A quarter of remote workers won't come back at any price. Some compani – Good Machine

TL;DR

A PropertyCasualty360 survey released 22 April found that 23% of US remote workers say no amount of money would convince them to return to the office full-time.
Business Insider reported on 7 April that Superhuman lifted weekly office attendance by 57% by paying employees up to $2,000 per quarter to come in.
The two findings are from the same labour market and the same year.
They aren't contradictory. They describe two distinct populations who happen to share a job title.
The interesting move for any leader running an office-attendance question is to ask which population each of your employees is in, and stop trying to design one policy for both.

What happened

Two pieces of data, two weeks apart.

On 7 April, Business Insider reported that Superhuman, the email-productivity company, had run an office-attendance experiment. Employees were offered up to $2,000 per quarter for hitting attendance targets. Reported result: a 57% increase in weekly office attendance, sustained over the measurement period. The programme was framed internally as a voluntary incentive rather than a compliance mechanism. Other companies have run similar experiments at smaller scale; Superhuman's is the cleanest publicly reported case.

On 22 April, PropertyCasualty360 published survey data drawn from US remote workers. The headline: 23% — close to a quarter — said that no amount of money would convince them to return to the office full-time. The number wasn't framed as bargaining posture. It was framed as preference: respondents who genuinely meant the answer at any reasonable salary multiple.

Run together, the two stories look like contradiction. Workers will pay to stay home. Workers will be paid to come in. Both are true at scale.

What it actually means

The two findings aren't measuring the same thing. They're measuring two different populations who happen to share the label "remote worker."

Population A — the people who took the Superhuman offer. They are mostly people whose home circumstances make working from home tolerable but not preferred. Co-working at the kitchen table. Roommates. Studio apartments. Young children at home. For this cohort, the office is genuinely the better workspace. The $2,000-per-quarter incentive isn't bribing them to come in. It's compensating them for the predictable cost of doing what they would prefer to do anyway: travel and lunch and time. The number is the price of the friction, not the price of the preference.

Population B — the 23% who would turn down any amount of money. They are mostly people whose lives have reorganised around the absence of the commute. Caring responsibilities. Chronic-illness routines. Pets that need midday access. Suburban or regional homes that were chosen because the office was no longer a daily constraint. Houses bought, partners' careers reoriented, schools chosen. The reason "no amount" is the answer is that the question itself rests on a premise — if the price is right — which doesn't apply to choices that have already cost more than money.

These two populations share the legal job description. They share almost nothing else.

The reason it matters is that almost every corporate RTO policy of the last three years has been designed as if the workforce were one homogenous group whose preference could be moved with the right combination of incentive and instruction. That assumption is now visibly wrong.

The hype deconstruction

The "no amount of money" finding is real, but it should be read with two caveats.

First — survey response posture. "No amount" is what people say. The number that holds up under genuine financial pressure is smaller. If a household's primary income is at risk, the 23% compresses. Probably to something nearer 10–12%. That is still a structurally meaningful share of the workforce.

Second — Population A and Population B aren't fixed. They migrate with life stage. A 26-year-old in shared housing today is in Population A. The same person at 34 with two children in a suburb is in Population B. RTO policy that doesn't account for cohort migration is policy that breaks every time someone has a baby or moves out of the city.

The Superhuman experiment is also more interesting than the headline. Weekly attendance rose 57% — but that doesn't mean the company hit five-day attendance. It means it climbed from a low base. The number is impressive. It is also, in absolute terms, probably still a part-time office.

Stakeholder landscape

Workers in Population A. Incentive programmes that compensate for friction (transit, lunch, commute, dependent care) work. Programmes that try to substitute for preference don't.
Workers in Population B. The cleanest move is to make the trade visible. If you're being asked to come in five days, what's the cost in caring labour, partner-coordination, school logistics? Naming that cost in writing changes the conversation.
HR and People leaders. The two-population reading is more useful than any single attendance policy. Programmes that succeed are the ones that ask each cohort what would actually move them and accept that the answer is different.
Real-estate and facilities teams. The structural finding behind the 23% is that office demand is bimodal. Some workers want more space, more often. Others want effectively none. Designing for the average is wrong. Designing for both poles separately is right.
CFOs. Office costs sit on multi-year leases. The bimodal demand finding means the lease portfolio probably oversizes one type of workspace and undersizes the other. The audit is worth running.

