This article covers what the research actually says, why your current number is probably flattering, and how to fix the specific gap that is costing you money.
In short
- The research is real and it is old. The best-known study is from 2011. Use it honestly.
- Most speed-to-lead reports have the wrong denominator — they exclude leads nobody ever contacted.
- Report the median and the 90th percentile, never the average alone.
- There are four timestamps and three gaps. Fix the biggest gap, not the most visible one.
- "Contact" needs a definition before the metric means anything.
- Faster follow-up on bad data is just faster failure. Speed multiplies whatever your system already does.
What the research says, and what it doesn't
The study everyone half-quotes is "The Short Life of Online Sales Leads," published in Harvard Business Review in March 2011 by James B. Oldroyd, Kristina McElheran and David Elkington (HBR; citation record). It reports two pieces of work.
An audit. The authors audited 2,241 US companies, measuring how long each took to respond to a web-generated test lead. Their results: 37% responded within an hour, 16% within one to 24 hours, 24% took more than 24 hours, and 23% never responded at all. Among companies that responded within 30 days, the average response time was 42 hours.
An observational study. Separately, they examined 1.25 million sales leads received by 29 B2C and 13 B2B companies in the US. Firms that tried to contact a potential customer within an hour of receiving a query were nearly seven times as likely to qualify the lead — which they defined as having a meaningful conversation with a key decision maker — as those that tried even an hour later, and more than 60 times as likely as companies that waited 24 hours or longer.
The article also names the causes the authors observed: pulling leads from CRM databases daily rather than continuously, sales forces focused on generating their own leads instead of reacting to inbound interest, and lead distribution rules based on geography and fairness rather than speed. Fifteen years later, that list still describes most of the companies we see.
Now the honest part. This is 2011 data, gathered before the current messaging landscape existed. The second study is observational, which means it establishes association, not proof of cause — firms that respond in minutes are plausibly better run in other ways too. And "qualify the lead" is their definition, not yours.
None of that makes it useless. It makes it directional evidence, which is what you should treat it as. Anyone citing a precise modern-sounding percentage on lead response without naming a methodology is repeating a number that has been copied across blog posts until its origin disappeared. Do not become a link in that chain. This is the same discipline behind revenue engineering generally: measure your own process rather than importing someone else's benchmark.
Why your current number is flattering
Three failures, in order of how much damage they do.
1. The denominator excludes your worst cases. Most CRM reports compute average response time across leads that received a response. Leads nobody ever touched have no response timestamp, so they drop out of the calculation entirely. Given the HBR audit found 23% of companies never responded at all, this is not a rounding error — it is the whole problem hiding in a blind spot. Fix: make "percentage never contacted" a headline number of its own, sitting next to the timing metric.
2. The average hides the tail. Response times are skewed: a large cluster of fast responses and a long tail of slow ones. The mean bounces around; the median is stable. Worse, an average can improve while the tail gets longer. Fix: report median and 90th percentile together. The median tells you about the normal day. The 90th percentile tells you about the leads you are losing.
3. "Response" is undefined. Does an automated confirmation email count? A dialed call that rang out? A rep opening the record? Each produces a wildly different number from the same data. Fix: define three levels and track them separately — attempt (we tried), reach (they answered), conversation (a real exchange with the right person). The HBR study used a conversation-level definition, which is the honest one and also the hardest to hit.
