Lead response time: what the research actually says

If you have read anything about sales enquiries in the last decade, you have met this sentence in some form: a Harvard study found that responding to a lead within five minutes makes you 100 times more likely to convert.

Almost every clause in it is wrong. It was not Harvard. The comparison was not five minutes against ten. The figure is not a conversion multiplier. And the study is now old enough to vote.

None of which means the underlying idea is false. It is not. But an argument built on a misquoted statistic collapses the moment somebody checks it, and the people you most want to persuade are usually the people who check. This article separates what the research established from what has been attached to it since.

The two studies that everyone merges into one

There are two distinct pieces of research behind the modern speed-to-lead conversation. They share a lead author, which is almost certainly why they have been welded together in the retelling. They measured different things, on different samples, four years apart.

What the 2007 MIT study actually measured

The Lead Response Management study was led by Dr James Oldroyd, then at the MIT Sloan School of Management, and produced in partnership with InsideSales.com. It examined three years of data covering six companies, more than 15,000 web-generated leads and over 100,000 call attempts. The analysis regressed response timing against outcomes using ordinary least squares with robust standard errors.

It asked a narrow and useful question: given a lead that arrives through a web form, how much does the delay before the first outbound call change your chances of reaching that person and of that conversation qualifying?

Three corrections follow immediately from that.

The comparator is 30 minutes, not 10. The widely circulated “five minutes versus ten minutes” version is a misquotation. The finding concerns the half hour after an enquiry arrives, which is a considerably less demanding standard than the folklore version implies, and therefore a more achievable one.

These are odds ratios, not revenue multipliers. “One hundred times more likely to make contact” is a statement about the probability of reaching a human being on the telephone. It is not a claim that you will close one hundred times more business. Repeating it as a conversion figure is the single most common distortion, and it is the one that makes experienced buyers stop listening.

The study did not measure close rates at all. It stopped at contact and qualification. Anyone citing it as evidence about revenue is extending it past where it goes.

The limitations worth stating out loud

We think a piece of research is more persuasive, not less, when its boundaries are described honestly. The 2007 study has several that matter.

• The sample is small and narrow. Six companies, weighted towards technology and financial services — sectors where web-form leads are well defined and outbound phone follow-up is the norm. A roofing contractor's enquiry mix does not obviously resemble a 2007 software vendor's.
• It measured the first call attempt, not the first answer. The clock ran from form submission to the moment a rep dialled. That is a different quantity from the one most businesses care about, which is when the customer actually heard from a person.
• It cannot fully separate speed from everything correlated with speed. Organisations that respond quickly also tend to have better-staffed desks, better CRM discipline and more motivated teams. The regression controls for some of this. It cannot control for all of it. Some of the measured effect belongs to being a well-run company rather than to speed as such.
• There is a commercial interest in the result. InsideSales.com sold lead-response software. A study co-produced by a vendor whose product is the implied remedy deserves to be read with that in mind. This does not make the finding wrong — the direction has been reproduced elsewhere — but it is a fact a reader is entitled to have, and most articles quoting the study omit it.
• It is from 2007. Before the smartphone was ubiquitous, before business messaging, before a buyer could reasonably expect a reply on WhatsApp at nine in the evening.

The 2011 HBR study is the more damning one, and nobody quotes it

Four years later, Oldroyd published “The Short Life of Online Sales Leads” in Harvard Business Review with Kristina McElheran and David Elkington. This is the Harvard connection, and it is a different study with different findings.

Rather than analysing one dataset of call logs, the researchers audited how quickly 2,241 US companies responded to a web enquiry. They also examined a much larger body of 1.25 million leads across 29 business-to-consumer and 13 business-to-business firms.

The paper also reported that firms responding within an hour were around seven times more likely to have a meaningful qualifying conversation than those responding an hour later. Note the size of that figure relative to the 2007 one. Seven, not one hundred. Different sample, different measurement, more conservative result — and much closer to what most businesses should plan around.

But the striking number is not the multiplier. It is the 23 per cent. Nearly a quarter of the companies audited did not reply to a genuine enquiry at all. Not slowly. Never.

That finding is more useful than the five-minute rule and it is almost never cited, probably because it does not sell anything. It reframes the problem. For a large share of businesses, the constraint is not the difference between five minutes and thirty. It is that a meaningful proportion of enquiries fall through the floor and nobody in the organisation knows it is happening.

Why the two studies got merged

The mechanism is easy to reconstruct. The same lead author produced both. The second appeared in a famous publication. The first contained the dramatic number. Over years of secondary citation, the dramatic number migrated to the famous publication, and the comparator drifted from thirty minutes to ten because ten makes a tidier headline.

The practical consequence is that a genuinely useful body of evidence now circulates in a form that a sceptical operations director can dismantle in about ninety seconds. If you are trying to make an internal case for changing how enquiries are handled, cite it correctly or do not cite it at all.

What has changed since 2007, and what has not

Assumption in the original research	Status in 2026
The response is an outbound telephone call	Weakened. For many buyers a text message, WhatsApp reply or web chat is a faster and more welcome first contact than a cold ring-back.
The enquiry arrives during working hours	Weakened. A large share of enquiries now arrive in evenings and at weekends, when the measured response gap is widest.
The buyer is contacting one supplier	Strengthened as a risk. Comparison sites and search make it trivial to submit the same enquiry to several suppliers in one sitting.
Attention decays quickly after submission	Holds. The buyer's intent is highest at the moment they press send, and it does not stay there.
Fast response requires more staff	Weakened. Acknowledgement, qualification and calendar booking can now be handled reliably without a person sitting by the phone.

