How to Price Window Cleaning: Per-Window Rates and AI Quotes From Photos

AI can give a rough window count from customer photos or Street View, but not a dependable one, so treat it as a first estimate. Price per window by dividing your hourly target by the windows you clean an hour, add a minimum charge, round, and visit first when the house is big.

The weak point is hidden glass. Street View shows only the side facing the road and may be years old, and customer photos rarely cover the back, so photo counts tend to come in low. Test the count against houses you have already cleaned before you quote from pictures alone, and keep the per-window sum as the part you control.

Follow me on Instagram@sagnikteaches

Start with your hourly target, not a market rate

Copying a competitor's per-window price copies their costs, their speed and their mistakes. Build yours from your own numbers, in three steps.

Connect on LinkedInSagnik Bhattacharya

1. Set the hourly figure you need to earn on site. It has to cover your wage, van, fuel, insurance, equipment, water-fed pole or squeegee kit, supplies, admin and a margin for non-working time. Illustratively, say $55 an hour on site.
2. Time a normal week. How many standard windows do you clean in an hour, including setup, frames and sills? Say 12 for a typical house.
3. Divide. $55 divided by 12 is about $4.60 a window.

Round to a figure that is easy to say and check, such as $5, and treat it as your base per standard window. These figures are placeholders to show the sum. Your speed and your costs will differ, and time in the van between jobs belongs in the hourly figure, not the per-window rate.

Subscribe on YouTube@sagnikteaches

A rate card with the modifiers that matter

A single rate does not survive real houses. Add a short list of modifiers so a customer can see why one window costs more.

Write your rule for what counts as one window at the top. A common choice is one opening or one sash equals one window, with divided windows charged separately. Whatever rule you pick, print it on the quote, because disputes usually start from different ideas of what a "window" is.

The minimum charge protects the small jobs

A flat with six windows at $5 is $30. Add driving and setup and you earn well below your hourly target. A minimum charge fixes that. Work it out: if a job costs you 20 minutes of travel and setup before you touch a window, and your hourly figure is $55, those 20 minutes cost about $18. Six windows at 12 an hour take another 30 minutes, about $28. The visit uses roughly $46 of your time and brings in $30. The $35 minimum on the rate card narrows the gap; if most of your small jobs look like this one, something nearer $45 closes it. Set it from your own timings, and review it when fuel prices or your speed change.

Rounding without annoying anyone

Customers react badly to odd figures like $83.40 for a window quote, and you waste time calculating them. Round the total, not each window, to the nearest $5. A 16-window house at $5 is $80 and needs no rounding. A 17-window house with three divided windows at $7 comes to 14 times $5 plus 3 times $7, which is $91, rounded to $90 or $95 as you prefer. Round up when the job is awkward and down when you want to win a regular. Keep your habit consistent so repeat customers see the same logic.

Worked quote: a detached house with hidden rear windows

A customer sends two photos of the front and side of a detached house and says there are "about 18 windows" with a conservatory at the back. All figures are illustrative and use the rate card above.

Count. The owner counts from the photos: eight front windows including a bay, four on the visible side, plus two upstairs on the side, so 14 visible. The customer says 18, so there are about four at the back. The conservatory is separate.

Price. Of the 18 windows, two bay sections are small-pane, so they go at the $7 divided rate. Outside only: 16 at $5 is $80, plus 2 at $7 is $14, so $94. Inside as well adds $3 on each of the 18, which is $54, bringing the total to $148. If the back holds two more windows than the customer remembered, those figures become $104 and $164. The conservatory roof is quoted separately after a visit.

Rounding and range. Because four windows are not seen in the photos, the owner quotes a range: "$95 to $105 outside only, $150 to $165 inside and out, conservatory quoted separately; final price confirmed once I've counted on the day." Time spent: about five minutes, helped by a prompt that does the arithmetic.

What AI can do with the photos

An assistant that reads images can list windows it sees on each photo, note types such as sash, bay or patio door, and flag obstructions like trees, scaffolding or a gate. It can also take your rate card and compute a price from a count you give it. What it cannot do reliably is count windows it cannot see, tell the difference between two panes and two windows, or judge how dirty the glass is.

I run a window cleaning business. Rate card:
standard window $5 outside, +$3 inside, patio or large door $10,
divided window $7, minimum charge $35, conservatory roof quoted separately.
Here are photos of a house. For each photo:
1. List every window you can see, with a type (standard, bay, patio door,
   divided, skylight, conservatory).
2. Say what you cannot see (rear, other side, upper floors behind trees).
3. Give a visible count only. Do not guess hidden windows.
Then calculate a price range using my rate card for the visible count,
showing the working line by line. Do not round. Flag anything that
needs a site visit.

Ask for working shown so you can check it. ChatGPT accepts images up to 20 MB each, so phone photos upload without trouble. What you would fix in the reply: assistants tend to add confident guesses ("approximately 6 windows at the rear") that are not in the photos. Your instruction to give a visible count only stops this.

