Priceguide research · September 2026
One million questions
What 171,797 homeowners told us about how they actually buy home services — including the finding that most home service businesses are closed for the majority of their own pricing conversations, and that it is getting worse every quarter.
On 10 April 2024, someone answered the first question inside a Priceguide estimator. Twenty-nine months later the count passed one million. This is not page views, impressions or visitors. One million times, a homeowner sat in front of a home service company's website and chose to type or click something specific about their own project, because they wanted to know what it would cost. As far as we know it is the largest single record of how homeowners behave when they are trying to find a price and nobody is watching them.
1. You are closed for most of your own pricing conversations
We took every estimate with a valid US postcode, converted the timestamp into the homeowner's own local time, and asked whether they were doing this during business hours.
That holds even if you are generous about what counts as a working week.
| Your working week | Estimates arriving outside it |
|---|---|
| Monday to Friday, 9am to 5pm | 54.9% |
| Monday to Friday, 8am to 6pm | 47.0% |
| Monday to Friday, 7am to 7pm | 41.3% |
Run a twelve-hour day, five days a week, and you are still absent for two in every five pricing conversations about your own services.
This is not a problem you fix by answering the phone faster. The homeowner is doing this in the gap between their working day and their bedtime, and that gap is precisely where most home service businesses do not exist.
2. And it is moving in one direction
A single snapshot tells you about homeowners today. A trend tells you where this is going. It is rising, steadily, every quarter.
Estimates arriving outside Monday to Friday, 9 to 5
There is an obvious objection to this, and it is a fair one. Our customer base changes. If we signed a lot of businesses whose customers happen to shop late, the average would move without any individual homeowner changing their behaviour at all.
So we ran it again on a fixed group: the 412 estimators live in both 2025 and 2026, compared only against themselves.
| Same 412 estimators | 2025 | 2026 |
|---|---|---|
| Outside Mon–Fri 9 to 5 | 54.1% | 57.0% |
| Weekend | 22.9% | 24.9% |
Same businesses, same industries, same estimators. Nearly three points in a year. The behaviour is genuinely shifting, not just the mix.
If it continues at this rate, the average home service business will be closed for 60% of its own pricing conversations some time in 2027.
3. Evenings, nights and the 3am estimate
The quietest hour is 3am, and even 3am is not empty. Nearly 500 people have sat down at three in the morning to find out what a job was going to cost.
It is worth thinking about who that actually is. Someone who cannot sleep because of a leak. A shift worker. A new parent. Someone who had a conversation about money over dinner and went downstairs to find out whether they could afford it.
Idle curiosity does not survive until 3am. Worry does.
4. Sunday is bigger than Saturday, and Monday is bigger than both
Share of estimates by day of the week
Saturday is the day people do things. Sunday is the day people plan things, and planning is when the pricing question gets asked.
Which reframes your Monday morning inbox. That enquiry did not appear on Monday at 9am. It was created on Sunday evening while you were with your family, and by the time you reply your real response time is not twenty minutes, it is fourteen hours. You were also, statistically, not the only company they asked.
5. It is a phone, and after dark it is almost always a phone
Since April 2026 we have been able to see what people are using. Of 154,907 human visits to an estimator:
Through the working day, mobile sits at around 56%, roughly level with desktop. In the evening it climbs past 75%.
So the out-of-hours half of your pricing demand is not only happening when you are closed. It is happening on a four-inch screen, one-handed, probably on a sofa. If your pricing page assumes a desktop and a patient reader, it is built for the wrong half of the day.
6. Homeowners are far more patient than the industry believes
There is a persistent belief in home services that people will abandon anything longer than three fields. A million answers say otherwise. The average completed estimate contains 6.3 answered questions, but the distribution is the real story.
Roughly one in six homeowners answered ten or more questions in a single sitting, voluntarily, to get a number.
The lesson is not to build a longer estimator. It is that length is not what kills completion. Pointlessness is. People will tell you the square footage, the access, the material, the condition and the timeline, as long as each answer visibly moves them closer to a price. What they abandon is the form that asks for a phone number and gives nothing back.
7. When the phone number is optional, seven in ten give it anyway
First, an honest caveat. Most Priceguide estimators require a name, an email address and a postcode before showing results, so the capture rate on those fields is close to 100%. That is a fact about how the product is built, not a discovery about homeowners, and we are not going to dress it up as insight.
