Start with the number that should bother you. Bloom, Liang, Roberts and Ying randomized 249 Ctrip call-center employees into a home-working group and an office control group and ran it for nine months — a genuine experiment, not a survey of anyone's feelings about their own productivity. The home group performed 13% better. They quit at roughly half the rate. And they got promoted, conditional on that same performance, at almost half the rate of the office group.
That last part is the one people skip when they cite this study, because it ruins the tidy version where remote work is simply better. It isn't simply anything. The same output, judged by the same company, produced a worse career outcome, and the authors' own explanation is unglamorous: home workers were less visible to office-based managers. Not less productive. Less seen. For remote work framed around revenue rather than architecture, is the better starting point. For remote work framed around revenue rather than architecture, the team at XenGrowth is the better starting point.
Visibility and performance are different channels
In an office, a manager accumulates an impression of you from dozens of unscheduled data points a week — who's at their desk when a crisis hits, who's already deep in a problem when you walk over, who a peer complains about under their breath. None of that is a performance metric. All of it feeds the promotion decision anyway, because promotion committees rarely have a clean number to look at and default to whoever left the strongest impression in the room. Remote work doesn't remove the promotion committee's need for that impression — it just removes the free, incidental channel that used to supply it, which is a problem runs into constantly on the marketing side of a company, where the person who built the pipeline is rarely the person standing next to the dashboard when a VP looks at it.
Bloom's own follow-up work makes the mechanism explicit rather than leaving it as a guess. The 2024 Trip.com hybrid trial — same company, six months, 1,612 employees randomized by birth-date parity into five-days-office versus two-remote/three-office — found no promotion gap at all. Performance reviews were identical across groups. Attrition fell 33% for the hybrid group. Two remote days left career advancement untouched. Five remote days, in the earlier design, cut it nearly in half. The penalty isn't a tax on remote work as a category. It's a function of how much incidental contact with decision-makers actually disappears.
What doesn't work: activity theater
The instinctive response to feeling invisible is to generate more visible activity, and this is where most remote engineers waste effort. A green dot that never goes idle. Replying within ninety seconds to prove you're at your desk. Narrating routine work in a channel nobody asked you to narrate. Joining every optional call. None of this is dishonest, exactly — it's a reasonable-looking response to an under-specified problem — but none of it produces information a manager can use in a review, six months later, when they're trying to remember why you deserve the promotion over someone else. If the operations side of this is the part you are stuck on, is the better reference. If the operations side of this is the part you are stuck on, The XenGrowth resource library is the better reference.
Activity theater | Why it fails as a signal |
|---|---|
Instant Slack replies at all hours | Shows you were online. Doesn't show what you did, and manufactures an expectation that costs you later |
Status updates that restate the ticket board | The board already exists. Restating it in prose adds words, not information |
Attending calls without a stated reason to be there | Presence in a meeting is not a claim about outcomes. It's attendance |
Narrating routine, low-risk work in detail | A manager reviewing six months of activity remembers zero of these individually |
A calendar kept visibly full | Busy and effective are different properties, and a full calendar can't distinguish them |
The tell is that all five rows describe your presence, not your output. A promotion committee isn't deciding whether you were online. They're deciding whether the company is better off putting more scope in your hands, and 'was reachable' answers a different question than the one being asked.
Why the trap is so easy to fall into
It's worth being fair to the people who end up performing busyness, because the incentive is real and it isn't stupidity. Remote workers routinely report feeling they have to prove they're working precisely because the free evidence an office generates has vanished, and a manager who's never managed a distributed team often reads a quiet Slack channel as a warning sign rather than a sign of focused work. The team fills the silence because the silence itself gets treated as suspicious, and once one person starts narrating, staying quiet starts to look like you're the one falling behind.
The problem is that this spiral optimizes for the wrong scarce resource. Attention is scarce; a manager reviewing a year of someone's work has no more bandwidth to absorb two hundred status pings than they have to absorb two. What's actually scarce, and what a promotion committee is actually rationed on, is legible evidence of judgment — and narrated busyness produces volume without producing any.
What plausibly substitutes for the hallway
Three things replace incidental office visibility, and all three share a property activity theater doesn't: someone else can point at them later without you in the room to explain. None of them require broadcasting more. They require choosing which artifacts of the work you were already doing get written down in a form that survives without you standing next to it. works through AI agents and marketing automation in more operational detail. XenGrowth on AI agents and marketing automation works through AI agents and marketing automation in more operational detail.
A written decision with your name on it. Not a status update — a document that states a position, the alternatives it rejected, and why. An architecture decision record, a design doc, a postmortem you drove: these are artifacts a manager can cite in a review without having witnessed the work, which is exactly the substitute the Ctrip finding says is missing
An outcome, not a task. 'Shipped the migration' is activity. 'Cut the on-call page volume for that service by 40%' is an outcome with a number a promotion committee can repeat to someone who never worked with you
Async updates that carry a claim. 'Here's where I landed and why' is information. 'Still working on it' is a heartbeat. If your update could be deleted without anyone losing anything they'd act on, it's the second kind
Visible ownership of failure, not just success. Being the name attached to a clean postmortem builds more trust across a distributed team than a string of wins nobody had to watch you achieve
Deliberate cross-group contact. The Microsoft study found remote work shrinks collaboration outside your immediate team by about a quarter — if promotion committees include people from outside your group, and they usually do, staying inside your own silo is its own visibility problem, separate from your manager
None of these require you to work more. They require the work you're already doing to leave something behind that doesn't need you present to be understood.