Testing photo counts against jobs you have already done

Before you trust it, test it. Take 15 houses you have already cleaned, where you know the true count. Use Street View screenshots or the photos customers sent, give the assistant only the pictures, and record its visible count against the real number.

An illustrative result might look like this. Of 15 houses, the count matched exactly on 6, was within one on 11, within two on 14, and one was badly wrong at nine windows short, a house with a rear extension and a side return not visible from the street. Typical errors are under-counts, because windows are hidden, not over-counts.

Now convert the error into money. At $5 a window, being out by two windows is $10, which is small. The costly errors are not the counts but the type and access: a big patio door counted as a standard window, a third storey not visible, a conservatory roof nobody mentioned. The nine-window miss, at $5 each, is $45, and the cost of that to you is the lost margin on a job you underpriced, plus the awkward conversation on the doorstep.

Write down what went wrong in each miss. In this illustration, of the nine counts that were not exact, four were short because trees blocked the view, three missed rear windows, and two came from old Street View images taken before an extension or loft conversion. These are the facts that go in your photo request and your quote wording.

Street View: handy, dated and incomplete

A Street View screenshot is a quick way to see a house before the customer replies. But it shows only what faces the road, may be several years old, and often misses extensions, loft conversions and rear windows. Imagery shows its capture date, so look at it. If the house has changed, the count will be wrong. Check the imagery provider's terms of use for commercial use before you build a routine around screenshots.

Use Street View for the front count and the type of windows, and use the customer for the rest. A reply that asks "Roughly how many windows in total, and are there any conservatories, skylights or windows above the second floor?" is more reliable than any image guess. For a related look at the same problem on another job, see pricing gutter cleaning from Street View.

What to ask customers to send

1. One photo of each side of the house you will clean, from the ground, showing all the windows.
2. A total count of windows, with any conservatories, skylights or patio doors listed separately.
3. Whether they want inside, outside or both, and whether frames and sills are included.
4. Anything that blocks access: parked cars, locked side gates, dogs, planters, scaffolding.
5. Any windows above the second floor.

Put this list in your first reply. An assistant can personalise it to the message, keep the tone friendly, and cut the reply to a text-sized message.

When to visit first

Photo quotes work best for ordinary houses where you can see most of the windows and the access is simple. Visit, or ask for a video walk-round, when any of these apply:

• More than about 25 windows, or a large, irregular building.
• Conservatories, glass roofs, skylights, or windows you cannot reach with your usual kit.
• Third storey or higher, or difficult access.
• Heavy dirt, paint, scaffolding marks or building work residue.
• A difference of more than two or three between the photo count and the customer's count.
• Any commercial premises, where access, hours and insurance rules matter. For those, per-pane quotes built from your labour sets out a different approach.

Visits cost time, so combine them: a quick look when you are already working on the same street often costs nothing.

A second example: a terrace with a regular round

Suppose a customer on a street where you already clean wants a quote. You know the house type from the neighbours, your count is easy, and the customer wants a four-weekly clean. In this illustration the house has 10 standard windows, $5 each is $50, and you give a regular-customer price of $45, rounded to the nearest $5. At four-weekly frequency, that is 13 cleans a year and $585 of annual revenue from one house, with almost no extra travel because it is on your round.

The pricing logic here is density: the cost of a stop on a street you already visit is lower than the same stop across town. Many cleaners give a small discount for regular and same-street customers for that reason. Have the assistant draft the offer wording and keep the discount rule in your own rate card, so it does not drift from customer to customer. For planning and rebooking a round of this kind, see window cleaning round software.

Quote wording that keeps you safe

Thanks for the photos, [first name]. I can see 14 windows from your pictures, and you've mentioned about four more at the back. My price is $5 per standard window outside ($7 for the small-pane bay sections), so between $95 and $105, and $150 to $165 including inside. The conservatory roof is quoted separately. The final price is confirmed on the day once I've counted, and I'll tell you before I start if it differs.

This does four jobs: it shows how the count was made, gives a range, names what is excluded, and promises to confirm before starting. Have the assistant keep the wording fixed so each quote follows the same pattern.

Before and after: a vague enquiry becomes a usable quote request

Before (customer message): "Hi, how much for my windows? House on the corner, bit of a mess from the builders."

After (assistant's draft reply, illustrative): "Hi [first name], thanks for getting in touch. To give you a price I need a few things: a photo of each side of the house from the ground, a rough count of the windows, and whether you'd like inside as well as outside. You mentioned building work, so please tell me if there's paint, cement or plaster on the glass, because that takes longer to remove. I charge per window and confirm the final price on the day."

The assistant picked up the builders' mess and turned it into a question, which an experienced cleaner would do on the phone. Check its draft for promises: if it adds "we can come tomorrow" or a price before the photos arrive, take it out.

