- On days they worked, 35 percent of employed Americans did some or all of their work at home in 2025. That figure, for your own workforce, is the input to the ratio.
- Ratios are set per team, not per company. A sales team and a finance team do not share desks at the same rate.
- 7 desks per person is a common landing point. Going below 0.6 without booking data is where the complaints start.
- The failure mode is never the maths. It is Tuesday, when everybody comes in at once and the station count is already built.
Short version: desk sharing works when the ratio is derived from attendance data and fails when it is derived from a property target, and the furniture package is where that mistake becomes permanent.
Set the ratio from attendance, then specify furniture that can move
The order matters, because furniture bought for a fixed layout is what makes a ratio impossible to correct later. On most fit-outs the ratio arrives after the drawings, which is precisely backwards.
Desk sharing changes what the furniture has to do. Stations become unassigned, so personal storage moves off the desk, and the layout has to survive being reconfigured once the first ratio turns out to be wrong. Modular office workstations are built around that assumption, with panel systems and pedestal options that come apart and reassemble rather than being bolted to a floor plan. The comparison points that matter when choosing between configurations are station count, panel height, and storage measured against cost per seat rather than against list price.
Specify for the ratio the client will have in two years, not the one going into this quarter’s announcement.
Start from the data you already have
The national picture gives you the shape of the problem, and your own badge data gives you the answer.
The Bureau of Labor Statistics reported in the 2025 American Time Use Survey that “on days they worked, 35 percent of employed people did some or all of their work at home”, published in the ATUS summary. That is a national average across every occupation, including the many that cannot be done remotely at all, so for an office-based employer the effective figure is usually higher.
Your own number is more useful and you almost certainly have it already, sitting in badge-in records or room booking data. What you want is not the average occupancy. It is the 95th percentile day, because that is the day the ratio has to survive.
Ratios that hold, by team pattern
The single company-wide ratio is the most common error, because it averages together teams with genuinely different behaviour.
| Team pattern | Typical attendance | Workable ratio |
| Client-facing sales | 2 days, unpredictable | 0.5 to 0.6 |
| Engineering | 2 to 3 days, clustered | 0.6 to 0.7 |
| Finance and operations | 3 to 4 days, regular | 0.8 to 0.9 |
| Leadership and exec support | 4 to 5 days | 1.0 |
| Contact centre or scheduled shift work | 5 days, rostered | 1.0 |
Applying 0.7 across all of those produces a floor that is simultaneously too tight for finance and half empty for sales, and the complaints will come from finance while the wasted space sits with sales.
The Tuesday problem
Every hybrid office has a peak day, and in most of them it is Tuesday, Wednesday or both.
An office at 0.7 desks per person that averages 55% occupancy sounds comfortable until you look at the distribution. If Monday and Friday run at 30% and Tuesday runs at 85%, the ratio is failing on one day a week, and one day a week is enough for people to conclude the policy does not work.
Three responses, in ascending order of difficulty. Stagger anchor days by team so the peaks do not coincide, which is free and works. Add bookable overflow space, which costs money and works. Or raise the ratio, which costs the most and is the honest answer if the first two have been tried.
What the ratio does not fix
Desk sharing is a space policy, and it is regularly asked to do a culture job it cannot do.
If people are coming in and finding nobody from their team present, the problem is coordination and not capacity. If they are coming in and taking calls all day in an open bay, the problem is the absence of enclosed space, and adding more shared desks makes it worse rather than better. The question of whether the space is set up for how work now happens is broader than the station count, and this piece on office relocation in 2026 covers the design, cost and programme decisions that sit around it.
The ratio is a necessary decision. It is not a strategy.
Put the ratio in the brief, with the evidence attached
The practical advice is short, and it is mostly about sequencing.
Get the attendance data into the brief before the station count is fixed. A number the client can check survives value engineering far better than one that arrived as a property target, and the data usually shows the office was already running at the proposed ratio informally.
Write a review date into the specification with a named metric, most usefully peak-day occupancy rather than average. And make the reconfiguration path explicit, because the question that comes back eighteen months later is never whether the ratio was right. It is what it costs to change it.
