors handled that.

One point in the instrument's favour holds regardless, ours. Whatever a badge measures, it measures the same way before and after, so a within-organisation comparison is protected from the instrument's biases in a way an absolute figure would not be.

Seventy Percent

The result, quoted directly.

The abstract states: "Contrary to common belief, the volume of face-to-face interaction decreased significantly (approx. 70%) in both cases, with an associated increase in electronic interaction."[1]

Four observations, ours.

Seventy percent is not a small effect. It is not a null result, not a marginal decline, and not a finding that needed a subtle analysis to detect.

"In both cases" is doing important work. Two separate organisations, and the result did not depend on which one.

The word "approx." is the authors being careful, and we read it as a summary of two figures that were close rather than as a precise estimate.

And the phrase "contrary to common belief" appears in the abstract rather than in a press release, which is unusual and tells you the authors knew what they had.

Socially Withdraw

The authors' interpretation, quoted and then examined.

The abstract states: "In short, rather than prompting increasingly vibrant face-to-face collaboration, open architecture appeared to trigger a natural human response to socially withdraw from officemates and interact instead over email and IM."[1]

Four observations, ours.

The word "appeared" marks this as interpretation rather than measurement, and we grade it accordingly. The 70 percent is the finding; the withdrawal is the explanation offered for it.

The explanation is plausible and is not the only one available. Noise, lack of privacy, and the difficulty of holding a confidential conversation would all produce the same measurement.

We did not obtain the analyses, so we cannot say whether the authors distinguished between these.

And for a business decision the distinction matters less than it seems. Whatever the mechanism, the interaction went away, and a firm cannot buy the mechanism separately from the layout.

It matters a great deal for a remedy though, ours. If the cause is noise, acoustic treatment helps; if it is being observed, only enclosure does, and a firm spending money to fix the problem needs to know which.

The First To Measure Both

A claim in the abstract worth pausing on.

It states: "This is the first study to empirically measure both face-to-face and electronic interaction before and after the adoption of open office architecture."[1]

Four observations, ours.

If that is right, and it was published in 2018, then the practice ran for decades before anybody measured both channels across the change.

Measuring both is the necessary part. A study measuring only face-to-face would have found a decline and missed that the exchanges moved, and one measuring only email would have found a rise and called it collaboration.

Which is a general lesson about metrics, ours. A single channel measured in isolation will show whatever the substitution did to it, and the substitution is the thing you wanted to know about.

We did not verify the priority claim, having surveyed no other literature, and we report it as the authors' assertion.

Priority claims are also the kind most likely to be contested, ours, and the kind we would least stake anything on. Nothing in this article depends on it being the first study, only on what it found.

Why Removing Walls Reduces Talking

Our own reasoning, offered as reasoning and not as evidence.

Four observations.

A wall is not only a barrier; it is also a permission. A closed door lets a conversation happen that an open floor does not, because the conversation has somewhere to be.

Most workplace conversations are mildly private. A question about a client, a disagreement about a decision, a personal matter, and none of them wants an audience of eleven.

And an open floor makes every conversation a performance for everyone in earshot, which raises the cost of starting one and lowers the cost of typing instead.

We flag this as our own account and not the paper's. The authors describe social withdrawal; the reason we have given is ours, and neither of us has demonstrated it.

The Floor-Space Trade, Priced

Our own arithmetic on an invented firm. Every figure below is ours.

Twenty staff, at $32 per square foot per year. Private offices at 150 square feet each come to 3,000 square feet and $96,000 a year. Cubicles at 110 come to 2,200 and $70,400. Open plan at 60 comes to 1,200 and $38,400.

So moving from private offices to open plan saves $57,600 a year.

Four observations.

That saving is real, and we want to be clear about it. It is a genuine, recurring, easily calculated reduction in cost, and it is why the decision gets made.

It is also the only figure in the decision anybody computes, in our experience, because it is the only one that appears on a lease.

The figures are invented and the ratio is not unreasonable. Open plan roughly halves the floor area per person against cellular offices, which is the whole commercial logic.