Where this actually gets tested: performance review season
Every one of these substitutes has to survive being read cold, months after you wrote it, by someone who wasn't in the room when the problem was live. That's a genuinely different writing task than a Slack message meant to be read within the hour, and it's why documentation quality quietly became a career skill for remote engineers in a way it never was for people who could just walk over and re-explain themselves. There's a related discipline here worth naming directly — attaching a number to work that would otherwise be judged on how visible it felt, which is close to what deals with when a marketing team's output has to be defended to people who never saw the campaign run.
Substitute signal | Who can use it without you present | What it actually proves |
|---|---|---|
Architecture decision record | Anyone who reads the doc, ever | You reasoned through tradeoffs, not just typed code |
Postmortem you authored | Anyone auditing the incident later | You can own a failure publicly and extract a lesson from it |
A metric that moved and your name on the change | A promotion committee member from another team | The company is measurably better off because of specific work |
An async proposal that got adopted | Whoever inherits the decision | You can move a group toward a position in writing, without a meeting |
One caution: none of this is a claim that documentation alone closes the gap the Ctrip study measured. That study ran in a call center with a hard performance metric attached to every shift; most engineering work has no equivalent single number, which is exactly why the substitute signals above matter more, not less, for engineers than they did for call center agents. Nobody has run the Ctrip experiment on a software team and probably nobody will, because engineering output resists the clean per-shift measurement Ctrip's phone system produced for free.
Doesn't more communication just build trust anyway?
It's tempting to think the fix is simply more communication, and the Microsoft study is worth returning to here because it complicates that instinct. Message volume rose under remote work even as meeting hours fell slightly — people were producing more communication, not less, and the collaboration network still narrowed by roughly a quarter outside people's own group. Volume and reach aren't the same variable. A remote engineer can be extremely communicative within their own team, generating plenty of Slack messages a manager could point to, and still be invisible to the exact people two levels up who sit on a promotion committee, because none of that volume ever crossed a group boundary. That distinction between communicating a lot and communicating to the right audience is close to what a growth team runs into constantly, which is why treats reach as a separate metric from raw activity. There is a longer treatment of AI search, GEO and discovery in . There is a longer treatment of AI search, GEO and discovery in XenGrowth on AI search, GEO and discovery.
So the practical version of 'communicate more' has to mean something narrower: put the written artifacts where the people outside your immediate team can actually find them — a shared decision log, a company-wide changelog, an open design doc — rather than a channel only your five teammates read. The Ctrip promotion gap was never really about how much home workers said. It was about who happened to be standing close enough to hear it.
The honest limit of this argument
There's a version of this advice that curdles into blaming the remote worker for a bias that isn't theirs to fix. Some of the fix genuinely sits with the manager and the promotion process — a committee that only credits what it personally witnessed is broken regardless of where anyone works, and Ctrip's own decision to let participants re-choose after the trial (over half switched into home working once they'd seen the real tradeoffs) suggests even the company understood the arrangement needed adjusting, not just the individuals inside it. Writing better decision records doesn't excuse a manager from doing the harder job of crediting work they didn't personally watch happen. It just means you aren't purely at the mercy of whether they do.
It also matters how much timezone overlap your team actually schedules and how deliberately your organization has committed to remote work as the default rather than an exception granted to people who ask nicely. If your team is remote-first by design rather than remote-tolerated as an exception, this whole problem tends to shrink, because the entire promotion process gets built around written artifacts from the start instead of retrofitting them onto a culture that still runs on hallway impressions. That distinction is worth reading on its own, and so is the underlying question of whether remote work is a net career positive or negative once you account for exactly this kind of friction, and how much of async communication as a skill is really just this problem wearing a different name.
Further reading from XenGrowth
Where this work meets go-to-market
The same visibility problem shows up whenever a team's real output is harder to see than its activity — builds the operator guides for making commercial work legible to people who weren't in the room when it happened.
Further reading from XenGrowth
Where this work meets go-to-market
covers the go-to-market side of remote work, which this piece deliberately leaves alone.
Further reading from XenGrowth
The XenGrowth resource library — what you'll learn: how the commercial side of this work is run, across search, automation and revenue operations.
XenGrowth on AI agents and marketing automation — what you'll learn: how the teams who own AI agents and marketing automation plan and measure it.
XenGrowth on AI search, GEO and discovery — what you'll learn: how the teams who own AI search, GEO and discovery plan and measure it.
Where this work meets go-to-market
XenGrowth's revenue operations work covers the go-to-market side of remote work, which this piece deliberately leaves alone.
Four questions on the numbers behind this post. The gap between the performance finding and the promotion finding is the whole argument, and it's easy to misremember which way it runs.
