The Inbound Response Benchmark: What Actually Happens After You Submit a B2B Demo Request

By Ashish Aman · 27 August 2026 · 10 min read

The statistics the entire speed-to-lead category rests on are between nine and nineteen years old, and the most-quoted one has no traceable source. Nobody has re-measured. So Marklinea is running the Inbound Response Benchmark: we are submitting genuine demo requests to 100 B2B software companies and measuring two things — how long until someone replies, and how long until the buyer actually has an answer. Method below, published before the data. Results Thursday 1 October 2026, free to read.

Why re-measure something everyone already knows?

Because almost nobody has checked how old the evidence is.

Here is what the category actually rests on:

The 2007 study predates the iPhone. It predates live chat as a default, mobile-first browsing, marketing automation as it exists now, and every AI assistant a buyer might use to shortlist vendors before ever filling in a form.

And the most-quoted number in the entire category cannot be traced to anything. It is credited to McKinsey, Forrester, InsideSales and MIT depending on where you read it, the attributions contradict each other, and the McKinsey study does not appear to exist. We know this well because we had it on our own homepage. We removed it, and wrote about why.

Removing a bad number leaves a gap. This is us trying to fill it with a real one.

What the benchmark measures

Two clocks, started by the same event.

The gap between them is the thing nobody has ever published, and it is the number this whole exercise exists to produce.

We are also recording four things about the response itself:

1. What the first reply actually contained — an auto-acknowledgement, a calendar link, a templated sequence opener, or a real answer to what was asked.
2. How many touches before a substantive answer arrives, if one ever does.
3. Whether a substantive answer arrives at all within 10 business days. We expect this to be the most uncomfortable column, and HBR's 2011 finding that 23% never respond is the figure it should be read against.
4. Whether the response is human or automated, and whether that is disclosed.

Every request includes one specific, easily answerable product question. Not a trick — the kind of thing a real evaluator asks, answerable in two sentences by anyone who knows the product. The point is to find out whether a question asked at the front door ever reaches somebody who can answer it.

How we are running it, including the part we are uneasy about

Method first, and in full, because a benchmark whose method appears after its findings is a benchmark whose method was chosen to fit them.

The sample. 100 B2B software companies with a public demo request or contact form. Mixed by company size and by how the sale is run — product-led and sales-led both. The list is fixed before any submission goes out and does not change afterwards.

The submission. A real, working, identifiable email address. A real company name — ours. A genuine product question. Nothing invented, no fake persona.

And here is the uncomfortable part. We are not telling companies in advance that they are being measured. If we did, we would be measuring how quickly a company responds when it knows it is being timed, which is not a number anyone needs.

We are not comfortable enough with that to skip past it, so here is exactly what we are doing to keep the imposition as small as we can make it:

• No meetings are taken. Any booking offer is declined immediately, so nobody's calendar is occupied.
• Any sequence is unsubscribed from as soon as the measurement window closes.
• No company is named in the results, ever. This is a category picture, not a league table, and we have no interest in embarrassing anyone's SDR team.
• If any company asks what this was, we tell them, and we will remove their data on request.

Reasonable people will disagree about whether that is a fair trade. We think publishing the reasoning openly is better than either hiding the method or quietly measuring the wrong thing. If you think we have got that call wrong, tell us — that argument is worth having in public and we will publish it.

The second arm. Companies can volunteer their own funnel to be measured, with full access to their own instrumentation, which lets us measure things a blind submission cannot see — internal routing time, how long a lead sits before assignment, what happens overnight. These results are reported separately and never blended with the blind sample, because a company confident enough to volunteer is unlikely to be a slow one and mixing the two would flatter the average.

What we expect to find, written down in advance

Stating expectations before the data means we cannot claim afterwards that the result is what we thought all along. Some of these will be wrong. That is the point of writing them down.

• First response times have improved substantially since 2011. Fifteen years of marketing automation, instant chat and SLA tooling did something. We expect the 42-hour figure to look badly dated.
• Time to resolution has barely moved. Automation made the acknowledgement faster. It did not make the answer faster, because the answer still requires a person who knows the product to be available. This is the core hypothesis.
• The gap between the two clocks is measured in days, and it is wider at larger companies with more routing between the form and a human.
• A meaningful share never produce a substantive answer at all, and the number will be higher than anyone finds comfortable.
• Product-led companies resolve faster than sales-led ones, not because their people are better, but because they have fewer handoffs between the form and someone who can answer.

If clock 2 turns out to be fast across the board, we will publish that, and it will undermine a fair amount of what we have argued this month. That risk is what makes the exercise worth anything.

How to measure your own funnel this week

You do not need to wait for us, and you do not need any tooling. Twenty enquiries and an hour.

1. Take your last 20 inbound enquiries. Not a flattering sample — the last 20.
2. Record the enquiry timestamp.
3. Record when a human first responded. That is your speed to lead.
4. Record when the buyer's actual question was first answered. A meeting invitation does not count. A brochure does not count. That is your speed to resolution.
5. Subtract, and average both. The difference between the two averages is your gap.
6. Count how many never reached step 4 at all. This is usually the largest group and the most uncomfortable.

Send us your two numbers and your rough company size, and we will include them anonymously in the results and send you the full dataset. No company names, no attribution, no sales follow-up — if you send a number and nothing else, you get the results and nothing else.

What happens to the results

Published Thursday 1 October 2026, free, no email required to read them.

The full dataset goes up alongside the write-up — every timestamp, every classification, the complete company list by category and size but never by name, and the ones we had to throw out and why. If our conclusions are wrong, we would rather you be able to prove it than take our word.

That is the standard the numbers in the table at the top of this page did not meet. It seems like a low bar to hold ourselves to, given we spent a month pointing at them.

Common questions