Two costs on the saving side we have left out, ours, and both cut against open plan. The fit-out itself costs money, so the first year's saving is partly consumed by the change.

And a firm that later wants enclosure pays twice, having removed walls it must then rebuild, which makes the decision less reversible than it looks.

And the question this article asks is the one nobody asks. What does the other side of the trade cost?

Which is a question a firm is unusually well placed to answer, ours, because both quantities are in its own records: the rent is on the lease and the wages are on the payroll.

What It Costs

Our own arithmetic, and here we hit a limit we have to be open about.

The abstract reports a percentage decline and no absolute baseline, and we did not obtain the full text. So we cannot say how many hours a day of face-to-face interaction were lost, and we present a range instead of a figure.

Take a loaded cost of $95,000 a year over 1,900 hours, which is $50.00 an hour, and 230 working days.

At a baseline of half an hour of face-to-face interaction per person per day, a 70 percent reduction costs the firm 1,610 hours a year. At one hour: 3,220 hours. At 1.5 hours: 4,830. At two hours: 6,440.

Four observations.

Even the lowest row is a large number. 1,610 hours is roughly one full-time person's year, removed from a firm of twenty.

The baseline is the whole uncertainty and it is ours rather than the paper's. A reader who thinks their staff spend twenty minutes a day talking to each other should use that and will get a smaller figure.

What no choice of baseline changes is the shape. The loss scales linearly with the baseline and the saving does not, so the trade gets worse the more your people talk to each other.

And that is an uncomfortable direction. The firms where interaction matters most are the ones with the most to lose, and they are the ones most likely to have adopted the layout for collaboration.

That is a selection effect running the wrong way, ours. The argument for open plan is strongest where its cost is highest, which is a genuinely unfortunate structure and is worth naming plainly.

The Break-Even

The cleanest way to put it, ours, and it removes the need to value anything.

At a baseline of one hour a day, the firm loses 3,220 hours to save $57,600.

So the trade pays only if an hour of face-to-face interaction is worth less than $17.89, which is 36 percent of loaded hourly cost.

Four observations.

That framing is useful because it does not require anybody to price collaboration. It asks only whether an hour of colleagues talking is worth more or less than about a third of what you pay for their time.

Most owners we would put that to would say considerably more, which decides the question without any further arithmetic.

And that is the useful property of a break-even, ours. It converts a question nobody can answer, what is collaboration worth, into one everybody can, which is whether it is worth more or less than a specific number.

Somebody who says less has made a coherent decision and should proceed, and should also stop describing the change as a collaboration initiative.

There is no criticism in that, ours. A firm on thin margins choosing rent over conversation is making a real trade under a real constraint, and the objection is only to calling it something else.

And the figures are invented throughout, ours, but the structure is not. Any firm can run this with its own rent, wages and headcount in ten minutes, and the answer will be about its own building.

One refinement worth adding when you do, ours. Use the marginal rent rather than the average, since what matters is the space you would actually give up or take on, and a lease with years to run may make the saving notional.

The Substitution Is Not A Wash

The obvious defence, examined. Our own arithmetic on invented figures.

The abstract reports "an associated increase in electronic interaction"[1], so the exchanges moved rather than vanished. A defender would say nothing was lost.

Consider a question resolved at somebody's desk: one exchange, about three minutes elapsed, one interruption to the other person.

The same question over messaging, at an eight-minute average reply wait: two round trips takes 19 minutes elapsed. Three takes 29. Four takes 38. At a fifteen-minute wait, four round trips takes 66 minutes and six takes 99.

Four observations.

Three minutes becomes thirty-eight, on middling assumptions, and the typing time is a small part of it.

The elapsed time matters because the question is unresolved for the whole of it, and whatever depended on the answer is waiting.

And the waiting compounds along a chain, ours. If the answer feeds a second question to a third person, two thirty-eight-minute threads become a day, where two desk conversations would have been six minutes.

Every figure here is invented by us. The paper reports that electronic interaction rose and reports nothing about round trips or latency.

