I saw the same number in five different articles last month. The average estate agent takes 917 minutes to respond to a new lead. Over fifteen hours. Every article credited it to Inman’s 2025 Real Estate Technology Survey.

It sounded precise. It sounded recent. So I went looking for the survey.

The number comes from a 2014 study by WAV Group and Weichert. It measured 384 US brokers across eleven states, and nearly half of the enquiries they sent received no reply at all. Twelve years old. American secondary market. Individual agents, not development companies.

Nobody lied.

Somebody typed “2025” once, and the rest of us copied it.

Provenance check 01

This article is not really about that one number. Once I started checking, I found the same thing at every stage of the funnel — audience, cost per lead, website behaviour, response speed, qualification. Different sources, different decades, different industries. And one shared weakness that only becomes visible when you line them up in order.

Step One — Who is the buyer?

Most buyer research quoted in English-language property marketing comes from one place: the NAR Profile of Home Buyers and Sellers. It is a serious piece of work, published for over forty years, with a large sample and a transparent method.

It also carries a note that almost nobody repeats. NAR states plainly that the data covers primary residence buyers only — not vacation home buyers, not investors.

If you sell to international buyers, read that twice. The most-cited buyer research in our industry

excludes your buyer by design.

A Polish family buying an apartment in Spain is not in that dataset. Neither is anyone buying to rent out, or buying a second home by the sea.

The individual findings are real. In the 2025 edition the median buyer age reached 59, and first-time buyers fell to 21% of the market, the lowest share NAR has recorded since 1981. That is a true statement about American primary housing. It is not a statement about the Mediterranean.

Europe has its own numbers. The RE/MAX European Housing Trend Report 2025 covers 23 countries and puts the average first-time buyer at 31.3 years, from 28 in Malta to 34.7 in Switzerland and Greece.

You can see what happens next. Someone puts 59 and 31 on the same slide, and a theory about generational difference writes itself. But one figure is the median age of all buyers and the other is the average age of first-time buyers. There is no twenty-eight year gap. There is only a slide.

I have watched that comparison get made in a planning meeting, and I did not catch it at the time either.

Step Two — What should a lead cost?

Every marketing director has been asked this by a CEO. The honest answer — “it depends what we count as a lead” — sounds evasive, so we reach for a benchmark instead.

The usual one comes from WordStream and LocaliQ, drawn from more than 13,000 search campaigns across 23 industries. For Real Estate, cost per lead lands around $100 on Google. On Facebook, the same source puts real estate at $16.61.

Three things about that category are worth knowing. It holds estate agents, brokerages, letting agents, property managers and developers in a single average — a letting agent chasing rental enquiries sits in the same number as a developer selling €600,000 apartments abroad. It is not the market either; it is one agency’s client base, mostly small and mid-sized American advertisers who chose to work with that agency.

And cost per lead is calculated from whatever each advertiser tagged as a conversion. In one account that means a viewing request. In another, a brochure download. In a third, a tap on a WhatsApp button.

You can see the strain in the numbers themselves. A second widely-shared figure puts the average real estate cost per lead on Google Ads at $42. Two published benchmarks for the same metric,

more than twice apart,

both quoted confidently, rarely in the same article.

Then there is the comparison the data invites: $100 against $16.61. Six times cheaper. If you have ever moved budget on that logic and watched lead volume rise while qualified pipeline fell, the next section is where the money went.

Step Three — What happens on the website?

Ruler Analytics tracks conversion by traffic source rather than by site. Their channel breakdown is the most useful table in this article, because it explains the cheap lead.

Source: Ruler Analytics conversion benchmarks. Ruler defines a conversion as a qualified lead or sale, which is stricter than most platform-reported conversions.

Direct converts 3.7× better than paid social. Source: Ruler Analytics conversion benchmarks. Ruler defines a conversion as a qualified lead or sale, which is stricter than most platform-reported conversions.

Direct traffic converts roughly four times better than paid social. That is not variation around an average — it is two different kinds of visitor arriving at the same pages. Which means a site-wide conversion rate mostly

describes your traffic mix, not your website.

Change the mix and the number moves without anyone touching a page.

The same logic quietly undermines the most repeated website statistic in property: that 53% of visitors leave if a page takes more than three seconds to load. It comes from DoubleClick’s “The Need for Mobile Speed”, published in September 2016, measuring publisher sites in an advertising network over 3G connections that averaged nineteen seconds.

Speed matters. I am not arguing otherwise. But “abandoned” in that study means the visit ended before the page rendered — a visitor who waits, looks at your project and decides against it is not in the 53%. And in GA4 a bounce is now a session under ten seconds with no conversion and no second pageview, where Universal Analytics counted any single-page visit however long. When the industry migrated, bounce rates fell without a single visitor changing their behaviour.

Step Four — How fast do you answer?