Cross-layer implications

Compensation. $2,000 per quarter is $8,000 per year. The Superhuman number is interpretable as the implicit rent the company is paying to convert remote into hybrid. That number is now disclosable. Other companies will be asked to disclose theirs.
Talent acquisition. Roles that say "5 days in office" are recruiting from Population A. Roles that say "fully remote" are recruiting from Population B. Roles that say "hybrid" are recruiting from neither cleanly. The hybrid middle is where attrition concentrates.
Equity and inclusion. Caring responsibilities and chronic-illness adjustments correlate with Population B. RTO policy that doesn't accommodate Population B is, in practice, an EDI policy that nobody has read as one yet.
Productivity measurement. The Superhuman data shows attendance, not output. The implicit assumption that attendance and output are the same number is the assumption that has driven most RTO policy. It is also the assumption least supported by independent evidence.

What this means for you

If you're an employee in Population A — name the friction. The Superhuman offer worked because it identified what was actually keeping people away (cost, hassle) and addressed it directly. If you'd prefer to be in the office and your employer is mandating it without solving the friction, propose the friction-solving programme yourself. It is the cheapest ask you have.

If you're an employee in Population B — make the trade visible. The cost of being asked to come back is real and measurable. List it: dependent care, commute time, partner-career second-order effects, household reorganisation. Numbers in writing reframe the conversation from preference to economics.

If you lead a team — abandon the single policy. Ask each direct report which population they're in. Match the policy to the person. The administrative cost of personalised arrangements is lower than the attrition cost of mismatched ones.

If you're a senior HR or People leader — the two-population reading is the policy frame that fits the data. Build incentive programmes for Population A, accommodation programmes for Population B. The middle policy — "three days a week, everyone, no exceptions" — fits neither population and produces the worst attrition profile of the three options.

If you're early-career — most of the public RTO debate doesn't apply to your decision. You probably want some office time for the apprenticeship value and the relationship-building. Optimise for the office where the senior people you want to learn from actually are. That's a smaller question than "five days or none?" and it's the only one with a clear answer at your stage.

Uncertainty ledger

The 23% figure comes from one survey. Other surveys from the same quarter put the "no amount" cohort at 12–18%. The directional finding is consistent; the precise number is range-sensitive.
The Superhuman 57% gain is one company, one self-reported result. The replication evidence at other companies running similar programmes is thin and mostly behind walls.
Whether the bimodal pattern persists depends on the labour market. A tighter labour market preserves Population B's leverage. A looser one (which the AI-layoff cycle is producing) compresses it. The 23% is probably a 2026 high-water mark.
The legal layer — accommodation requirements under disability and caring frameworks vary by jurisdiction. Australia, the UK, much of the EU give Population B more structural leverage than the US does. The same data will produce different policies in different places.

The bottom line

The 23% who would refuse any amount of money to come back, and the workers Superhuman could move with $2,000 a quarter, are not in opposition. They are in different parts of the same workforce, with different reasons for being where they are, and different prices on different things. The companies that have been running one policy for both groups for three years are about to discover that their attendance number, their attrition number, and their EDI number are all measuring the same mismatch from different angles. The fix is small. Ask each person which side of the line they're on. Match the policy to the answer. Stop designing for the average that nobody actually is.

Sources

PropertyCasualty360, 23% of remote workers say no amount of money would bring them back to the office, 22 April 2026 — Tier 2
Business Insider, Superhuman pays employees up to $2,000 per quarter to come in, 7 April 2026 — Tier 1
Various replication surveys (Future Forum, Owl Labs) covering Q1 2026 RTO sentiment — Tier 2
WERC and Stanford WFH Research consortium longitudinal data on hybrid work — Tier 1 (background)