And the direction is not in serious doubt even so. A conversation compresses many exchanges into one continuous one, which is the thing a message thread cannot do.

Two exceptions we would grant, ours. A message thread leaves a record, which a conversation does not, and that is a real benefit for anything that needs to be referred back to.

And it does not require both people to be free at once, which is why it wins for anything genuinely asynchronous. The loss is concentrated in the exchanges that were never asynchronous to begin with.

One Interruption Becomes Four

The second half of the substitution cost, and we think it is the larger one. Ours.

Four observations.

A desk conversation interrupts the other person once. A four-exchange message thread interrupts them four times, spread across an hour.

So the substitution has traded one interruption for four, which is the reverse of what open plan is defended for on focus grounds.

The eighty-fifth article in this series examined the interruption literature and found the picture more complicated than the popular version. We are not claiming a specific recovery cost here, only that four interruptions cost more than one.

And the two costs compound rather than trading off. The elapsed time grows and the interruption count grows, so the substitution is worse on both dimensions at once.

We would resist overstating the interruption half, ours, because the eighty-fifth article found the popular account of interruption costs to be less settled than it is usually presented. Four is more than one, and how much more is not something we can put a number on.

What The Firm Actually Bought

Our reading of the whole episode, and it is the point of the article.

Four observations.

The abstract says the pursuit of increased workplace collaboration is what led managers to make this change[1]. That was the stated goal.

The measured result was face-to-face interaction down about 70 percent.

The actual benefit was floor space, which is real, valuable, and was not the reason given.

So the justification and the benefit are different things, ours. A firm that had argued for the cost saving would have made a defensible case. One that argued for collaboration bought the opposite of what it claimed.

A Correction To Our Own Draft

Reported here because this series reports its own errors in the body rather than quietly.

Four observations.

An earlier draft of this article cross-referenced our examination of the interruption literature as the eighty-first article in this series. It is the eighty-fifth.

We caught it by checking rather than by remembering, which is the only method that works. The article's own header carries its number, and confirming it took ten seconds.

It is a small error and it is exactly the kind this series has spent fifty-four bibliographic variants documenting in other people's work. A number recalled rather than checked is a number that will eventually be wrong.

And we would rather publish the correction than the clean version. An article about measuring instead of assuming should not quietly fix an assumption, which has been the standing rule here since the tenth.

Two things that rule has cost us over ninety-nine articles, ours. Several withdrawn calculations reported as their own sections, each of which would have been easier to delete.

And it has bought one thing worth more than the cost. A reader who finds an error we disclosed learns that we disclose them, which is the only evidence about a publication that is worth anything.

The Word In The Middle Of It

A term in the abstract worth stopping on, because it is doing quiet work. Ours.

The abstract describes the new layouts as "'open', transparency-enhancing architectures"[1], and lists transparency among the paper's keywords[3].

Four observations.

Transparency and collaboration are different things, and the layout delivers the first reliably. Everyone can see everyone.

Which is presented as a benefit and is also continuous observation by colleagues, whether or not anybody intends it that way.

And the authors' own account of what happened is consistent with that reading. People who are being watched withdraw, which is what the abstract describes.

We flag this as our own inference rather than the paper's argument. The abstract uses transparency descriptively, and the connection to withdrawal is ours.

The Loss That Arrives Late

A cost the annual arithmetic misses entirely. Ours, and offered as reasoning.

Four observations.

Most of what junior staff learn, they learn by overhearing. How a partner handles an awkward client call, how a disagreement gets resolved, what the firm actually does when something goes wrong.

An open floor supplies more of that in principle, which is the case for it. But a floor where the difficult conversations have migrated to messaging supplies less, because the conversations worth overhearing are exactly the ones that moved.

And a message thread is invisible to everyone not on it, so the substitution converts ambient learning into private correspondence.

The cost lands years later and is never attributed. A firm whose juniors develop more slowly will not connect that to a floor plan, and we cannot demonstrate the link either; this is reasoning and not evidence.

