Written for sales and marketing operations leaders who own web lead follow up: what the classic research really says, how old it is, and the Five9 architecture that gets a first dial out in minutes.
How fast should you call a web lead?
Call within five minutes, and sooner if you can. Oldroyd's 2007 lead response study found contact odds collapse between five and thirty minutes, and a 2011 Harvard Business Review audit found most companies taking hours or days. Those studies are old and their multipliers should not be quoted as gospel, but the direction still holds, and phone screening has made the fast first attempt matter more, not less.
The 2007 Lead Response Management study, with its provenance attached.
Nearly every speed to lead statistic in circulation traces back to two documents, and the most famous number of all is this: call inside five minutes and you are 100 times more likely to reach the lead than at thirty. The number is real and it has a specific home, which almost nobody who quotes it has read. In 2007 the software firm InsideSales.com engaged Dr. James Oldroyd, then a faculty fellow at MIT's Sloan School of Management, to work out when web generated leads should be called. The findings were presented at MarketingSherpa's B2B Demand Generation Summit in October 2007 and circulated as the Lead Response Management study.
It came in two parts. Part one was a survey of 495 companies, run while Oldroyd was finishing his PhD at Kellogg, and it mainly proved that nobody knew the answer: respondents agreed faster was better and could not say when or by how much. Part two is the piece everyone remembers: a behavioral analysis of three years of call data from six companies on the InsideSales.com platform, covering more than 15,000 leads and over 100,000 call attempts. The headline findings:
Read the provenance before you tattoo 100x on the sales floor. Six companies. Data drawn from the vendor's own platform, and the vendor sold software whose pitch was calling leads faster. Odds of contact and qualification, not revenue. Oldroyd himself noted that the clean patterns only emerge when the companies are pooled. None of that makes the study wrong. It makes it a directional finding from a small sample, published by an interested party, nineteen years ago.
2,241 companies mystery shopped by Harvard Business Review, most of them slow, some of them silent.
Four years later Oldroyd returned with co-authors Kristina McElheran and David Elkington in the March 2011 issue of Harvard Business Review, under the title The Short Life of Online Sales Leads. This is the second pillar of the canon, and the more uncomfortable one, because it measured what companies actually do rather than what happens once they dial.
The authors audited 2,241 US companies by submitting a test lead through each company's own web form and timing the response. 37 percent responded within an hour, 16 percent took between one and 24 hours, 24 percent took more than a day, and 23 percent never responded at all. Among firms that did respond inside 30 days, the average response time was 42 hours.
A companion dataset of 1.25 million leads received by 29 B2C and 13 B2B US companies supplied the multipliers this article is remembered for: firms attempting contact within an hour of the query were nearly seven times as likely to qualify the lead as firms attempting an hour later, and more than 60 times as likely as firms that waited 24 hours or longer.
The diagnosis aged best. Why so slow? Leads retrieved from CRM databases in daily batches rather than continuously. Sales teams rewarded for self generated pipeline, not for reacting to inbound interest. Routing rules designed around territory fairness rather than response time. We audit dialer estates for a living, and in 2026 we still find all three, frequently in businesses spending six figures a month generating the leads they then sit on.
Webhook to list to live campaign, with priority flags set, dedupe checked and alerts on silent failure. We are a Five9 partner and this is a build measured in days, not quarters.
The Lead Response Management study is nineteen years old. The HBR audit is fifteen. Both predate smartphone call screening, carrier spam labeling, and most of the lead aggregator market as it now operates. Quoting 100x as though it were measured last quarter is how marketing loses credibility with finance, and it deserves to.
The honest reading is that the multipliers are stale and the mechanism is not. Speed works for structural reasons that have not moved:
The honest answer on newer evidence: there is no modern replication with provenance as clean as the originals, and the vendor studies published since tend to repeat the direction with worse methodology and a product to sell. What we can offer instead is operational experience. On floors we run, leads dialed within a few minutes of arrival connect at a visible multiple of leads dialed after lunch. We will not put a number on that multiple, because it moves with lead source, caller ID health and time of day. The direction has never once reversed on us.
Spam labels and handset screening have raised the price of a slow callback.
Two developments since 2011 have made the slow callback strictly worse, and both live in the phone itself.
First, carrier analytics. US voice providers and third party analytics firms now block or label billions of calls a year, painting Spam Likely or similar onto caller ID before anyone decides whether to answer. The FCC's 2020 call blocking report records the collateral damage: one industry survey cited in it claimed 21 percent of lawful calls were being blocked, and another reported 45 percent of calling numbers rated severe or high risk even where the caller had consent to be called. That report is itself six years old now, and the machinery it describes has not gone away. A number that dials cold lists all day inherits a reputation, and the day two callback arrives wearing it.