The phone number is different. On most estimators it is genuinely optional, which makes it the one contact field that measures willingness rather than compulsion. Across the 343 estimators where it is optional, covering 88,921 estimates:
| Year | Gave a phone number where optional |
|---|---|
| 2024 | 70.6% |
| 2025 | 71.8% |
| 2026 | 74.1% |
Worth showing the working, because the naive version of this is badly misleading. Measured across all estimators, phone capture appears to climb from 72.5% to 85.8% in two years. It did not. The share of estimates coming from estimators that require a phone number went from 3.1% in 2024 to 45.1% in 2026. The blended average rose because the mix changed underneath it. Strip that out and the real movement is about three and a half points.
The finding that matters is the level, not the trend. Seven in ten people volunteer a phone number when nothing forces them to, because the exchange is obvious. They give you a way to reach them, and they get a price. Compare that with a contact form asking for the same number in return for the promise that somebody will get back to them at some point.
8. One in eight comes back
Some are pricing a second job. Many are re-running the same job with different choices, checking what happens to the number if they change the material, the size or the scope. That is a homeowner doing the thing they have never previously been able to do in this industry: exploring the price privately, without a salesperson in the room, before deciding whether to talk to anyone.
1,280 of them ran estimates with more than one business. Small in absolute terms, but it is the clearest signal in the dataset of what instant pricing eventually becomes. When two companies both publish a price, the comparison happens in an afternoon rather than across three weeks of missed calls.
9. The machines have started reading your prices
This one is small, new, and moving fast. Since April 2026 we can see when an AI assistant's crawler visits an estimator rather than a person — not search engines, which are a separate and much larger 13.7% of traffic, but the crawlers behind AI assistants: GPTBot, ClaudeBot, OAI-SearchBot, Meta's agent and others.
Share of visits from AI assistant crawlers, 2026
The practical point is simple. When a homeowner asks an AI assistant what a new roof costs in their area, the assistant answers from whatever it can read. A business whose pricing exists only inside a salesperson's head cannot be included. A business that publishes real numbers can.
10. The demand is compounding
| Period | Estimates |
|---|---|
| January to August 2025 | 42,460 |
| January to August 2026 | 92,575 |
One seasonal note worth having, since it is consistent across both years: December is the trough and January jumps hard. Between December 2025 and January 2026, volume rose 41% in a single month. People decide over the holidays and act in the new year.
What we would do with this if we ran a home service business
Assume the pricing conversation starts without you. Most of it already does, and the share grows every quarter. The only real question is whether it happens on your website or on a competitor's.
Measure response time from when the homeowner wanted the answer, not from when you saw the enquiry. On a Sunday evening enquiry answered at 9am Monday, your true response time is fourteen hours.
Build the pricing experience for a phone in the evening, because that is the single most common context it will be used in.
Stop treating length as the enemy. Ask everything you need in order to give an accurate number. People who are serious will answer. People who will not answer six questions were never going to sign a contract.
Publish real numbers where a machine can read them. The AI share of traffic is under 1% today and tripling. That is exactly the point at which acting is cheap.
A note on the numbers
Every figure here comes from Priceguide's own production data as of 7 September 2026, covering all 171,797 completed estimates. No sampling.
How the question count is calculated
The figure of 1,073,641 counts distinct answered question fields and is deliberately conservative. A fuller parse puts the real figure closer to 1.1 million.
Why only the phone number is reported for contact capture
Contact capture is reported only where it measures a choice. Name, email and postcode are required fields on most estimators, so their capture rates approach 100% by design and are not reported as findings. The phone number figure is calculated only across the 343 estimators on which it is optional, so it reflects what homeowners chose to do rather than what the form obliged them to do.
How timing is calculated
Timing is calculated in the homeowner's own local time, derived from the US postcode they entered and its timezone, with a standard daylight saving adjustment. Estimates without a valid US postcode — mainly Canadian, UK and Australian — are excluded from timing analysis entirely rather than being forced into a US clock. That leaves 120,060 estimates behind every timing percentage above.
Why the trend is reported twice
The trend analysis is reported once across all estimates and once across only the 412 estimators live in both 2025 and 2026, so that a change in our customer mix cannot be mistaken for a change in homeowner behaviour. Both show the same movement in the same direction.
Device, AI crawler and early-period data
Device and AI crawler figures cover April to September 2026 only, which is the period for which that data exists. Device timing is anchored to US Eastern and is approximate to within an hour. Figures from 2024 and early 2025 come from a much smaller base and include our own testing, so they are shown for context but no trend claim rests on them.
See what this looks like on your own website
Priceguide builds online pricing estimators for home service businesses, so a homeowner gets a real number on your website at the moment they ask for it — including at 10pm on a Sunday.
Try a live estimator · Work out what it is worth to you · See pricing
Questions about the data or the method? Email [email protected]. Happy to share more detail with anyone writing about it.