One thing a firm can do about it regardless, ours, and it costs nothing. Put a junior on the call rather than forwarding them the summary, which restores the ambient learning by a different route and does not depend on any floor plan.

What Actually Survives

Our reading, stated directly.

Five statements.

Face-to-face interaction fell by about 70 percent in both organisations, on the abstract.

It was measured with wearable devices and communication servers, not with surveys, which is why the finding carries weight.

Electronic interaction rose, and the authors describe the pattern as a response to socially withdraw.

On our own arithmetic, the floor-space saving pays only if an hour of interaction is worth under about a third of loaded hourly cost.

And on our own arithmetic the substitution is not neutral, trading three minutes for closer to forty and one interruption for four.

Those five are what we would defend, ours, and the first three come from an abstract confirmed against five independent records, which is the strongest confirmation position this series has reached.

Not An Argument For Private Offices

The obvious misreading, addressed. Ours.

Four observations.

Nothing here says every firm should build walls. The study measured what happened in two organisations that removed them, and reports a change rather than an optimum.

Private offices have real costs beyond rent. They isolate junior staff from overheard expertise, which is how a great deal of professional learning happens, and that is not in the arithmetic.

And the alternatives are not two. Enclosed rooms available to book, quiet zones, and a mix of settings are all in play, and the study speaks to none of them.

What changes is what a firm should expect from removing walls, which was the practical content of the belief and which the measurement contradicts.

And that is the modest, defensible version of this article's claim, ours. Not that open plan is wrong, but that one specific thing it is sold on did not happen, in the two places anybody has measured properly.

Two Headquarters Is Two

The limitation, given its own section. Ours.

Four observations.

Two corporate headquarters is a small number of organisations, however many people they contain, and both were large firms.

A twenty-person accounting practice is not a Fortune headquarters. The result may not transfer, and we cannot say, having not obtained the paper or surveyed the wider literature.

Against that, the finding held in both cases, which is better than one, and the direction was the same.

And the magnitude was similar enough that the abstract reports a single approximate figure for both, ours. Two organisations landing near the same number is more informative than two landing far apart, though it remains two.

And there is a general point about how to read this, ours. The value of the study is that it measured something nobody had measured, not that it settled the question, and treating it as settled would be the error this series keeps documenting.

The Honest Case For Open Plan

Because criticism is cheap unless you can state the other side. Ours.

Four observations.

The cost saving is substantial and certain, at roughly half the floor area per person on our invented figures, and it recurs every year.

Flexibility is real. A firm that grows or shrinks can reconfigure open space in a weekend and cannot reconfigure walls.

And some work genuinely benefits. A team doing one thing together, in short bursts, on a shared problem, is the case open plan was imagined for and may well be the case where it works.

Note what that case has in common, ours. The conversations are about the shared work, so nobody minds them being overheard, which removes the mechanism the paper's authors describe.

The honest version of the argument is therefore available. We are doing this to save money and to stay flexible, and we accept it will reduce spontaneous conversation, which is a defensible position and is not what gets said.

Two reasons it is not what gets said, ours. Cost is an unpopular reason to give staff, and collaboration is a popular one, so the argument that gets made is chosen for its reception.

Which is a poor reason to make an argument and a very common one. The cost of doing so is that the firm then believes its own justification and measures nothing.

The Question Under The Question

Where this now lands, given how offices have changed. Ours.

Four observations.

The study observed organisations transitioning to open plan, and a great many firms have since transitioned to people not being there every day, which changes what the space is for.

If an office is used mainly on the days people come in to see each other, the layout's effect on face-to-face interaction becomes more important rather than less, because interaction is now most of the reason to be there.

And the substitution the paper measured has already happened by other means. A firm whose staff are remote two days a week has moved a large share of its exchanges to messaging regardless of its floor plan.

We flag that none of this is in the paper, ours. It was published in 2018 and studied a world that changed shortly afterwards, and our reasoning about what that implies is reasoning.

What Would Change Our Mind

Stated in advance, because a criticism nothing could answer is not a criticism. Ours.

Four things.

