Lead response time statistics from every major study from 2007 to 2026

Lead response time statistics get repeated across the internet with the wrong source attached. Sample sizes get inflated. The "5-minute rule" gets credited to a paper that does not contain it. Studies that no longer have a recoverable primary source get cited as primary. The result is that the same handful of figures bounce around B2B marketing decks with very little of the underlying rigor still attached.

This page is the table we wanted to find when we audited the underlying research for our speed-to-lead in the agentic era post. Every major lead response time study from Oldroyd's 2007 InsideSales paper through the 2026 vertical benchmarks, with the actual sample sizes, the primary URLs where they exist, and explicit flags on the rows where the primary source is unrecoverable. Where a widely-quoted figure does not hold up, we say so. The misattribution callouts under the table are the part most teams citing these stats do not realise they need.

Every major lead response time study, 2007 to 2026

Study	Year	Sample size	Key finding	Industry	Source
Oldroyd / InsideSales Lead Response Management Study [2]	2007	~15,000 web leads at 6 companies over 3 years	100x more likely to make contact within 5 minutes vs 30 minutes; 21x more likely to qualify within 5 minutes	Cross-industry B2B	Secondary (PDF on HubSpot CDN; original leadresponsemanagement.org retired)
HBR "The Short Life of Online Sales Leads" [1]	2011	1.25 million leads at 42 U.S. companies (29 B2C, 13 B2B)	7x more likely to qualify within 1 hour vs longer; 60x more vs 24 hours	Mixed B2C and B2B	Primary (paywalled but verifiable)
Drift State of Conversational Marketing [4]	2018	433 B2B companies tested via manual form submission	Average first-response: 47 hours; only 7% replied within 5 minutes	B2B SaaS	Secondary (original Drift report retired post-Salesloft acquisition)
Chili Piper Average Vendor Response Time Benchmark [3]	2022	Hundreds of B2B vendors tested via manual form submission	Average first-response: 4 hours 50 minutes; 7% sub-60-second; 30% never respond	B2B SaaS	Primary
RevenueHero B2B Response Time Study [4]	2024	1,000 B2B companies	Average first-response (responders only): 1 day, 5 hours, 17 minutes; 63.5% never respond	Cross-industry B2B	Secondary (via Apten aggregation)
Hatch HVAC Speed-to-Lead Analysis [4]	2024	132,000 home-services campaigns	88% take more than 5 minutes to reply; only 3% sub-60-second; 37% modal response time of 1 day	Home services	Secondary (via Apten aggregation)
Hennessey Digital Legal-Vertical Speed Study [4]	2025	Law firms (sample size not disclosed)	Median first-response: 13 minutes; 26% never respond; 25% under 5 minutes	Legal	Secondary (via Apten aggregation)
Apten / Blazeo 2026 Speed-to-Lead Benchmark Report [4]	2026	573 service-industry companies across 6 verticals	AI-assisted teams meet <15-minute standard 62.5% of the time vs 39.1% for manual teams; 81.2% lead leakage above 1 hour	Service B2B	Primary
Artificial Analysis frontier-LLM latency benchmark [5]	2026	Live testing of Claude Sonnet 4.6 across multiple providers	Time-to-first-token: 1.37 seconds; output speed: 44.3 tokens per second; sub-10-second 300-token response feasible	LLM infrastructure	Primary

The last row is not a lead response study. It sets the technical floor that an agentic SDR can actually hit on the same curve, which is the part the 2011 paper could not have anticipated.

Common misattributions to ignore

The following figures circulate widely. Several B2B vendor pages and aggregator blogs publish them as primary research. They do not hold up.

The HBR 2011 study did not survey "2,241 companies." The HBR study sampled 1.25 million leads at 42 American companies (29 B2C, 13 B2B). The "2,241 companies" figure that appears in several recent comparison tables likely comes from a separate, later InsideSales report that we were unable to locate the primary source for. The two studies should not be conflated.

The "5-minute rule" is not in HBR 2011. The 5-minute cliff, the 100x contact-rate multiplier, and the 21x qualification multiplier all come from the earlier 2007 Oldroyd / InsideSales Lead Response Management Study, not from the 2011 HBR piece. The HBR piece anchors the 1-hour and 24-hour thresholds. Both are real findings. They come from two different papers with two different sample sizes.

The Oldroyd 2007 study did not cover "100+ companies." The original 2007 paper covered approximately 15,000 web leads and 100,000 call attempts across 6 companies over 3 years. The "100+ companies" figure that gets attached to the 2007 paper in some secondary sources belongs to a separate replication study. Most articles citing the 2007 paper conflate these.

