When Does a City Deserve an Office? What 883 Company-City Markets Reveal

"We have people in Denver now, should we get space there?" is one of the most expensive questions in workplace operations, because it is almost always answered with judgment instead of evidence. Someone counts heads, someone else remembers a request from a manager, and a decision worth six figures over three years gets made on a feeling about momentum.

We can put numbers on it. Croissant has booked more than 4 million workspace hours since 2015, and within that sits a specific dataset: every occasion a company started using workspace in a market it had not used before. There are 949 of those company-market relationships, and 883 of them are old enough to watch for a full twelve months. We tracked all 883 from their first booking to their twelfth month.

The finding is blunt. Most new-city workspace demand never becomes anything.

What happens to a new market in its first year

The median new market generated four bookings in its entire first year, all inside a single month. Just over half never produced a second month of activity at all. And 53.6% never involved more than one person: one employee, in one city, booking a handful of days.

Read the curve carefully, though, because it has a shape worth noticing. Attrition is brutal early and then almost stops: 35.1% survive to months 2–3, 28.5% to months 4–6, and 27.9% are still active in months 7–12. Between month three and month twelve, the population barely shrinks.

Three signals that separate real markets from noise

If nine out of ten new markets will not reach weekly demand, the useful question is whether the ones that will are identifiable early. They are, and the strongest signal is not the one most companies watch.

Signal 1: how many people, not how many bookings

Volume in the first month does predict survival, and it predicts it steeply:

But headcount in the first month predicts it more sharply, and it is far easier to observe:

The jump from one person to two is the single largest discontinuity in the dataset. A market that starts with one employee reaches sustained demand 12.0% of the time. A market that starts with two reaches it 39.0% of the time, more than three times as likely, on the strength of one additional person.

And 80.4% of all new markets start with exactly one person. Four out of five new-city workspace requests come from an individual, and individuals are a weak predictor of durable local demand. That is not a criticism of the request; it is a reason not to answer it with real estate.

Signal 2: someone books a meeting room

This is the finding we did not expect, and it is the most operationally useful one in the study.

A market where someone books a meeting room is more than three times as likely to become durable. The reason is intuitive once you see it: a desk booking means one person needed somewhere to work, while a meeting room booking means several people needed to be in the same place at the same time, and that only happens where there is a team, a customer, or a recurring reason to gather.

The catch is rarity. Only 2.7% of new markets ever produce a meeting room booking. So it is a high-precision, low-recall signal: when you see it, take it seriously; you will not see it often.

Signal 3: the market survives its first quarter

Given how flat the curve becomes after month three, survival through the first quarter is close to a proxy for survival through the year. Of markets still active in months 4–6, most are still active in months 7–12. The first quarter does the filtering; everything after it mostly confirms.

This is why committing space in month one is a bet against your own data. You are paying to remove the only cheap filter available to you.

A decision framework you can actually apply

Combining the three signals into thresholds, in the order you will encounter them:

Every row above the last one is a row where a lease loses. That is not an argument against offices. It is an argument for putting the office decision after the evidence rather than before it.

What this means for how you budget

Three consequences follow from a 9.1% activation rate.

Your default answer to a new city should be pre-decided. If nine out of ten markets will not mature, the answer to "can we get space in Denver?" should be a standing policy (on-demand access, available immediately, no approval chain), not a project. Companies that treat each request as a decision spend more time deciding than the decision is worth.

Budget for the tail as an operating expense, not a capital commitment. The long tail of markets is real: in our wider dataset, 37.5% of all company bookings happen outside the company's largest market. It is genuine demand, but it is thin and mobile, and it should be funded the way variable demand is funded.

Instrument the first quarter. The signals that matter, namely how many distinct people booked, whether anyone booked a room, and whether month four saw activity, are all observable in the first ninety days, and all invisible if workspace is bought through expense reports. A market you cannot measure is a market you will over-commit to.

Methodology and limits

Croissant has booked more than 4 million workspace hours since 2015: 611,226 booked sessions, each counted as a standard seven-hour workday. Logged attendance averaged 3.8 hours per session, so booked hours describe reserved capacity rather than time physically present.

This study covers company-attributed usage only. A company-market relationship is a company account and a metro market it has booked in; there are 949 in total, of which 883 have at least twelve calendar months of observable history after their first booking. "Sustained" means activity in six or more separate calendar months of the first twelve. "Weekly-equivalent" means 24 or more bookings across months 7–12. Bookings include coworking day visits, guest visits, meeting room bookings, and private office bookings; cancelled bookings and sessions longer than 24 hours are excluded.

Two limits matter. The activation bands are descriptive, not causal: markets that open with several people may differ from single-person markets in ways beyond the count, and we are not claiming that adding a second booker causes durability. And this measures workspace booked through Croissant, so a company's leases and local arrangements outside the platform are invisible here, meaning the true rate at which company demand justifies fixed space could be higher than 9.1% for organizations already committed elsewhere.