Replication in smaller organisations. Two large headquarters is the study's principal limit, and a similar measurement in firms of twenty or fifty would tell a small business owner far more.

Evidence that the decline reverses over time. People adapt, and a measurement taken shortly after a move may catch a disruption rather than a steady state; we do not know the observation window.

Evidence that the electronic increase compensates. Our arithmetic says it does not, and our arithmetic is invented; a study measuring resolution times either way would settle it.

And evidence that the layout was not the operative change, which a reorganisation occurring alongside the move would supply, and which we cannot rule out from an abstract.

We list these because the alternative is a confident article, ours. An abstract, however well confirmed, leaves four open questions on a finding this consequential, and naming them is cheaper than pretending otherwise.

Your Own Office

The practical application. Ours, and not advice on office design or property.

Four points.

Run the break-even yourself. Your rent per square foot, your headcount, your loaded hourly cost, and the area difference between layouts. Ten minutes.

State the real reason in the proposal. If it is cost, say cost. The finding here is only embarrassing to a firm that claimed collaboration.

And that principle outlives this particular decision, ours. A justification chosen for how it will be received rather than for being true commits the firm to a claim it will never test, which is how most of the beliefs examined in this series survived.

Provide enclosed space that can be booked, if you go open, since the mechanism on any account involves conversations having nowhere to go.

One number to put on that, ours and invented. Two enclosed rooms for twenty staff costs perhaps 300 square feet, or about a sixth of the saving, which is a small price for the only remedy that addresses the mechanism.

And watch what happens to your messaging volume. A sharp rise after a move is consistent with what these authors measured, and you already have that data.

One caution before reading anything into it, ours. Messaging volume rises over time in most organisations regardless, so the comparison worth making is against the trend rather than against the level.

You Can Measure This Yourself

The unusual thing about this topic, ours, and the reason we chose it to finish on.

Four observations.

Almost every finding in this series is out of reach of a small firm. You cannot run a replication of a hormone study or assemble 500 brand histories.

This one you can approximate. Your messaging platform records volume by day, and a move date is a natural before-and-after.

It is not a controlled study and we will not pretend otherwise. Volume rises for many reasons and a move coincides with others, so a change tells you something and not everything.

Two ways to make it slightly better, ours, and both are cheap. Compare the same months in the prior year, which nets out seasonality.

And look at face-to-face meeting counts in calendars alongside it, since a rise in booked meetings against a fall in unbooked conversation is the pattern the paper would predict.

But it is your own building and your own people, which is the standard this series has argued for throughout, and it costs an afternoon.

That standard is the thread running through all ninety-nine of these, ours. A measurement of your own firm beats a published figure about somebody else's, however good the study, because your firm is the one you are deciding about.

Bibliographic Note

The series keeps a count, and this article produced two.

The publication date is given as 19 August 2018 by the journal and a medical index[1][2], and as 2 July 2018 by an open archive and an academic index[3][4].

And that academic index gives the paper as volume 373 with no issue and no article number[4], where the journal's record is 373(1753), 20170239[1].

Three observations, ours.

The date discrepancy is the online-first problem, where a paper appears before its issue, and it is the most benign kind of variant we record. Both dates are correct about different events.

It is nonetheless the reason we have seen this described as a July paper and an August one, and a reader searching by date would be defeated by either.

That brings the running count of bibliographic variants across this series to fifty-four.

Fifty-four across ninety-nine articles is the number the hundredth will have to account for, ours. We did not go looking for a single one of them, and they turned up simply because we checked more than one record for each paper.

What To Do

Do not adopt open plan to increase collaboration. Two organisations measured before and after found face-to-face interaction fell by about 70 percent in both.

Note how it was measured. Wearable devices and communication servers rather than surveys, which is why this finding is worth more than most workplace research.

Run the break-even for your own building. On our own invented figures the saving pays only if an hour of interaction is worth under about a third of loaded hourly cost.

Count the substitution honestly. On our own arithmetic a three-minute conversation becomes a thirty-eight-minute thread, and one interruption becomes four.