Second, handset screening. Apple documents iPhone settings that silence unknown callers straight to voicemail, and a screening mode that asks an unknown caller to state a reason for calling before the phone rings at all. Filtered call lists tuck unknown voicemails away where they may never be seen. None of this is exotic. It ships in the standard settings of the phone in your lead's pocket.
Put those together and the evidence needs restating for 2026. In the 2007 data, a thirty minute delay cost you contact odds. Now a next day callback often costs you the ability to be perceived as a caller at all, because you have moved from expected call to unknown number, and the handset is actively working against the latter. The first attempt, made while your brand is the most recent thing the lead typed into a form, is the one call most likely to be answered on purpose. First attempt timing matters more today than when the studies were written, not less. And if your numbers already carry labels, deal with the spam label problem first, because speed cannot outrun a poisoned caller ID.
The urgency is front loaded. Minutes matter for the first attempt, then the curve flattens fast: the gap between one minute and thirty is enormous, the gap between thirty and ninety is real but smaller, and by day two you are arguing over scraps. The 2007 study had a sharp version of this: once roughly 20 hours had passed, each additional dial actually hurt the odds of ever contacting and qualifying the lead. Hammering stale records does not just waste agent time, it burns caller ID reputation you will want later.
The cadence we run for high intent web leads, presented as operator practice rather than laboratory fact:
| Window | Attempts | Reasoning |
|---|---|---|
| First 5 minutes | Attempt 1 | The expected call. The highest connect odds you will ever have on this record. |
| Same shift | Attempt 2 | Different hour, same day. Catches the school pickup and the meeting that overran. |
| Days 1 to 3 | Attempts 3 to 6 | Front loaded and rotating day parts: morning, midday, early evening within permitted hours. |
| Days 4 to 7 | One attempt per day, tapering | Returns are thin here. Text or email between calls where you hold consent for the channel. |
| After day 7 | Move to a nurture list | Stop dialing daily. Re-approach on a schedule with fresh day parts and clean numbers. |
Two cautions. Cadence intensity is regulated in parts of the US market: frequency caps and quiet hour windows exist in some states and sectors, so the table above assumes you have checked what applies to your book. Rules in this area keep moving and enforcement follows its own schedule, so verify against the current texts and take advice from qualified counsel where the stakes justify it; our piece on lead generation consent rules covers the consent side. And cadence never rescues a slow first attempt. A perfect eight touch sequence started four hours late is a well organized way to call a cold record.
A webhook into the dialer beats a scheduled import every time.
Most teams we meet say they call leads quickly. The architecture usually disagrees. The standard estate looks like this: forms write into the CRM, an integration or an intern exports on a schedule, and a CSV lands in the dialer at ten and two. Median speed to lead: hours, dressed up by whichever leads happened to arrive just before an import ran. If your leads travel by CSV, you do not have a speed to lead program, you have a batch job with ambitions.
The pattern that actually delivers minutes is boring and has been buildable for years. The form, or the aggregator's delivery endpoint, fires a webhook at the moment of submission. That webhook writes the record straight into a dialing list on a campaign that is already running and already staffed, flagged for immediate treatment rather than joining the back of the queue. On Five9 this is native: the Web2Campaign API accepts an HTTP POST that adds a record to a chosen list, with a flag (F9CallASAP) to prioritize it for dialing at once. No middleware platform is required, though one can help with field mapping and retries.
The build takes days. Keeping it honest takes discipline, because the failure modes are quiet:
We insist on that last point from scar tissue: a write that fails silently is the worst outcome in any pipeline, so insert volume gets an alert, not a dashboard someone might remember to look at. The full architecture, including out of hours handling, is laid out on our speed to lead page.
Before citing any lead response time study in a meeting, run the 2011 audit on yourself. Submit a lead through your own form on a Tuesday at 11am. Do it again on Saturday at 9pm. Time the first human attempt. Most leadership teams have never done this, and most are unpleasantly surprised, which is precisely the result the HBR authors got out of 2,241 companies.
Then instrument it properly. The definition worth using runs from the timestamp the lead was created at source, meaning the form submission or the moment an aggregator posted the record to you, to the timestamp of the first outbound dial attempt against that record. Not first activity, not CRM task creation, and not the time the record reached the dialer, which conveniently excludes the batch delay you are trying to expose.