The speed-to-lead clock: four timestamps, three gaps
Here is the MitHub framework we use to find out where the time actually goes. Most teams have one number. You need four timestamps, because the three gaps between them have completely different fixes.
| # | Timestamp | What it means |
|---|---|---|
| T1 | Lead created | The prospect submitted, called or opted in |
| T2 | Received by your system | The record exists in the CRM and is assignable |
| T3 | First attempt | Someone or something tried to make contact |
| T4 | First conversation | A real exchange with the right person happened |
| Gap | Name | Usual cause | Usual fix |
|---|---|---|---|
| T1→T2 | The plumbing gap | Batch imports, a broken integration, a form posting somewhere nobody watches | Event-driven intake |
| T2→T3 | The routing gap | No owner assigned, unclear ownership, rep not notified | Automatic routing with an escalation |
| T3→T4 | The persistence gap | One attempt and stop; wrong channel; wrong hours | A defined cadence across channels |
Measure all three before touching anything. In our experience the gap teams assume is the problem is rarely the biggest one. Everybody blames reps for T2→T3 while a nightly batch import is quietly burning eighteen hours in T1→T2 — a gap no amount of rep coaching will close.
This is the same "map the process with numbers" discipline as MitHub's Follow the money chapter, applied to a single stretch of the funnel. Put the result in a revenue leak map so the gap sizes are visible next to everything else that leaks.
Fixing each gap
T1 → T2: the plumbing gap
This should be near zero, and it is the cheapest to fix.
Move from pull to push. Instead of a scheduled job that asks "any new leads?", have the source system send the lead the moment it arrives. In n8n that means a webhook: the node gives you separate test and production URLs, and lets you respond to the caller immediately rather than after the whole workflow finishes (n8n Docs) — so the form never waits on a slow enrichment step.
Two traps. First, a source still pointing at a test URL loses leads silently every day you are not testing; document which URL each source uses. Second, webhooks fired during an outage are gone. Keep a scheduled sweep as a safety net that catches anything the real-time path dropped. See CRM automation with n8n for the pattern.
T2 → T3: the routing gap
A lead with no owner has no response time — it has an expiry date.
- Assign automatically, by territory, product, branch or round robin, with a named fallback owner when no rule matches.
- Make "no owner resolved" fail loudly. n8n's Stop and Error node lets you force a failure on conditions you choose, which then triggers your error workflow (n8n Docs). A lead falling through the rules should page someone, not sit invisibly.
- Notify where people actually look. A CRM task nobody opens is not a notification.
- Set an SLA with an escalation. Untouched after N minutes, it goes to a manager; after 2N, it reassigns and the miss is logged.
The mechanics are covered in lead routing.
T3 → T4: the persistence gap
One attempt is not follow-up. This gap is about cadence and channel, and it is where most of the remaining loss lives once the first two gaps are closed.
Decide in advance: how many attempts, over how many days, across which channels, in which order, and when you stop. Write it down, then automate it so it happens identically whether or not the rep is having a good week.
This is also where AI voice agents have become genuinely useful — not because a machine is more persuasive, but because it is available at minute two and at attempt six, which humans are not. See AI voice agents for sales and follow-up. Before you automate any outbound contact, confirm the consent, suppression and permitted-hours rules that apply to your market and channel; speed is not a defence.
Two ways speed-to-lead projects fail
Speed onto bad data. Calling within thirty seconds on a wrong number is a faster way to reach nobody. Speed multiplies whatever your system already does, including its mistakes. Fix contact-data quality alongside response time, or you will show a dramatically improved metric and an unchanged pipeline.
Speed without capacity. If reps already cannot work the leads they have, responding faster produces more untouched conversations, not more revenue. Check the constraint before you optimise the step in front of it — otherwise you have moved the queue, not shortened it.
Your first week
- Define contact. Attempt, reach, conversation. Write the definitions down and tell the team.
- Instrument the four timestamps. T1 and T2 usually exist already. T3 and T4 often need a field.
- Publish the honest baseline: median, 90th percentile, and percentage never contacted. Do not clean it up.
- Size the three gaps and pick the largest.
- Fix that one gap only, and watch the same three numbers for two weeks.
- Write up what changed with the before-and-after chart.
That last step is not admin. A documented before-and-after on a real process, with numbers you can defend and a method you can explain, is exactly the proof of work that gets someone hired into this field — and it is the shape of the case study at the end of MitHub's Faculty of Revenue Reverse Engineering.