What we publish
A private office rate per seat per month, a coworking day pass rate and meeting room rates for every market in each covered country. Each rate carries a confidence tier, a line saying what it is built from, and an 80% range. Launched partners also see office size bands and meeting rooms in six sizes. Rates refresh monthly. New markets are added as their rates pass our accuracy test. A market without enough published rates shows no rate.
Month-over-month change is measured on venues present in both months, so a change in which venues list does not read as a price change. Past months are revised when new rates arrive; the latest release replaces earlier figures. We publish no forecasts.
Three rates behind every price
Every price on MarketTide comes from one of three rates, each counted per person so markets compare fairly. The product prices you see, such as an office for five or a room for eight, are built from them.
| Rate | What it measures | Prices built from it |
|---|---|---|
| Seat Rate | A seat in a private office, per month. The market rate, estimated and tested. | Private office by size: smaller offices cost more per seat |
| Day Rate | A coworking day pass for one person. The market rate, estimated and tested. | Coworking day pass |
| Seat Hour Rate | One seat in a meeting room for an hour: a room's hourly price divided by its seats. The median of listed rooms of that size. | Room Rate: the typical hourly price of a 4-person and an 8-person room |
Bigger rooms cost less per seat, so a Seat Hour Rate is always quoted with its room size. Seat Rate and Day Rate use the full method below. Meeting room rates are the median of rooms listed in the market, with no model, and publish once 10 rooms of a size from 3 venues and 3 operators are listed; no operator carries more than half the weight. Every rate shows an 80% range under it.
Where the rates come from
MarketTide reflects published flex workspace rates from operators around the world: rates listed publicly on the LiquidSpace marketplace and rates operators publish or send us, including standard rate cards that operators publish across their venues. We blend them with public statistics, such as income, GDP per head and population, from national statistics agencies. A census of known flex venues tells us which operators are present in each market.
Where few venues publish their own rates, the model also learns from price evidence we hold privately, such as rates operators quote for one client, restated to a list-price basis. That evidence shapes the model but never appears as a local rate, and no published figure can be traced back to it. Only publicly listed rates count as local rates.
Location search uses place and postcode data from GeoNames (CC BY 4.0) and the US Census Bureau. It finds the market for a place; it doesn't locate street addresses.
How we estimate
For each market we take the published rates of the venues there, adjust each for its operator and average them. That is the local rate. Separately, a model predicts the same rate from public statistics such as income and population. The published rate blends the two. The more venues publish rates in a market, the more weight its own rates carry: a market with many published rates is measured almost entirely from them, and a market with none is estimated by the model alone. The 80% range widens as local data thins.
Where few venues publish their own rates, the rate starts from published standard rate cards, adjusted to the local market with public statistics. The market page says so.
| Step | What it does | Standard method |
|---|---|---|
| One price per venue | Each venue counts once, so one chain cannot swamp a market | Equal weight per outlet, as in US Bureau of Labor Statistics price indexes |
| Operator adjustment | Removes each operator's price level before averaging, so markets compare fairly | Fixed-effects regression, as in hedonic and CPI quality adjustment |
| Market model | Predicts a market's rate from public statistics, so markets with few published rates still get an answer | Small-area regression, as in US Census Bureau county estimates |
| Credibility blend | Weights local rates against the model by how many local rates there are | Credibility weighting (empirical Bayes), as in insurance ratemaking |
| Holdout testing | Hides each market, predicts it from the rest and measures the error before every release | Holdout cross-validation, as in Zillow Zestimate error reporting |
Confidence
Reading a rate
Three US markets from the October 2026 release, from the most local data to none. The less local data behind a rate, the wider its range.
| Market | Seat Rate | 80% range | Local rates carry | Line under the rate |
|---|---|---|---|---|
| Los Angeles-Long Beach-Anaheim, CAVerified | $543 | $519 to $566 | 94% | Mostly published rates in this market |
| Atlanta-Sandy Springs-Roswell, GAEstimated | $421 | $377 to $470 | 58% | Published rates in this market, blended with the model |
| Nevada County, CAEstimated | $425 | $352 to $513 | 0% | Model estimate from public statistics. Few venues here publish their own rates. |
Los Angeles is measured almost entirely from its own published rates, so its range is narrow. In Atlanta its own rates carry 58%, the model carries 42% and the range widens. Nevada County has none, so the model sets the rate and the range is widest.
How accurate it is
Before every release we hide each tested market, predict its rate from everything else and compare the two. Median error on private office rates, and how often the 80% range held the true rate:
| Country | Markets tested | Median error | 80% range held the true rate |
|---|---|---|---|
| United States | 150 | 9.1% | 88% |
| Canada | 17 | 7.9% | 94% |
| Germany | 53 | 11.0% | 94% |
| United Kingdom | 33 | 17.3% | 79% |
| India | 38 | 16.9% | 84% |
| Australia | 24 | 16.6% | 83% |
The test predicts each market with its own rates hidden. Add just 2 local venues and median error is 7.9% in the United States, 11.1% in Germany and 8.1% in the UK.
Cities outside these countries are tested by leaving out one city at a time. The 80% range held the true rate in 92% of tests in Asia-Pacific and 73% in Europe. The largest city error was 6.4% in Asia-Pacific and 17.6% in Europe.
For comparison, Zillow reports about 7% median error for homes not on the market, and above 12% in low-data states. That is why we publish tiers and ranges, not single numbers alone.
What we never publish
- Any rate that could reveal one operator's price. A market's listed rates count only with at least five venues and no operator above a quarter of them. Below that, the market shows the model rate. Standard rate cards that operators publish themselves are public and always count.
- Booked prices, signed deal prices and quotes made for one client.
- Booking counts or any figure that reveals booking volume.
- One operator's price level against another's.
Rates are estimates, not appraisals or offers.