Talk-to-Listen Ratio: What the Numbers Support, and What They Do Not

Somebody in your organisation is about to tell a rep to listen more. It is good advice with nothing attached to it, so the rep nods and changes nothing. The talk-to-listen ratio exists to put a figure on that instruction, and it half succeeds: it is the rare call behaviour you can count without an argument, and it is also easy to read as a verdict when it is barely a hint.

This is for managers running call reviews who want to know what the metric genuinely supports, what the two best-known datasets actually said, and which companion measure does more work.

The Definition, and What It Leaves Out

The ratio is simply the share of a recorded conversation in which the seller is speaking, set against the share in which the buyer is. Conversation intelligence tools produce it automatically by timing each speaker, so a call reported as sixty to forty had the seller holding the floor for three fifths of it.

What it cannot see is everything that matters about those minutes. It does not know whether your talking answered a question or filled a silence, whether the buyer's share was a considered explanation or a polite reply, or whether anybody agreed a next step. It is a shape, not a judgement, and the reason managers reach for it anyway is that it requires no interpretation to produce. That is a real virtue and a real limit at the same time.

Two Studies, Nine Years Apart, Different Answers

Most of what circulates on this subject traces back to two analyses from the same vendor, and they do not agree with each other. The version written up by Dan Morgese sets both out.

The earlier one, from 2016, produced the figure people still quote. It identified a "golden ratio" of 43% talking to 57% listening, and reported that speaking for more than 65% of a call was associated with weaker conversion and win rates. The write-up does not describe that study's sample or method, which is worth knowing before anybody builds a coaching standard on it.

The 2025 re-analysis covered 326,000 calls of at least ten minutes and landed somewhere else entirely. The average call ran at 60% seller talk time, with won deals at 57% and lost deals at 62%. Note what that means: today's winners talk for more than half the call, and the separation between winning and losing is roughly five points. Morgese says so himself, observing that the difference between the two "isn't massive".

The more interesting number in that update is about consistency rather than level. Weaker sellers went from 54% talk time on deals they won to 64% on deals they lost, while the strongest ones held roughly the same share either way. The ratio is not describing a skill so much as describing composure. Talking more is what a deal going badly feels like from the inside, and the measurement catches the symptom.

Why Sellers Drift Over the Line

1. Silence reads as failure. Two seconds of nothing feels like a problem to the person being evaluated and feels ordinary to the person thinking.
2. The deck exists and wants to be used. Preparation converts into airtime unless somebody decides it should not.
3. The questions were closed. A yes-or-no question returns a word, and the rep fills the rest of the minute. Our discovery framework built on three problems rather than fifty questions is the fix for that one.

Counting Questions Is a Different Thing Entirely

Asking more is not the same as listening better, and the same dataset makes the point bluntly. Calls that ended in a win carried roughly fifteen to sixteen questions, while lost ones averaged around twenty, which inverts the instinct that a questioning rep is a curious one. Past some threshold, questions stop being interest and start being a form.

The academic literature has been pointing in the same direction for a long time. Stephen Castleberry, David Shepherd and Rick Ridnour built a measure of interpersonal listening for selling and tested it on "a sample of 604 salespeople from a variety of firms and industries", finding that scores on the scale correlated with both performance and experience. That study dates from 1999 and relies on self-reported listening, so it proves nothing about causation, and we are citing it because nothing with a comparable sample has replaced it since. What it establishes is narrower and still useful: listening in selling has been treated as a measurable construct for decades, and nobody serious has ever defined it as the inverse of talking.

A 2019 analysis of 6,200 recorded dealership calls, written up for F and I Magazine, gives a picture of what the measurable version looks like in practice. Marchex found its top quartile distinguished by repeating and clarifying what the caller had said, asking about specific needs, and consistently getting the caller's name. Those are behaviours an observer can tick off. The setting is consumer car buying rather than B2B, so do not import the findings, but do import the method: name the behaviours, then count those.

What to Measure Alongside the Percentage

The single most useful companion is the longest unbroken stretch of seller speech in the call. A conversation split evenly down the middle that contains one three-minute lecture is a lecture, and the ratio will tell you it was balanced.

There is a benchmark for the hard moment. Chris Orlob's analysis of 67,149 recorded demos found the typical seller responds to an objection with an unbroken reply of 21.45 seconds. That one measurement says more about how somebody handles pressure than a whole-call average ever will, and we work through what to do about it in our piece on handling objections. If your recording tool reports longest turn, pull it. If it does not, a stopwatch across five calls gets you there in twenty minutes.

Never Give a Rep a Number to Hit

Here is where we part from most of the coaching advice built on this metric. A talk-to-listen target handed to an individual rep is worse than no metric at all, and the reason is not philosophical. The separation between winning and losing calls in the best available dataset is around five points, which is comfortably inside the range a rep can move by changing nothing except how often they say yes while the buyer speaks. Give somebody a number and you will get the number. You will not get the behaviour, and you will have spent your credibility buying a statistic.

Use it as a trigger for a conversation instead. A rep sitting far from the group on a particular call type is a prompt to go and listen to one of those calls, which is a different managerial act from reporting a figure in a review. That distinction is the same one we draw in coaching behaviour rather than inspecting deals, and it belongs in the weekly slot described in our thirty-minute one-to-one system.

The missing piece is a dataset that separates cause from effect. Engaged buyers talk more because they are engaged, so a seller's lower share may be a consequence of a good deal rather than a cause of one, and none of the published analyses can pull those apart. If somebody ran a controlled study where reps were instructed to shift their share and win rates moved, a target would become defensible. Until then the honest position is that the ratio detects, and only coaching changes anything.

An Hour of Call Review, Start to Finish

Take three recent discovery calls from one rep, never a single call, and compare like with like rather than setting a demo beside a first conversation. Note the seller's share and the longest uninterrupted turn in each. Ask the rep to find one moment where stopping would have helped, and agree one change. Review three fresh calls a fortnight later. The ratio did not improve anybody in that loop. It told you which hour to spend, which is all a blunt measure should be asked to do. On a cold call the opening line decides more than any of this, as our breakdown of why bad-time openers backfire sets out.