How the SquareFoot Flex Index is built

A per-desk, per-month price index for private offices and desks in serviced and coworking centres, one city at a time, outside the United States. Version 2026.10. Every number on a SquareFoot market page and in the JSON feeds follows the rules on this page.

1. What is measured

The asking or quoted price of a private office or dedicated desk in a serviced office or coworking centre, expressed per desk per month in the local currency, with a USD conversion at the mid-market rate on the day the price was seen. Hot desks, meeting rooms and virtual offices are collected and shown by product but never enter the headline office median. These are list and quote prices, not signed-lease rents: the index tells you what operators ask this month, which is what a tenant will be quoted when they start a search.

2. Sources, and what is never used

• Operator and landlord pages that permit automated reading. Before a site is read, its robots.txt is checked against the exact paths and its terms of use are reviewed; a site whose terms forbid automated access, or that answers with a bot wall, is read by email only or not at all. Pages are fetched slowly, from one identified crawler, no more than once a month per site.
• Availability and price mailings that operators and brokerages send to a SquareFoot mailbox after SquareFoot subscribed. Each sender is attributed by a dedicated address, so a price is always traceable to who sent it and when.
• Quotes received on inquiries. When an operator replies to a SquareFoot inquiry with a price, that price is recorded as a quoted observation. Quoted and asking observations are counted separately in every feed.
• Partner feeds from brokerages and operators that share inventory with SquareFoot under agreement.

SquareFoot does not scrape listing portals or any site whose terms prohibit it, does not bypass access controls or CAPTCHAs, and does not create accounts by machine. Every source carries its approval record: who approved it, when, and on what channel.

3. Extraction and normalisation

• Documents (pages, emails, PDFs, spreadsheets) are stored verbatim, then read into structured rows: centre, address, district, operator, product, desks, floor area, the price exactly as written, its currency and its basis (per desk per month, per office per month, per square metre per year…). Nothing is estimated; a field the document does not state is empty.
• Per desk per month is derived from the stated basis. A monthly office price with a stated desk count is divided by the desks. When only floor area is given, desks are derived from the market's square-metres-per-desk assumption (published in each feed as sqm_per_desk_assumption) and the row is flagged price_derived. Weekly, daily and annual bases are converted by calendar arithmetic (52 weeks, 12 months).
• Currency. A price stated in a currency other than the market's is converted at the daily mid-market rate; the USD figure uses the same day's rate. The rate and date travel with each observation.

4. Gating

Before a market's median is published it is also compared with what other publishers state for the same city (brokers' and marketplaces' own averages or medians, with their period and scope). Each feed lists those references and our deviation from each; a median more than 25% away from two or more references is held back and flagged for review rather than published as the headline.

A row is rejected when its city does not resolve to a covered market (so a UK operator's Bristol page never lands in London), when the per-desk price falls outside a plausibility band for the market, when the model's confidence in the extraction is low, or when the source is no longer approved. Rejected rows are kept with their reason and counted in the daily report. Accepted rows upsert one listing per centre and product; every price seen is appended to a price history, so a listing's current price and its history are both available.

5. When a city index is published

A market page and its feed exist as soon as a city has inventory, but the index is marked published, and the page allowed into search indexes, only when the city has at least 50 listed spaces, 25 with a price, from at least 2 independent sources. Below that the page says so and reports counts without a headline median. The page's price guide on squarefoot.com is written only when 30 or more office rows carry a stated price.

24 of 545 markets currently meet the publication gate. Counts refresh continuously; this table was rendered 2026-10-31.

6. Update cadence

• Inbound mailings and quotes are processed within 30 minutes of arrival; accepted rows are published hourly.
• Coverage is snapshotted daily, and the publication gate re-evaluated with it.
• Operator pages are re-read monthly.
• Price guides on squarefoot.com are regenerated daily from the same data.

7. Known limits

• Many operators, above all in Asia and the Gulf, quote private offices on inquiry only. Those centres appear with their desk products or with no price; the quoted-price channel is how their numbers enter the index over time.
• "From" prices are the smallest office an operator publishes; a team of eight will usually pay less per desk than the from-price suggests.
• Asking prices move with incentives (free months, fit-out) that the index cannot see.
• Derived per-desk prices depend on the square-metres-per-desk assumption; the feed states how many office rows are derived versus stated.
• Coverage is of flexible, serviced and coworking space. Conventional leases are out of scope here.

8. Licence and citation

Feeds and the published index figures are licensed CC BY 4.0. Cite as "SquareFoot Flex Index, <city>, <month year>" with a link to the market page. Each feed carries a ready how_to_cite string. Questions about the data: [email protected].

9. Changes