The first clean costs more than the regular ones

A house that has not been cleaned for a year takes longer than the same house cleaned every four weeks. Many cleaners price the first visit above the regular rate and drop to the standard rate once the glass is in good order. In this illustration, a 12-window house is $60 at the regular rate, with a first-clean price of $75 if the glass and frames are heavily soiled. Write the reason on the quote: "The first clean takes longer because of built-up dirt; later visits are priced at $60."

Ask the assistant to draft the explanation in a friendly tone, and decide the rule yourself: how much extra, and when it applies. Judging the dirt from a photo is unreliable, since glass often looks fine in a picture and dirty in person. Use a conditional line: "If the windows need extra work on the first visit, I'll tell you the price before I start."

Frequency and the price per window

Regular customers cost you less per visit because there is less dirt and no re-quoting, and a stop on your existing round costs less to reach. An illustrative frequency table, with all figures placeholders:

Check each line against your hourly target before you adopt it. At 12 windows an hour and $55 an hour, the four-weekly price of $55 only works if light dirt lets you finish in an hour or less. Your timing data decides, not a table borrowed from somewhere else. Record minutes on site for a month and see which frequency bands really are quicker.

The Street View quote that missed an extension

Here is an illustrative case. A cleaner used a Street View screenshot to quote a "small house" at $40. On arrival there was an extension at the back, a glazed porch and eight more windows than the photo showed. The job took twice as long. The customer had not said anything, because nobody asked. The fix was a standing question in every reply: "Are there any extensions, porches or conservatories at the back or side?" plus a rule that Street View counts are labelled "front only". In this illustration, surprises then drop from one job in five to about one in twenty-five in the cleaner's log.

The figures are invented to show the pattern. The method to copy is the log: record every job's quoted count, actual count and the cause of any miss. Within two months you have your own accuracy numbers and the right questions to ask.

Price rises and rate reviews

Review the hourly target every few months, and again after a change in fuel, insurance or equipment costs. When rates go up, tell customers properly; a price increase letter in text and leaflet versions covers the wording. Doing the sum behind the rate each time keeps you honest about whether your $5 is still $5.

If you want to measure windows rather than count them, window measurement apps are a different tool for a different job, better suited to glazing quotes than to cleaning.

Checking your quote accuracy after a month

Track three numbers: the share of quotes where the final count matched the quoted count within one window, the share where you changed the price on the day, and the average minutes to produce a quote. If the first is above about eight in ten and the second is low, the method works. If price changes on the day are common, the weak link is usually the photo request or a missing question, not the assistant.

Pricing is a business decision, and AI is a calculator and a drafting aid within it. The per-window rate is yours, the count is a shared estimate between you, the customer and the images, and the visit is the fallback when any of the three disagree.

Questions about pricing window cleaning from photos

Can AI count windows accurately from Street View?

Only the ones it can see, and the image may be years old. Rear windows, extensions, upper floors behind trees and side windows often do not show. Use it for a first count of the front, ask the customer for a total, and visit when the house is large or the two counts differ.

Should I price per window or per hour?

Work out your price per hour first, then turn it into a per-window rate using how many windows you clean in an hour. Customers like per-window prices because they are easy to check. You keep your hourly target behind it and add a minimum charge for short jobs.

How do I count a window with several panes?

Decide one rule and publish it. Many cleaners count each opening or sash as one window and charge extra for small-pane or divided windows. State the rule in your quote so a customer's count and yours mean the same thing.

When should I visit before quoting?

Visit when the house is large, has conservatories, skylights, high or awkward access, heavy dirt, or when the photo count and the customer's count disagree by more than a couple of windows. For commercial buildings, always visit.

Further reads

• AI for Window Cleaning: What Helps a One-Van Business and What Doesn't — What helps a one-van window cleaner and what does not.
• Commercial Window Cleaning Contract: Proposal, Access and Safety With AI — Proposal, access and safety wording for commercial contracts.
• How to Get Window Cleaning Customers: Leaflets, Local Search and AI — Find the customers to price once your rate card is set.
• How to Price Pressure Washing Jobs With an AI-Built Rate Card — The same rate-card method applied to pressure washing.
• AI for Cleaning Services: End-of-Tenancy Quotes and Photo Checklists — Photo checklists and quotes for end-of-tenancy cleans.
• Solar Panel Cleaning Business: Quote and Route Jobs With AI — A worked pricing sum by panel count and access class, plus where AI helps with quotes, route clustering and recurring-plan reminders.
• Stump Grinding Business: Can AI Price Jobs From Photos? — How to price stump grinding from customer photos: the text to send, a diameter-based sum, and the details that still need a site visit.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI help centre (image upload limits) and the capture dates shown in Google Street View; all rates, counts and the 15-job test are illustrative, not market data.