Ask a room of sales managers what separates their best reps and someone will say "they listen more". Ask how much more and the room goes quiet. Talk-to-listen ratio is the metric that turns that instinct into a number, and it is the most common single metric in conversation analytics for good reason: it is simple, it is measurable from any transcript, and it correlates with behaviours that good salespeople recognise. It is also widely misused. This guide explains what the ratio measures, where the 40-to-50-percent target comes from, what else belongs on a call scorecard, a fully published scoring formula you can check, and how to coach with the numbers without coaching the humanity out of your team.
Quick answer: Talk-to-listen ratio is the share of a call's speaking time taken by the host. Most coaching models target 40 to 50 percent, so the prospect speaks at least as much as the rep. Measure it from a transcript with speaker labels, pair it with questions asked, filler words and monologue length, and read every score alongside the recording. PepoSmart scores each of those four components out of 25 and publishes every threshold so the score can be verified rather than trusted.
Key takeaways
- The ratio measures balance, not quality. A rep at 20 percent talk time is not automatically better than one at 45.
- The 40-to-50-percent target is a coaching heuristic supported by conversation analytics, not a law. Discovery calls run lower; demos and proposals run higher.
- One metric invites gaming. Four metrics, read together with the recording, do not.
- Publish the formula. A score a rep cannot check is a score a rep will not trust.
- Coach one behaviour at a time, over weeks. Trend lines matter more than any single call.
What talk-to-listen ratio actually measures
The ratio is the host's share of the total speaking time on a call, usually expressed as a percentage. A 45 percent talk ratio means the rep spoke for 45 percent of the time that anyone was speaking, and the prospect for 55 percent.
It is calculated from a transcript with speaker labels. Each speaker's words or seconds are summed, and the host's total is divided by the grand total. Silence is excluded. Overlapping speech is hard to attribute, which is why noisy calls and calls with heavy crosstalk produce unreliable ratios, and why very short calls swing wildly on a single long answer.
Two things the ratio does not measure: whether the rep said anything useful in their share, and whether the prospect's share was engaged conversation or a polite monologue about their weekend. Those limits are why it is a starting point rather than a verdict.
Where the "ideal" ratio comes from
The most quoted figure is from Gong, whose analysis of a large body of recorded sales calls popularised a talk-to-listen ratio of around 43:57 for top-performing reps. Other conversation-analytics vendors have published broadly similar findings: reps who close more tend to talk less than half the time, and the gap between average and top performers is a matter of several percentage points rather than a different order of magnitude.
The honest reading of that research is that talking less is associated with better outcomes across large samples, and that the effect is strongest on discovery calls, where the rep's job is to learn. It is not evidence that any individual call with a 60 percent ratio was a bad call. Presentations, technical demos and negotiation calls legitimately run higher.
That is why PepoSmart's scoring treats 40 to 50 percent as the full-marks band and tapers the score on either side rather than penalising a single point over 50. The exact rule, from the published call-score methodology: full marks for a host talk share of 40 to 50 percent, a reduced score between 30 and 40 and between 50 and 60, and one point lost for each percentage point outside the 30-to-60 band.
Why one metric is not enough
Any single metric can be gamed, and reps are smart. Told to lower their talk ratio, a rep can ask "does that make sense?" every ninety seconds and let the prospect ramble. The ratio improves; the call does not.
A good scorecard uses a small set of mechanics that are hard to satisfy simultaneously without actually running a good conversation:
Questions asked. A rep who asks 10 to 15 real questions in a 30-minute discovery call is learning. Fewer, and the call was a pitch. Far more, and it was an interrogation. PepoSmart benchmarks this at 10 to 15 questions per 30 minutes for full marks.
Filler words. "Um", "uh", "like", "basically". A few are human. A high count signals a rep who is thinking out loud rather than listening, and it is one of the fastest things to improve with awareness.
Longest monologue. A single uninterrupted stretch over about a minute is where prospects check their email. Even a rep with a healthy overall ratio can lose the call in one four-minute feature tour.
Talk dominance. A separate check on whether the host's share crossed clearly excessive thresholds, because a call at 80 percent is a different problem from a call at 55.
Together, these four make the scorecard difficult to game and easy to coach. The coaching scorecards page describes how they roll up into trends and a team view.
A fully published 100-point call score
Here is the complete formula PepoSmart uses, published so it can be checked. Every component is worth 25 points and every count is normalised to a 30-minute call, so a 15-minute call and a 60-minute call are scored on the same scale.
| Component | Points | Rule |
|---|---|---|
| Talk ratio | 25 | Full marks for a host share of 40 to 50 percent. Reduced between 30 and 40 and between 50 and 60. Falls one point per percentage point outside the 30 to 60 band. |
| Questions asked | 25 | Full marks for 10 to 15 questions per 30 minutes; 20 points for 7 to 20; 15 for 5 or more; 10 for 3 or more; 5 for at least 1. |
| Filler words | 25 | Starts at 25 and loses 2 points per filler word per 30 minutes, to a floor of 0. |
| Engagement | 25 | Starts at 25. Loses up to 15 points for the longest monologue, at thresholds of 45, 60, 90 and 120 seconds, and up to 10 more where the host's talk share exceeds 65, 75 or 85 percent. |
The total maps to a plain-language rating:
| Score | Rating |
|---|---|
| 85 to 100 | Excellent |
| 70 to 84 | Good |
| 55 to 69 | Average |
| 40 to 54 | Needs work |
| 0 to 39 | Poor |
Alongside the score, the scorecard reports the underlying metrics: the talk-to-listen ratio and host versus guest talk time, the number of questions, the filler-word count with a breakdown of which words and how often, the longest monologue, and words per minute. Coaching targets are adjustable, so a team can set its own talk-ratio band, minimum questions and filler ceiling if its calls run differently from the defaults. The details are in the sales intelligence documentation.