Say the real reason out loud. Cost and flexibility are defensible arguments and do not require the collaboration claim.

Give conversations somewhere to go if you go open, since on any account of the mechanism the problem is that they have nowhere.

Expect the effect to be larger where interaction matters more. The loss scales with how much your people talk and the saving does not.

And check your own messaging volume around your last move. It is the one finding in this series a small firm can approximately test on itself.

The Limits Of This Analysis

Several caveats matter. This article discusses research on workplace interaction and is not advice on office design, property or employment. Everything is verified to August 2026. We did not obtain the paper's full text, only its abstract, confirmed verbatim against five independent records; we therefore cannot describe the two organisations, their sizes, the observation periods, how the badge data were processed, what statistical tests were used, or what else was changing during the transitions. We have no absolute baseline for face-to-face interaction, since the abstract reports only a percentage, which is why our cost section presents a range rather than a figure and why the baseline is entirely ours. We surveyed no other literature on office design, a substantial limitation given how large that literature is, and we cannot say whether this study is representative or has been replicated or contested. The finding rests on two corporate headquarters, both large organisations, and we do not know whether it transfers to a small firm. The authors' interpretation, that open architecture triggers a response to socially withdraw, is marked in their own abstract with the word appeared, and we have not upgraded it; competing explanations such as noise and lack of privacy would produce the same measurement and we could not distinguish them. Our own account of why removing walls reduces talking is our reasoning and rests on no evidence we obtained. All arithmetic is ours, and only the 70 percent comes from the paper: the twenty staff, the rent of $32 per square foot, the areas of 150, 110 and 60 square feet, the loaded cost of $95,000 over 1,900 hours, the 230 working days, the three-minute conversation, the eight and fifteen minute reply waits and every round-trip count are invented by us to demonstrate a structure. The break-even of $17.89 is a function of those invented figures and should not be quoted as a general result. And we make no claim about badge measurement quality, having obtained no methods.

Related Reading

Behavioural Finance Ask The Winner Who Was First The self-report problem this study was designed to avoid entirely. Behavioural Finance The Paper Says The Opposite What the interruption literature actually found, which is more complicated than the popular version. Behavioural Finance Fourteen Percent Come Back Someone Else Another workplace belief that survives because nobody measured the thing it claims.

Frequently Asked Questions

Does open plan increase collaboration?

In the two organisations studied, no. The abstract reports that contrary to common belief, the volume of face-to-face interaction decreased significantly, by approximately 70 percent, in both cases, with an associated increase in electronic interaction.

Why is this study more credible than most office research?

It measured behaviour rather than asking about it, using wearable devices and electronic communication servers, and it observed the same organisations before and after their own transitions rather than comparing different firms with different layouts.

Did the conversations just move online?

Electronic interaction rose, so they moved rather than vanished. On our own invented figures that is not neutral: a three-minute desk conversation becomes something closer to a thirty-eight-minute message thread, and one interruption to the other person becomes four.

Is the floor-space saving worth it?

On our own arithmetic for an invented firm of twenty, the saving is about $57,600 a year, and at a baseline of one hour of face-to-face interaction per person per day the firm loses 3,220 hours. The trade pays only if an hour of that interaction is worth less than $17.89, or 36 percent of loaded hourly cost.

Should firms go back to private offices?

Nothing here says that. The study reports a change rather than an optimum, private offices have real costs including isolating junior staff from overheard expertise, and the options are not only two. What changes is what a firm should expect from removing walls.

Does this apply to a small business?

We cannot say. The finding rests on two large corporate headquarters and we did not obtain the paper or survey the wider literature. The value of the study is that it measured something nobody had measured, not that it settled the question.

Can I test this in my own firm?

Approximately, and this is the one finding in this series where that is true. Your messaging platform records volume by day and an office move is a natural before-and-after. It is not a controlled study, since volume rises for many reasons, but it is your own building and your own people.