Velocify's "391% conversion lift in under 1 minute" has no recoverable primary source. Velocify published the figure in a 2016 whitepaper, but Velocify was acquired and rolled into ICE Mortgage Technology, and the original document is no longer hosted. Every contemporary citation traces back to the same chart with no primary URL. Treat the figure as widely-repeated industry folklore rather than primary research. Apten's 2026 leakage data and Chili Piper's 2022 benchmark cover the same conversion-lift dimension with traceable sources.

"78% of buyers purchase from the first vendor to respond" (Lead Connect, 2020) appears in several decks but is a single-source vendor-published statistic without a published methodology. Conflict-of-interest flag: Lead Connect sells the response-time product the statistic justifies. Use with caution, and never as the load-bearing claim.

"85% of callers will not call back after a missed call" (Ringba / CallRail). Widely repeated, especially in call-tracking marketing, but we could not find a primary methodology or sample size for it. Treat as folklore.

When you see any of these in a competitor's comparison table, you are looking at a piece of content that did not audit its own sources.

What the studies actually agree on

Across the durable, well-sourced studies (rows 1, 2, 3, 4, 6, 7, 8 in the table), four findings hold up consistently from 2007 through 2026:

1. Response-time decay is steep and starts within the first hour. Both Oldroyd 2007 (minute-by-minute) and HBR 2011 (hour-by-day) show the curve. Apten 2026 confirms 81.2% lead leakage above the 1-hour mark.
2. The gap between contact rate and qualification rate matters. Faster response disproportionately improves the chance of reaching a decision-maker, not just leaving a voicemail. HBR's 7x figure is about qualification, not contact alone.
3. Most companies still fail the bar. Industry-median response time has been measured in hours-to-days every year since the original studies. The bar has held; vendor performance against the bar has not.
4. First-responder advantage is real and economically meaningful. Vendors that consistently win the first-touch race convert at meaningfully higher rates per Chili Piper 2022 and the Apten 2026 leakage data.

The findings that do not replicate well across studies (the 391% Velocify number, the 78% Lead Connect number, the 85% Ringba number) are precisely the ones that came from single-vendor sources without published methodology. The pattern is consistent. Single-vendor stats inflate. Multi-source academic / benchmark stats hold.

What changed between 2007 and 2026

Three things, and only one of them is in the direction the early researchers would have predicted.

Industry-median response time has gotten worse, not better. The 2011 finding was supposed to motivate teams to compress their first-response window. Across the well-sourced 2018, 2022, 2024, and 2026 benchmarks, the median has stayed in the multi-hour range, and the share of companies that never respond has actually grown (Drift 2018: ~38% non-respond; RevenueHero 2024: 63.5% non-respond).

Industry-best response time has gotten dramatically better, for the small share of vendors that automated first touch. Apten 2026 shows AI-assisted teams clearing the under-15-minute standard 62.5% of the time vs 39.1% for manual teams. That is the bifurcation: a small leading edge has compressed the curve to seconds, and a large long tail has not moved.

The technical floor has moved from minutes to seconds. A frontier LLM in 2026 has first-token latency around 1.4 seconds and output speed around 44 tokens per second per the Artificial Analysis benchmark [5]. An end-to-end agentic SDR (webhook ingest, dedup, KB retrieval, brand-voice composition, send) clears 30 seconds with margin. Thirty seconds is the practical SLA, not the floor of what is possible.

What this means for inbound response in the agentic era

The 2011 paper anchored its argument at the 1-hour mark because that was the realistic edge case for a well-run human sales team. In 2026, the relevant edge case for a well-built AI inbound SDR is the 30-second mark, and the underlying curve has not changed. Faster response, grounded in context, still drives a multiplicative improvement in qualification rate. The leverage point moved. The leverage did not.

We argued the longer version of that case in speed-to-lead in the agentic era, including the closed-loop attribution argument that the 2011 paper could not have seen because the ad-platform Conversions APIs did not exist yet.

The two sister posts cover the rest of the agentic-inbound-SDR thesis. Signal-reactive nurture makes the case for what happens after the first touch, and why drip cadences are now an anti-pattern. FAINT plus GPCT versus BANT covers the qualification framework that fits an agent with infinite patience.

Related reading

• Speed-to-lead in the agentic era for the full thesis on why the 2011 curve compounds with closed-loop attribution
• Signal-reactive nurture vs drip cadences for what happens after first touch
• FAINT plus GPCT, the AI-era qualification frameworks for the qualification questions an AI inbound SDR should be running through

If any of the figures in the table above turn out to have a better primary source than the one we found, or if a 2026 study supersedes one of the older rows, let us know and we will update this page. The point of the table is for the rigor to be auditable.