Report the distribution, never the average. One weekend of unworked leads poisons a mean, and a respectable average can hide a rotten tail. Median and 90th percentile tell the truth: a floor with a 3 minute median and a 9 hour p90 does not have a speed problem, it has a night and weekend problem, which is a staffing and expectation decision rather than a plumbing one. Segment by lead source and by hour of arrival, because an aggregator feed and a white paper form deserve different targets.
Speed to lead measures the race to the phone. What happens after the dial, meaning connects, contacts and conversations per hour, is its own discipline, and our outbound benchmarks piece covers where those numbers tend to land.
Operating targets from floors we run, not laws of physics.
Targets should follow the economics of the lead, not a slogan. These are the operating targets we set on floors we manage, offered as experience rather than research:
| Lead type | First attempt target | Why |
|---|---|---|
| Shared or aggregator leads | Under 1 minute | The record was sold to several buyers. It is a race, and second place is voicemail. |
| Exclusive quote requests (paid search, landing pages) | Under 5 minutes | High intent and nobody else is dialing them yet. Minutes are cheap insurance. |
| Call back requests | Under 5 minutes | The lead literally asked for a call. Every elapsed minute reads as indifference. |
| Content downloads, webinar signups | Same business day | Low immediate intent. A sub minute response buys little here, so save the urgency budget. |
| Out of hours arrivals | First minutes of next opening | Acknowledge instantly by email or text where you hold consent, then dial at open. |
| Aged and re-engagement records | Timing barely matters | Day part rotation and caller ID health decide these, not response speed. |
The 100x figure is old enough to vote, and it should be retired from slide decks. What replaces it is not a newer statistic. It is your own decay curve, measured by source, and an architecture that gets attempt one out while the lead still remembers your name. That is a plumbing project and a staffing decision, and both are very fixable.
The direction is valid; the exact multipliers are not current. The five minute finding comes from a 2007 study of six companies' call data, so quoting 100x as a live number is not defensible. But intent decay, competition on shared leads, and modern call screening all still reward a first attempt within minutes, and our floor data consistently shows leads dialed within minutes connecting at a multiple of leads dialed hours later.
From the 2007 Lead Response Management study by Dr. James Oldroyd, then at MIT, using InsideSales.com platform data: three years of calls from six companies, more than 15,000 leads and over 100,000 call attempts. It found the odds of contacting a lead called at five minutes versus thirty minutes dropped 100 times, and qualification odds dropped 21 times. It is a real finding with real limits: a small sample, vendor data, and nineteen years of age.
In March 2011, HBR published an audit of 2,241 US companies that were each sent a web test lead. Only 37 percent responded within an hour, 23 percent never responded at all, and the average response among firms that answered within 30 days was 42 hours. A companion dataset of 1.25 million leads found firms attempting contact within an hour were nearly seven times as likely to qualify the lead as firms an hour slower.
As operating targets: under one minute for shared or aggregator leads, because several buyers hold the same record; under five minutes for exclusive quote requests and call back requests; same business day for low intent content leads. These come from floors we run rather than published research, and the right target for your operation depends on lead economics, staffing hours, and what your own connect data shows at each response time.
No. Averages hide the tail that does the damage. A handful of unworked weekend leads can drag a mean into hours while most leads get called in minutes, or a respectable mean can conceal a rotten 90th percentile. Measure from lead creation at source to first dial attempt, then report the median and the 90th percentile, segmented by lead source and hour of arrival. The distribution tells you whether you have a plumbing problem or a staffing problem.
Use a webhook rather than scheduled imports. Five9's Web2Campaign API accepts an HTTP POST that writes a record directly into a chosen dialing list the moment a form is submitted, and the F9CallASAP flag marks it for immediate dialing rather than joining the back of the queue. The destination campaign must be running and staffed when the record lands, and insert volume should be alerted on so a silent failure cannot go unnoticed for days.
Front load the attempts: one or two in the first hour, a handful more across the first three days rotating day parts, then taper through the first week before the record moves to a nurture schedule. The 2007 research found that once a lead had gone stale, at roughly the 20 hour mark, additional dials actively hurt qualification odds. Frequency caps exist in some US states and sectors, so check the rules that govern your book before fixing a cadence.
Occasionally someone is startled that you called before they had put the phone down. In our experience they still take the call, and the connect and qualification numbers comfortably outweigh the odd surprised prospect. What actually reads badly is the reverse: a callback two days later from a number their handset labels as spam. If sub minute feels too aggressive for your brand, an agent opening that references the form softens it immediately.
Score your operation in the free 17-question Five9 Health Check. Five minutes, no sign-up hoops, instant results.
We run outbound floors on Five9 daily, and we can usually tell within one conversation whether your speed to lead problem is plumbing, staffing or measurement. Bring your median and your p90, or let us help you find them.
Talk to us