How Distributed Companies Actually Manage Workspace Across Multiple Cities

Ask how a distributed company provides workspace and you will usually get a list rather than a system. A membership somebody signed years ago. Two leases inherited from an acquisition. A handful of direct arrangements with local spaces. Reimbursements for everyone else. A broker on call for the market that matters this quarter. Nobody designed it. It accumulated.

What has been missing is a measurement of how fast it accumulates. Across 287 company accounts and 71,395 bookings, we can count the actual suppliers a company touches as its demand spreads. The answer is that adding a market does not add a workspace relationship. It adds several, and it keeps adding them long after the demand itself has stopped growing proportionally.

How many venues and operators does a distributed company actually use?

Grouping companies by the number of metro markets they book in, and reporting medians because the distributions are heavily skewed:

A company in a single market uses a median of 6 venues from 5 operators. A company in 21 or more markets uses 251.5 venues from 183 operators across 14 countries. That is not a procurement list anyone maintains by hand, and it is the number most workspace plans never produce because no single system holds it.

Where the primary market finally stops dominating

Earlier Croissant research found that the largest market holds roughly 85% of a company's usage all the way from three markets to twenty, so geographic expansion adds a tail rather than redistributing demand. This data extends that finding and locates its limit.

Through twenty markets the pattern holds: the primary market keeps about 85% of usage even as the company operates in a dozen places. Past twenty markets it breaks, and usage outside the primary market jumps to 56.3%. That is the point at which a company stops being a headquarters with satellites and becomes genuinely distributed.

The venues-per-market column is the more useful operational number, and it is not monotonic. It falls from 6.0 to 3.8 as companies add their first ten or twenty markets, which is efficiency of a sort: new markets get a couple of venues each. Then it rises again to 8.2 in the most distributed band. Companies at that scale are not adding thin new markets any more, they are thickening the ones they have, with a median of 61 venues in their largest market alone.

A third of every company's venues are used exactly once

The most stable number in this analysis is also the most awkward one.

Roughly a third of the venues a company uses are used once and never again, and that share barely moves no matter how large or distributed the company becomes. Meanwhile the share of venues carrying the bulk of usage tightens from half to a quarter. So scale produces two things at once: a harder core of frequently used venues, and an undiminished tail of one-time visits.

Platform-wide the tail is even more pronounced. Of 1,325 venues used by company accounts, 18% were used exactly once, and just 2.7% of venues account for half of all company bookings. 11.8% account for 80%.

This is the practical case against per-market procurement. You cannot negotiate 183 operator relationships, and you would not want to, because two thirds of them will produce a handful of bookings and a third will produce exactly one. The tail is not waste to be eliminated. It is an employee in a city you have no presence in, needing a desk for a day. It has to be servable without a contract, which is a different requirement from anything a lease or a single-provider membership is built to satisfy.

Product complexity arrives late, and then all at once

Workspace type stays remarkably simple until companies get genuinely distributed:

Below eleven markets, more than nine companies in ten book coworking and nothing else. At eleven markets and above, meeting-room usage appears in 44.4% of companies, and only in the most distributed band does any company use all three types. The inflection is not gradual, which matters for anyone selecting a supplier: the requirement you have at four markets tells you very little about the one you will have at fifteen.

The administrative load per market rises too

Counting the administrative events attached to a company account, meaning cancellations, visit extensions, seat holds, prepaid hour reserves, credit quotes and cancelled room bookings, the median per active market runs 24.5 in the single-market band, dips to 19.2 in the 4–10 band, then reaches 60 in the most distributed band. Administrative events grew 64.6 times from the first band to the last, against 25 times for market count.

Per booking, though, the load is almost flat, at roughly 0.8 to 1.1 administrative events per booking across every band. That is the useful distinction. Distribution does not make each booking harder to administer. It multiplies the number of bookings and the number of places they happen in, and the coordination cost lands on whoever owns the account. Which is usually one person, and usually not their main job.

The six functions a distributed workspace system has to cover

Read against the data above, a workable system needs six things rather than a better vendor list:

1. Demand capture. Employees ask in the moment, in cities you may not serve yet. The system has to accept a request it cannot immediately fulfil from an existing relationship.
2. Policy. Rules about who books what, where, and how often, applied before spend happens rather than discovered afterwards. This is the same lever that makes employee choice compatible with budget control.
3. Fulfilment across the tail. A third of venues will be used once. Any model that requires a contract per venue fails that third by construction.
4. Transaction. One payment relationship, not 183. This is the single largest reduction available, and it is invisible on an org chart.
5. Administration. Cancellations, extensions, and holds at roughly one event per booking, at whatever volume distribution produces.
6. Intelligence. Usage by market and by person, which is the input to every subsequent decision about whether a given city has earned a fixed commitment.

Most companies have some version of the first and the last, patchily. The middle four are where the accumulation happens.

How to evaluate your current model

Four questions, each answerable from data you already hold:

Count your venues and your suppliers separately. If you can produce a market count but not a venue count, you are managing a map rather than a supply base. The venue-to-market ratio is the number that tells you whether you are thin in many places or thick in a few.

Find your one-time venues. Expect around a third. If yours is far lower, you are probably refusing requests you should be serving. If it is far higher, employees are improvising because coverage in their city is inadequate.

Check whether your primary market still holds 85%. Below twenty markets it very likely does, which means your central arrangement still deserves the attention it gets. Past that point, more than half your usage sits outside it and central arrangements stop being the main event.

Count invoices and contracts, not spend. Spend understates fragmentation because a large lease and forty one-off day passes can cost the same. The count of relationships is what determines administrative load, and the finance treatment of workspace as a variable cost depends on being able to state it.

Methodology and limits

This analysis covers company-attributed usage on Croissant: 71,395 bookings across 287 company accounts, from the first booking on record through 1 September 2026, restricted to companies with at least ten bookings in the window. A market is a metro area. A venue is a single workspace location. A booking is one coworking day visit, guest visit, meeting room booking, or private office booking; cancelled bookings and sessions longer than 24 hours are excluded. Medians are reported throughout because every distribution here is skewed, and band sizes are shown so small cohorts read as small.

Operator identity needs stating plainly, because there is no operator or brand field in the data. An operator is inferred from the payout entity attached to each venue, normalised to group the same business across spelling and legal-suffix variations. That identity is resolvable for a minority of venues in the catalogue, covering 1,335 venues and 1,066 distinct operators, so every operator count in this article is a floor rather than a total. The direction and the ratios are reliable; the absolute counts understate.

Two measures were tested and set aside. Venue management platform, meaning whether a venue runs on a given booking system, looked like a promising second fragmentation measure, but Croissant's own integration is present on all but a small fraction of the venue catalogue, so the metric largely measures us rather than the customer's vendor mix. It is excluded from every claim above. Administrative event counts depend on features that were introduced at different times, so the per-market rise is partly a function of newer accounts having access to more of them, and should be read as directional.

The most distributed bands are small: 9 companies in the 11–20 band and 6 above 20. Per-company medians with their band sizes carry the argument. Pooled shares at that end would describe two or three accounts rather than a cohort, so they are not reported. Restricting to the trailing twelve months leaves too few companies above ten markets to report at all, which is why the all-time window is used throughout.

Finally, this measures workspace booked through Croissant. Leases, broker engagements, and direct operator contracts held outside the platform are invisible here, which means the venue and operator counts describe one channel of a company's workspace supply and understate the total number of relationships it manages. The fragmentation described above is therefore a lower bound on the real thing. For how the underlying demand behaves as it spreads, see how workspace demand fragments across markets.