The limitation, stated plainly
The score measures conversational mechanics in the transcript. It does not measure whether the call was useful, whether the product fit the buyer, or whether the deal will close, and it should be read alongside the call itself rather than used on its own to evaluate a person. The underlying counts also inherit transcription errors, so a bad microphone lowers a score. Treat a score as the reason to open the recording, not as a substitute for it.
How to measure talk ratio without buying anything
You do not need a platform to start. For a handful of calls, this works:
- Record the call with the platform's own recording, with consent.
- Run it through any transcription tool that labels speakers.
- Count words per speaker (a spreadsheet does this) and divide.
- Count questions by searching for question marks in the host's lines.
- Count filler words with find-and-replace.
This takes about fifteen minutes per call, which is fine for a coaching sample and impossible at scale. Once a team wants every call scored, a notetaker that joins the meeting and produces the metrics automatically is the practical route. The AI notetaker guide explains how those tools work and what to check before choosing one.
A coaching routine that uses the numbers well
Numbers are only useful if they change behaviour. Here is a routine that managers who coach with call metrics tend to converge on.
Weekly: one metric, one rep, one call
Pick the metric each rep should work on this month and review one call a week against it. Not the whole scorecard. One thing. A rep working on monologue length should hear their own four-minute stretch and practise breaking it with a question. Next month, move on.
Monthly: look at trends, not calls
A single call is noise. Thirty days of calls is signal. Look for the direction of each rep's talk ratio, question count and filler count over the month. Praise movement, not levels; a rep who moved from 70 to 58 percent did more work than one who sat at 45 all month. PepoSmart raises trend alerts when a metric improves or slips compared with earlier calls, which saves the manager from building the chart by hand.
Quarterly: tag outcomes and look for correlation
Tag meetings as won, lost, active or stalled. Once there are enough of each, compare the metrics of won and lost calls. Your team's own pattern is worth more than any vendor's benchmark: maybe your winners ask more questions, or maybe they present more and the ratio runs higher than the textbook. PepoSmart's win/loss correlation surfaces once at least three won and three lost meetings are tagged, and the analytics view compares performance across reps for team owners and admins.
Talk ratio by call type
Because the "ideal" shifts with purpose, set expectations per call type rather than per rep.
| Call type | Reasonable host talk share | What to watch instead |
|---|---|---|
| Discovery | 35 to 45 percent | Questions asked; longest monologue |
| Demo | 50 to 60 percent | Monologue length; whether questions were invited between sections |
| Proposal or pricing review | 45 to 55 percent | Whether objections were surfaced and answered |
| Negotiation | 40 to 50 percent | Filler words under pressure; talk dominance |
| Customer check-in | 30 to 40 percent | Commitments captured; sentiment trend |
Set the target on the event type. A team that books its discovery calls and demos as separate event types can coach each against the right band, and can see in the call prep sheet which coaching reminder applies before the next call starts.
Common objections from reps, and honest answers
"The score punishes me for presenting." On a demo, yes, a higher ratio is expected. Set the target per call type, and read the engagement component, which cares about monologue length rather than the overall share.
"The transcript got my words wrong." It happens. Audio quality drives transcription quality, and a headset fixes most of it. A score is a prompt to open the recording, and the recording settles it.
"My best call last month scored 62." Probably true. A score describes mechanics, not merit. The point of tagging outcomes is to find out whether your team's mechanics predict its results, not to assume they do.
"I don't want to sound like a robot." Nobody does, and nobody should. The coaching goal is one behaviour at a time, over weeks, until it is natural. A rep who has stopped monologuing does not sound robotic; they sound like someone who is listening.
Where the numbers go next
A scorecard is most useful when it is not a separate document. In PepoSmart the score sits on the meeting record next to the summary, the action items and the follow-up draft, rolls into 30-day trends and a team view, and feeds the coaching reminders on the next call's prep sheet. For managers, the sales solution page describes how coaching, deal intelligence and CRM sync fit together; for anyone comparing conversation-analytics tools, the meeting intelligence guide sets out what each output is for.
Summary
Talk-to-listen ratio is a good first metric because it is simple and measurable, and a poor only metric because it can be gamed and it measures balance rather than quality. Target roughly 40 to 50 percent host talk time on discovery calls and adjust by call type. Pair the ratio with questions asked, filler words and monologue length, publish the formula so reps can check their scores, coach one behaviour at a time, and tag outcomes so your own data can tell you what actually correlates with winning. PepoSmart's call score is published in full on the product-facts page, and the free plan includes two AI recordings so you can see your own numbers before deciding anything; see the pricing page for plan allowances.
