AI Receptionist

The Speed-to-Lead Stat Everyone Quotes Is From 2007 — Here's What Our Call Data Says

The famous 100x speed-to-lead statistic is from a 2007 MIT Sloan study with InsideSales.com, not Harvard. Here is the correct attribution, what has aged badly since, and what our own call data says about response time in 2026.

An open printed research report on a cluttered desk beside a smartphone lit up with an incoming call, late afternoon light across the pages

I've sat through a lot of pitch decks this year, and the same slide keeps showing up: contact a lead within five minutes and you're 100 times more likely to reach them, 21 times more likely to qualify them. Almost every version of that slide credits Harvard.

Harvard didn't run it. And the study is old enough to vote.

The short answer on speed to lead

The 100x and 21x figures come from a 2007 Lead Response Management study by Dr James Oldroyd at MIT's Sloan School of Management, done with InsideSales.com, covering roughly 15,000 leads and about 100,000 dial attempts. It is real research. It is also nearly two decades old, and buyer behaviour in 2026 looks nothing like buyer behaviour in 2007.

Where the Harvard mix-up comes from

There is a Harvard connection, just not the one people cite. In March 2011, Harvard Business Review published The Short Life of Online Sales Leads by James Oldroyd, Kristina McElheran and David Elkington. Same lead author, different study. That one audited 2,241 US companies by submitting test web forms and timing the replies.

The findings were bleak. About 37% of companies answered inside an hour. Roughly 23% never answered at all. Among the ones that did respond, the average was around 42 hours. Firms that made contact within the hour were close to seven times more likely to qualify the lead than firms that waited even one hour longer.

So: MIT Sloan owns the 100x number, HBR published the 2011 follow-up, and somewhere in the retelling the two collapsed into "a Harvard study." If you're quoting it in a deck, quote it properly. Sales audiences notice.

What's aged badly

Three things.

First, in 2007 a web form was a strong buying signal. Filling one in took effort. Now a lead can arrive from a Google Local Services ad, a chat widget, a WhatsApp click-to-message button, or a comparison site that has already sent the same enquiry to four of your competitors. The intent behind those is wildly different, and treating them as one bucket is how you end up with a beautiful five-minute SLA and a terrible close rate.

Second, the baseline moved. When almost nobody responded inside an hour, being fast was a genuine edge. Plenty of your competitors now have an automated first touch. Speed has quietly gone from an advantage to a floor.

Third, and this is the one that annoys me: the original study measured dials, not conversations. A five-minute dial that goes to voicemail is not speed to lead. It's speed to nothing.

What our own call data shows

We run AI voice agents for home services, clinics and property firms, so we get to watch this from the inside rather than from a slide.

A few patterns hold up consistently across our deployments:

  • Under 60 seconds is a different animal to under 5 minutes. When the callback fires while the person is still on the site, connection rates sit dramatically higher than when it fires four minutes later. The person has already opened another tab by minute four.
  • The second and third attempt do most of the heavy lifting. A large share of the enquiries we eventually connect with are not reached on attempt one. Human teams almost never get to attempt three. Software does, without sulking about it.
  • Evenings and weekends are where the gap lives. The single biggest lift we see for a plumbing or HVAC client isn't a faster daytime response. It's that the 8:40pm call gets answered at all.
  • Speed doesn't rescue a bad list. On a cold outbound campaign for an HVAC client we hit a 6% reply rate across 30 warmed inboxes. That came from targeting and copy. Responding faster to the wrong people just gets you rejected sooner.

Here's what most people get wrong

They treat speed to lead as a stopwatch problem when it's mostly a coverage problem.

Everyone benchmarks median response time, because it's the number the CRM gives you for free. Median response time is close to useless. It tells you how quickly you handle the leads you handle, and says nothing about the ones that died in a voicemail box at 9pm on a Saturday.

The number I'd actually track is the percentage of enquiries that received a real conversation within 24 hours, across every hour of the week. Run it for a month. Most owners are shocked by how much of their pipeline never got a single human interaction, let alone a fast one.

Fix coverage first. Then optimise the stopwatch.

What to do this week

Pull your last 100 enquiries. Tag each one with the hour it came in and whether anyone ever spoke to that person. Not "called" — spoke. If more than a quarter of them never got a conversation, your problem isn't a five-minute SLA. Your problem is that nobody was there.

That's the gap an AI receptionist is actually good at closing, and it's the honest reason the maths tends to work out. If you want to see the arithmetic on that, we ran it properly in our AI receptionist ROI breakdown, and there are more benchmarks in our 2026 statistics roundup.

FAQ

Is the speed-to-lead study from Harvard or MIT?

MIT. The 100x contact and 21x qualification figures come from Dr James Oldroyd's 2007 Lead Response Management study at MIT Sloan, run with InsideSales.com. The Harvard link is a separate 2011 Harvard Business Review article by the same lead author.

How old is the 5-minute rule?

The core data is from 2007, so it predates smartphones being the default browsing device, live chat widgets, and most modern lead sources. Directionally it still holds. The exact multipliers should not be quoted as current fact.

What is a good lead response time in 2026?

Under a minute for inbound web enquiries in competitive local services, and never longer than an hour. But response time only matters if the response is a conversation rather than a missed call.

Does responding in 5 minutes really make you 100x more likely to connect?

That was the 2007 finding for 5 minutes versus 30 minutes in a specific dataset. It is not a universal constant. Treat it as evidence that decay is steep, not as a number to put in a board pack.

Why do so many companies still respond slowly?

Because coverage costs money. Answering every enquiry inside a minute, at 2am, on a bank holiday, is a staffing problem before it's a process problem. That's the whole reason automated answering exists.

Does speed to lead matter for outbound as well as inbound?

Less. Outbound response speed matters at the reply stage, not the first-touch stage. If someone replies to a cold email, calling back the same hour is worth a lot.

How do I measure speed to lead properly?

Measure time to first conversation, segmented by hour of day and lead source, plus the share of enquiries that never got a conversation at all. Median time to first touch will flatter you.

Can an AI receptionist actually respond faster than a person?

On availability, yes, and that's the real difference. It answers the fourth simultaneous call and the Sunday-night call. A good human receptionist beats it on nuance during business hours.

Next step

Do the 100-enquiry audit. If the coverage gap is real, start a 15-day free trial and point your after-hours calls at it for two weeks. You'll have your own data instead of someone else's from 2007.

Sources