Kluup opens on the question you came with, puts the headline figure first and keeps the evidence one click away. Every picture below is the real product, photographed from a running Kluup.
Was there a real, workable buyer on the end of this call at all?
Given whoever that turned out to be, what did the rep do with it?
Each gate carries the words it was decided on.
A next step is only a next step if the lead owns it — a booked visit, a reservation, a deposit, a second call on a day they named. “I'll send it over and ring you next week” does not pass.
A competent but unremarkable call is five or six out of ten on a dimension, and the product says so. Scores are only comparable within a call type and a channel, and the league table keeps them apart.
2 call types are included, because almost every floor runs more than one motion — outbound plus demos, or outbound plus inbound. Each gets its own dimensions, its own gates and its own playbook, and calls are only ever compared within the same one.
Changing a dimension is not free, and the product says so before you do it. Adding or redefining a dimension means re-running old calls from the audio Kluup still holds. The editor shows what that will cost and what it will re-score — and says plainly which calls fall outside the 90-day audio window and therefore cannot be re-scored at all, rather than inventing a number for them.
Building a scorecard from a new set of calls is the one step here we have not yet run all the way through on our own. It is built, and the screen you would review it on is built; see the note at the foot of this page.
The model chooses which words mattered. The machine decides when they were said.
The analyst cannot write a timestamp. Kluup finds the quoted words in the recording's own word-level timing and reads the clock off the audio. Where a call reaches Kluup with no recording — the demo workspace above is one — the times come from the transcript's own turns instead, and every figure that rests on that clock carries ≈ and the words estimated from the transcript, not measured.
The default vocabulary. It lives in your scorecard, so the labels and their colours are yours to change.
The breakdown is a separate operation from the score, deliberately. Scoring is included in the seat; the breakdown is the expensive half, so it is its own decision and its own meter. Score every call; break down the ones worth an hour.
A ranking that ignores coverage is measuring the sample, not the rep.
Scores are only comparable within a channel. A rep working warm inbound will out-convert a rep working a cold list on lead quality alone, and a table that ignores that measures the funnel. Kluup keeps them apart and says so on the screen.
Short dials still count as calls attempted. Anything too short to contain a conversation never reaches a model and never consumes an allowance, but it is not quietly dropped from the count either. A rep with two prior calls gets “no baseline yet” instead of a change worked out from one call.
A rep can see everything about themselves, and nothing about anyone else.
A rep sees their own calls, their own scores and the quotes behind them. They do not see another rep's quotes, another rep's scores, or the league table. That is enforced by the database, not by hiding a menu item.
Disputes travel with the record. A rep can contest a score; the disagreement stays attached to the call, so anyone who reads that score later sees it. Only a manager can mark a dispute resolved, and removing it entirely means deleting the call. The rep view and the dispute route are included for every customer, cannot be switched off, and are not a paid tier — they are what makes a scored rep lawful to manage.
connector, upload, or look around the demo first
Before anything is authorised, the connect screen shows the exact scopes asked for, what Kluup reads, what it never reads, and what happens next. It samples one recording to check whether your calls are dual-channel and tells you what you will get before you finish setting up.
Drop files in. Kluup probes each one, gates out anything too short to contain a conversation, splits the channels and transcribes them. Works with whatever your phone system exports.
A ready-made, clearly-labelled fictional floor with scored calls, a league table and full breakdowns, so you can judge the product before you hand it a single real recording. It is a step inside the trial rather than a public sandbox — you sign up first, and no card is asked for.
Kluup never ingests a CRM's transcript. It takes the audio and transcribes each channel on its own hardware, because a stereo recording puts the rep in one ear and the lead in the other — which makes “who said this” a recorded fact rather than a guess. The transcript is held in memory for the length of one analysis and is never written to disk.
If your recordings are single-track, Kluup will not guess. You still get activity, the lead gates, the verdict and the score; you do not get the breakdown, attributed quotes, talk ratio or the timeline. The full policy is on the pricing page, on the page before you sign up.
The same rule runs inside a call. A share of the talking Kluup could not attribute to a speaker is left blank and says why — never printed as 0%, which would read as a rep who said nothing.
so you can rule Kluup out quickly if it is wrong for you
We would rather you read this here than find it out in week two. Two things are built but have not yet been run end to end by us: scoring a call you have just uploaded, and building a scorecard from a new set of calls. Getting the recording in, checking its length and transcribing it are done and timed. If the scoring step cannot run, the call stops there and tells you; nothing is filled in for it.
Everything else on this page — the scores, the breakdowns, the timeline, the league table, the rep view, the dispute route and the allowance meter — is drawn from calls that really went through it.
14 days, 10 full breakdowns, no card. Or take the demo-workspace step inside the trial and judge it on someone else's floor first.