The last row is the one that changes the economics. In 2007 the only way to answer an enquiry at 8pm on a Sunday was to pay somebody to be available at 8pm on a Sunday, which for most businesses was not worth it. That trade-off no longer holds in the same form, which is why the research has become more actionable with age rather than less.

What holds up regardless

Strip away the contested multipliers and a robust core remains.

The decay is real and it is front-loaded. Every serious measurement of this, across different samples and decades, finds the same shape: a steep fall in the first hour, then a long flat tail. The exact gradient varies by sector and channel. The shape does not.

The mechanism is obvious once stated. Someone submitting an enquiry is, at that moment, actively in-market, thinking about the problem, sitting at a device. Twenty minutes later they are doing something else. Two days later they have either forgotten or bought elsewhere. No statistic is needed to accept this; it describes ordinary human attention.

It is a relative race, not an absolute one. When a buyer contacts four suppliers, the first credible response takes a disproportionate share of the outcome. This is the part that matters commercially, and it is why response time behaves less like an efficiency metric and more like a competitive position.

Never-responded is the largest single leak. It does not show up in an average response time, because a lead that was never answered has no response time. It has to be counted separately or it stays invisible.

How to measure your own response time honestly

Most businesses that measure this measure it flatteringly. If you want a number you can act on, define it carefully.

• Fix the start of the clock. Form submitted, call first rang, message received. Not when it appeared in someone's CRM queue.
• Fix the end of the clock. The first response that a customer would recognise as a real reply. An automated “thanks, we've got your enquiry” is not a response; it is a receipt. Counting it is the most common way organisations flatter themselves here.
• Report the median and the 90th percentile. An average is dominated by the tail and hides it at the same time. The median tells you the normal case; the 90th percentile tells you what your worst-served customers experienced.
• Count non-responses separately. Every enquiry that received nothing at all, as a raw count and a percentage. Given the HBR finding, expect this to be higher than anyone internally believes.
• Segment by hour and by day. Aggregate figures conceal the failure mode. Most businesses are competent between ten and four on weekdays and absent the rest of the time, which is when a meaningful share of enquiries arrive.
• Segment by channel. Phone, web form, email, WhatsApp, portal. These usually differ by an order of magnitude, and the worst channel is rarely the one people assume.

Do this for one month of real enquiries before changing anything. The measurement frequently reorders the priority list on its own, and it protects you from buying a solution to a problem you have not confirmed you have. That principle is the same one set out in what a growing business should automate first.

What to do about it: process before technology

Once the measurement is honest, the fixes fall into a clear order, and the first ones cost nothing.

Decide who owns an enquiry. A surprising share of slow responses are not capacity problems but ownership problems: the enquiry arrives somewhere shared, everyone assumes someone else has it, and it ages. Named ownership with a visible clock resolves more delay than any software purchase.

Define what happens out of hours. Not aspirationally. Decide explicitly whether an enquiry arriving at 7pm on Friday is answered that evening, first thing Saturday, or Monday morning — and make the answer visible to the customer.

Separate acknowledgement from qualification. These are different jobs with different urgency. Acknowledgement should be immediate and should tell the customer something specific and true about what happens next. Qualification can follow, and benefits from being done properly rather than quickly.

Then automate the part that is genuinely mechanical. Answering at a predictable moment, capturing the same facts every time, checking a calendar and booking a slot are all rule-governed tasks that do not degrade when they happen at 3am. Judgement, pricing, negotiation and anything involving an unhappy customer are not, and should route to a person with the context already gathered. Where that line sits is the subject of what AI agents should and should not be allowed to automate.

The order matters. Automating an undefined process reliably produces a fast, consistent version of the wrong thing.

Common questions

Who did the study behind the five-minute rule?

Dr James Oldroyd, then at MIT Sloan, with InsideSales.com, published in 2007. Not Harvard Business Review, which published a separate 2011 study by the same lead author with different and more conservative figures.

Is it five minutes versus ten minutes?

No. Five minutes versus thirty. The ten-minute version is a misquotation that has been copied forward for years.

Does “100 times more likely” mean 100 times more sales?

No. It is an odds ratio for making telephone contact. The study did not examine close rates.

Is 2007 research still usable in 2026?

Treat the specific multipliers as dated and setting-specific. Treat the shape — sharp decay in the first hour — as sound, because it is consistent across later independent measurement and matches the obvious mechanism.

What is the single most useful number in this literature?

The 23 per cent of audited companies that never responded at all. It is the largest and most fixable failure, and unlike the five-minute rule it requires no statistical argument to act on.

The short version

Quote it properly or leave it alone. The MIT study established that the odds of reaching and qualifying a web enquiry fall steeply between five and thirty minutes, in a small sample, in 2007, measured to first dial, by researchers working with a vendor of response software. The HBR study established that a quarter of companies never reply at all and the average reply takes 42 hours.

Taken together they support a modest, defensible claim: responding quickly and reliably to enquiries is worth more than most businesses assume, and the largest available gain for most organisations is not shaving minutes off an already-good process but closing the gap where enquiries currently receive nothing.

That is a smaller claim than the folklore. It is also one you can take into a board meeting without it falling apart.

If you want to see how this applies to a specific operation, read about Lead Recovery, which includes a calculator for estimating the value of missed enquiries from your own figures, or the sector-specific version for estate agencies.