Navigating the New Normal: Building Cohesive Teams in a Hybrid World — Epoche C1
What this piece reports, and what it does not The problem is building and holding together a working team whose members are in the same room only some of the time, and the four strategies usually offered for it are deliberate informal-communication channels, periodic in-person gatherings, closer attention by managers to individual wellbeing, and clarity of expectations combined with autonomy over method. An earlier version of this piece presented those four as the shared conclusions of three named managers in Berlin, Lyon and London, quoted directly. Those managers are not identifiable and this collection conducts no interviews; their remarks should be read as illustrations of positions widely held in management practice, and nothing in what follows rests on them as evidence. That is not a small correction, because the four strategies are worth assessing and the assessment has to come from somewhere. It happens that each of them touches a question on which good evidence exists — some of it randomised, some of it from natural experiments at very large scale — and the evidence supports two of the four as stated, redirects the third, and identifies a trap in the fourth that the managerial version of it walks into. This piece sets out that evidence. The erosion of informal connection, which was measured The starting claim — that what disappeared was not formal communication but the incidental kind — is correct, and it has been quantified. Yang and colleagues (2022) used the natural experiment of a company-wide shift to remote work: they took anonymised communication and calendar data from more than 61,000 employees of Microsoft in the United States, covering the months before and after the firm went fully remote in 2020, and reconstructed the collaboration network — who worked with whom, how often, through which medium. Two changes stand out. The network became more siloed : the proportion of collaboration occurring within tightly connected groups rose, while the ties that connect otherwise separate groups — bridging ties, the connections along which information reaches people who would not otherwise have received it — became fewer. And it became more static : the rate at which new ties formed and unused ones lapsed slowed, so the network people had in March was substantially the network they still had in September. Alongside this the medium shifted, with less synchronous conversation and more asynchronous written exchange, which is efficient for transmitting decisions and poor for the exploratory talk that creates ties in the first place. This is a sharper finding than the coffee-machine anecdote it vindicates. The deficit is not that colleagues stopped speaking; within existing groups, communication held up. It is that the network stopped rewiring. Everything that depends on rewiring degrades: hearing that another team has already solved your problem, discovering a collaborator, and — the case the earlier version raised about onboarding — joining at all. A new hire's position is the extreme case of the general finding, since a person who joins a static network has no existing ties to fall back on and arrives precisely when tie formation has slowed. That inference is not itself measured in the study, and it is offered as an inference, but it follows directly from what the study did measure. Does cohesion drive performance, and when? Before spending effort on cohesion it is worth asking what the return is, and the question has a meta-analytic answer with important conditions attached. Beal, Cohen, Burke and McLendon (2003) pooled the studies relating group cohesion to group performance and found the association positive but strongly moderated. Two moderators matter here. The relation is considerably stronger when performance is measured as behaviour — what the team actually does — than as outcome , which is contaminated by factors outside the team's control. And it is stronger where the workflow is highly interdependent, so that members must coordinate continuously, than where members work on separable pieces. The practical consequence is that the four strategies are not uniformly worth their cost. A team of specialists working largely in parallel, whose interfaces are occasional and well defined, will get little performance return on cohesion investment, whatever it does for morale. A team whose work is reciprocally interdependent — where each member's next step depends on what another has just done — is where the investment pays. Deciding which kind of team one has is a prior question that the standard advice does not ask. What the randomised trials found about hybrid work itself The premise that the arrangement is a problem to be mitigated deserves testing too, and here there are actual randomised experiments rather than surveys. Bloom, Liang, Roberts and Ying (2015) randomised 249 volunteers among the call-centre staff of a large Chinese travel agency to work from home four days a week for nine months, keeping the rest in the office as controls. Measured performance rose by about 13 per cent, and — this is the part worth having — the increase decomposed. Roughly nine percentage points came from working more minutes per shift: fewer breaks, fewer sick days, and fewer late starts, since the commute had gone. Roughly four points came from a higher rate of calls per working minute, which the authors attribute to a quieter and less interrupted environment. Attrition among the home workers fell by about half. Two results ran the other way: promotion rates conditional on measured performance roughly halved, and when the arrangement was later offered as a choice, more than half of the volunteers chose to return to the office. Bloom, Han and Liang (2024) tested the hybrid arrangement specifically, randomising 1,612 graduate employees at a large firm to two days at home and three in the office, or five in the office, over six months. Performance ratings did not differ. Promotion rates over the fol