This is the one everybody knows: respond within five minutes and you are 21 times more likely to qualify the lead than if you wait thirty.

The original document is easy to find and worth twenty minutes of anyone’s time. It is a presentation given by InsideSales.com at MarketingSherpa’s Business-to-Business Demand Generation Summit in October 2007, analysed by Dr James Oldroyd, then a faculty fellow at MIT Sloan. Three years of data, six companies, fifteen thousand leads.

Provenance check 02

The detail I keep returning to is not the sample. It is that the authors were honest about not knowing why speed worked. They offer a few guesses, and the main one they call presence: someone who has just submitted a web form is sitting at their computer desk, probably near their phone. Call immediately and they pick up.

That is a description of an office worker in 2007 next to a landline. Now picture the buyer you actually have: on a phone, on a sofa, at 22:40, in a different time zone, thinking about a country they visited once on holiday.

The mechanism does not survive the journey. And if the mechanism does not transfer, the number cannot either.

There is also a closing section in that 2007 paper explaining that the research shifted the company’s positioning, and that its new callback dialler opens fresh ground in web marketing. That does not make the data false. It does mean this was never neutral evidence, and we have treated it as neutral evidence for nineteen years.

What actually helped me was a different piece of work. Mike DelPrete sent secret shoppers at online property enquiries and published both measures of the same result.

Source: Mike DelPrete, secret shopping study. Enquiries were sent during weekday business hours.

Same dataset. Same enquiries. Two opposite conclusions. Nearly half of the enquiries were never answered at all. Source: Mike DelPrete, secret shopping study. Enquiries were sent during weekday business hours.

“The industry takes eight hours to reply” and “half the industry replies within forty minutes” are both true, and they lead to opposite decisions. Response data is dominated by its tail — the enquiries nobody ever touched. Averages get dragged into that tail. Medians do not. If your dashboard shows average response time, it is mostly reporting your failures, which is useful, but it is not telling you how fast your team is when the process works.

One more finding, from British mystery shopping by Homeflow: agents answered rental enquiries within an hour 70% of the time, and sales enquiries only 40% of the time. Same agents. Same inbox. Different type of enquiry, and a thirty-point difference in behaviour.

Which is the same problem as the cost-per-lead benchmark, one step further down the funnel.

Step Five — What happens after the call?

By this stage the benchmarks stop pretending. Published MQL-to-SQL rates sit at 12–21% by sector, 15–35% in one summary, and 39% in another. Contact attempts before a rep gives up: 1.3 in InsideSales data from 2008 to 2012, while other research puts the number needed to reach a prospect at eight.

One aggregator says the quiet part out loud. Every benchmark article defines conversion differently — 2.9%, 31%, 62% — and all of them are correct, because they are measuring different stages of different funnels.

That sentence is the whole article.

Each borrowed number failed in the same place, and it was never the sample size or the date. It was the definition of the thing being counted.

NAR counted primary-residence buyers. WordStream counted whatever each advertiser tagged. DoubleClick counted visits that ended before rendering. Oldroyd counted qualified leads and openly did not count sales. Five careful measurements of five different objects, all filed in our decks under one word.

This matters most if you have a qualification step. When a call centre screens enquiries before sales sees them, you have a stage that exists in your funnel and in nobody else’s benchmark. Your MQL is not their MQL. Your SQL is defined by what your team writes in a notes field on a Tuesday afternoon.

Finally — Why we all do it anyway

The easy answer is laziness, and I do not think that is right. I have been on the wrong side of this often enough to say so.

Property marketing has almost no public evidence base of its own. There is no NAR for European developers selling to international buyers. There is no shared definition of a qualified lead, and no dataset anyone can point to. So we import — B2B research from 2007, American secondary-market surveys, publisher page-speed studies. Not because we believe they fit, but because the alternative is standing in front of a CEO with no number at all.

And that, I think, is what these statistics are really for.

A benchmark is not knowledge. It is a way to end an argument.

“The industry average is €100 per lead” closes a conversation. “Our cost per qualified buyer rose 40% because we scaled a channel that produces enquiries rather than buyers” opens one. The second sentence is more useful and far more expensive to say. It requires you to have measured it. It invites follow-up questions.

It puts you on the hook.

I am not arguing that we stop using external data. Some of it is good, and knowing the shape of the wider market has real value. The fix is smaller and more irritating than that: before a number enters a decision,

know what it counted.

Who was in the sample, which definition was used, who paid for the study, and whether the mechanism behind the finding still exists in your situation.

Which leaves a question I cannot answer for you.

Your own account holds a few hundred leads. Maybe a few thousand. It is small, messy and incomplete, and it would never survive peer review. But it was collected in your market, from your buyers, using your definitions, in this decade.

How much of your last budget decision rested on that — and how much rested on six B2B companies in 2007?