How SanCrew’s San Francisco Cleaning Price Data Is Calculated

This page documents the figures published on what house cleaning costs in San Francisco. It explains where the prices come from, how each figure is calculated, what is left out, and what the data can and cannot tell you. If you are citing the figures, this is the page that says what they mean.

Where do the prices come from?

They come from the same pricing system customers use to compare crews on SanCrew. When someone enters an address and a home size on the booking page, the platform asks every crew serving that address what they charge for that home, and shows the answers side by side. The published figures are those same answers, summarised.

Concretely: each figure is the price a customer would be charged at checkout, with the platform’s commission already included, exactly as it appears on the crew’s card. Nothing is re-derived, adjusted, or modelled. The page reads this data live and refreshes hourly, so a crew who changes their rate this afternoon changes the published figures within the hour, without anyone editing the page.

Who sets the prices?

Crews on SanCrew set their own rates. The figures here summarise the prices customers can see when comparing crews on SanCrew. SanCrew handles the booking and the billing.

How is the average calculated?

It is a plain arithmetic mean: every qualifying crew’s listed price for that home, added up and divided by the number of crews. Nothing is weighted — a crew who completes fifty jobs a month counts exactly as much as a crew who completes two. Nothing is trimmed either: no high price is removed for being high and no low price for being low, so the mean reflects the whole pool rather than a tidied version of it.

The minimum and maximum are not percentiles or modelled bounds. They are the actual lowest and highest prices listed by a crew for that home at the moment the data was read. If one crew lists far above the rest, the maximum shows it.

How big is the sample?

Every figure carries the number of crews it is drawn from, and that number moves as crews join, leave, or change which services they offer. Sample sizes are not large — this is a marketplace with a defined set of crews, not a survey — which is exactly why the count is printed next to the figure rather than buried in a footnote.

A crew whose listing is incomplete is excluded rather than counted as zero. A missing or zero per-unit rate means that crew has not finished pricing that service, and treating that as a $0 price would silently drag the average down. Those exclusions are counted and reported alongside the figure.

Read live when this page was generated, not typed in. “Crews contributing” is the largest number of crews any single home layout drew on for that service.

What area does the data cover?

Every figure is quoted for one San Francisco address, in ZIP code 94115. The same address is used for every reading, which is the point: comparing this month to last month is only meaningful if the question stayed identical.

Two things that follow, and both matter if you are citing this. The data does not support neighbourhood-level claims — it is one address, not a grid across the city, and we do not publish per-neighbourhood prices because we have not measured them. And it is not a census of cleaners in San Francisco; it describes the crews who serve that address through SanCrew.

Are these prices paid, or prices asked?

These are prices crews list for the home configuration described — what a customer would see and could book at that moment. They are not an average of completed bookings. The difference is real: customers choose which crew to book, so what actually gets paid reflects that choosing, while a listed price reflects what is on offer. If SanCrew ever publishes transacted prices, it will be a separate dataset, labelled as such, not a quiet redefinition of this one.

What is excluded from the figures?

• •Add-ons. Inside the oven or fridge, interior windows, laundry, wall washing and the rest are priced separately by each crew and are not in these figures. A booking with add-ons costs more than the figure shown.
• •Tips. Tipping is optional and decided after the job, so it is not part of any listed price.
• •Hourly bookings. Hourly work is priced per hour rather than by rooms, bathrooms and kitchens, so a fixed home layout has no meaning for it. Hourly is not in this dataset at all — including it would compare two different things.
• •Incomplete listings.A crew missing a required per-unit rate for a service is excluded from that service’s figures and counted in the exclusions, never averaged in as a zero.

Promotional and referral pricing is also outside the dataset: the figures are the standard prices shown to a customer with no code applied.

Why is there no average for extreme deep cleaning?

Because the spread is too wide for a mean to mean anything. On extreme deep cleaning, and on post-construction, the highest listed price for the same home runs several times the lowest — not because crews disagree about a fixed job, but because the job itself varies enormously. An extreme clean of a lightly used flat and one of a kitchen with years of buildup are different days of work, and crews price that uncertainty differently. An average across those two would describe neither. Where that happens SanCrew publishes the observed range and the sample size, and leaves the average blank — in the tables and in the CSV alike. Where the spread is narrow enough for a mean to be representative, the mean is published.

What are the limitations?

This describes the crews and prices available through SanCrew, not the entire San Francisco cleaning market. Agencies, individual cleaners working off-platform, and crews on other marketplaces are not in it, and nothing here should be read as a city-wide market rate.

We think that is a strength rather than something to bury. Most published cleaning-price figures are estimates or survey recollections with no stated sample. These are actual listed prices, from a named and countable set of businesses, for precisely defined homes, at a stated address, with the sample size printed next to every figure and the exclusions declared. Narrow and verifiable beats broad and unfalsifiable.

Two other limits worth stating plainly. Sample sizes are small enough that one crew joining or leaving moves an average visibly, so treat small differences between home sizes or services as noise rather than signal. And because figures refresh hourly, two people citing the page on different days can legitimately quote different numbers — which is why the citation format on the hub page carries a date.

Can I use this data?

Yes, with attribution and a link. The full citation format, and the terms, are on the price data page. Please link to that page rather than to a dated copy, so the link keeps pointing at current figures.

Questions about the method, or a request for a cut of the data we do not publish? [email protected].