References

1. Journal record for Bernstein, E. S., & Turban, S. (2018), The impact of the 'open' workspace on human collaboration, Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1753), 20170239, dated 19 August 2018, DOI 10.1098/rstb.2017.0239, reproducing the abstract: that organizations' pursuit of increased workplace collaboration has led managers to transform traditional office spaces into open, transparency-enhancing architectures with fewer walls, doors and other spatial boundaries, yet there is scant direct empirical research on how human interaction patterns change as a result of these architectural changes; that in two intervention-based field studies of corporate headquarters transitioning to more open office spaces the authors empirically examined, using digital data from advanced wearable devices and from electronic communication servers, the effect of open office architectures on employees' face-to-face, email and instant messaging interaction patterns; that contrary to common belief, the volume of face-to-face interaction decreased significantly, by approximately 70 percent, in both cases, with an associated increase in electronic interaction; that in short, rather than prompting increasingly vibrant face-to-face collaboration, open architecture appeared to trigger a natural human response to socially withdraw from officemates and interact instead over email and instant messaging; that this is the first study to empirically measure both face-to-face and electronic interaction before and after the adoption of open office architecture; and that the results inform our understanding of the impact on human behaviour of workspaces that trend towards fewer spatial boundaries. Note: the journal's own record and the source of every quotation in this article. We obtained the abstract only, not the full text. royalsocietypublishing.org
2. National medical index record for the same paper, giving Philos Trans R Soc Lond B Biol Sci, 2018 Aug 19;373(1753):20170239, DOI 10.1098/rstb.2017.0239, PMID 29967303, reproducing the abstract identically, recording the article as part of a theme issue on interdisciplinary approaches for uncovering the impacts of architecture on collective behaviour, listing keywords of collaboration, collective intelligence, communication, interaction, spatial boundaries and transparency, recording the authors' declaration of no competing interests, and identifying the measurement instrument as a sociometric badge. Note: an independent index, used to confirm the abstract verbatim and as our source for the instrument name and the competing interests declaration. pubmed.ncbi.nlm.nih.gov
3. Open archive copy of the same paper, dated 2 July 2018, listing the authors' affiliations at Harvard Business School and Harvard University, listing keywords of interaction, transparency, collaboration, communication, spatial boundaries and collective intelligence, and reproducing the abstract identically. Note: an open archive, used as a second independent confirmation of the abstract and as our source for the affiliations and keywords. Recorded also as one source of the publication date variant. ncbi.nlm.nih.gov
4. Academic index record for the same paper, giving Philosophical Transactions of the Royal Society B: Biological Sciences, 2018, volume 373, published 2 July 2018, and reproducing the abstract identically. Note: an indexing service, used as a third independent confirmation. Recorded also as the source of two bibliographic variants: it gives a volume with no issue or article number, and dates publication to July where the journal dates it to August. semanticscholar.org
5. Publisher's own blog post about the paper, recording that it achieved the publisher's highest attention score to date in a very short time, that the study was conducted by a researcher at Harvard Business School and one at Harvard University, that it used new technology to track behaviour in an office before and after the space was made more open, and that following the change the volume of face-to-face interaction decreased significantly by around 70 percent with an associated increase in electronic interaction. Note: a publisher's blog, NOT a peer-reviewed source, flagged. Used only to establish how widely the paper was discussed and as a fourth confirmation of the headline figure. royalsociety.org
6. Knowledge management magazine article reviewing the paper as part of a series on the most-discussed journal articles of 2018, quoting the abstract's finding and its interpretation that rather than prompting increasingly vibrant face-to-face collaboration, open architecture appeared to trigger a natural human response to socially withdraw from officemates and interact instead over email and instant messaging. Note: a trade magazine, NOT an academic source, flagged. Used as a fifth confirmation of the abstract's wording and to establish the paper's reception. realkm.com

This article discusses research on workplace interaction and is not advice on office design, property or employment. The paper's full text was not obtained, only its abstract, confirmed verbatim against five records. No absolute baseline for face-to-face interaction was available, and no other literature on office design was surveyed. All arithmetic is the authors' own; only the 70 percent figure comes from the paper, and every business figure is invented to demonstrate a structure.