Sales & Leads · 26 August 2026 · 14 min read

Sales scripts: the call frame, the mistakes, and a full example

A working phone script is a decision tree, not a monologue: seven blocks, 2-3 qualification questions that change the offer, and a last block with a date in it. Here is the frame, the seven mistakes that show up in almost every script we review, a full invented call block by block, and how to tell whether yours is being followed at all.

THE FRAME, FOUR PHASES1Open, earn the minute2Qualify, three questions3Offer, mirrored back4Close on a dateSeven blocks collapse into four phases, and the fourth is the one floorsquietly skip.

Scripts have a bad name and mostly deserve it, because most of what gets called a script is a monologue printed on one page and read aloud. That version does deserve the reputation. The version that works is not a monologue at all.

A working phone script is a decision tree. It says what has to happen, in what order, and what the caller is allowed to do at each branch. It does not say the words. Confusing those two things is the whole problem, and it is why a good rep resents a script and a weak rep hides behind one.

Where this frame comes from

Ten years of running call centres, up to fifteen hundred live operators at peak, across France, Mexico and beyond. The block structure below and the list of mistakes further down are what survived reviewing those floors' calls, not an industry standard and not research. The example call near the end is invented start to finish, and is labelled again where it begins.

What a script is actually for

Three jobs, and only the third one is why anyone should care.

That third job is the one that pays for the exercise. Almost no call is lost because someone phrased the value proposition awkwardly. Calls are lost because nobody asked what the client already has, or because the conversation ended on "I'll get back to you" instead of a date. Those are omissions, and omissions are exactly what a frame catches.

The frame: seven blocks

Every call that works has the same skeleton underneath it, whatever the product. Seven blocks, in this order:

  1. Who you are and why you are calling. 10-15 seconds, no more.
  2. The right to the conversation. Not "is now a good time" — a stated length, so the other person can say yes to something specific.
  3. The right person. Confirm it before you spend anything on qualification.
  4. Qualification. 2-3 questions that change what you are about to offer.
  5. The offer, built out of the answers you just got.
  6. The objection. Expected, not survived.
  7. The next step, with a date and a time in it.

The timing is tighter than most people expect. The first three blocks together are about 30 seconds. Qualification runs 1-1.5 minutes. A cold call that is working rarely goes past 4 minutes end to end, and a cold call that runs twelve minutes is usually a rep who has been talking, not a prospect who is interested.

A script is not what you say. It is what you are not allowed to skip.

The qualification block

Three questions, and the test for each one is brutal: if the answer cannot change what you offer next, the question is filler and it is costing you the minute you have.

The second test is phrasing. A question that can be answered with one word will be, and then you are the one who has to fill the silence. Compare:

Three questions of that kind and you know more about the account than a ten-minute pitch would have taught you, and the prospect has done most of the talking, which is the other thing you wanted.

The offer block

Mirror first. Say back what you just heard, in their words and in one sentence: sixteen machines, no contractor, repairs when something breaks, downtime up to a week. Then two or three sentences of offer, aimed at the thing they actually said. Then hand the initiative back with a question.

The mirror is not a rapport trick

It does two hard jobs. It proves you were listening, which buys you the next minute. More usefully, it gives the prospect a chance to correct you before you have priced anything — and they will correct you, roughly a third of the time, on something that would have made the offer wrong.

Building the first version out of your own calls

Nobody should write a script from scratch, and nobody outside the sales floor should write one at all. Pull 20-30 recordings from different reps — the strong ones and the weak ones, deliberately — and go looking for four things:

That is your first version. Then revise it monthly against new recordings, because the objections move: a competitor changes its pricing, a regulation lands, a season turns, and the script that was right in March is answering a question nobody is asking in September.

The first ten seconds, where most calls die

A refusal in the first seconds is not a decision about your offer. It is a reflex against being sold to, and it fires before the person has processed a word you said. Which means it can be answered — once.

One recovery attempt, not three

Push past the first refusal once. Pushing twice is where a rep stops selling and starts damaging the number they are calling — and the brand behind it. Databases are not burnt by bad offers, they are burnt by third attempts.

A full call, block by block

What follows is invented — an imaginary service company for office equipment, an imaginary prospect, no recording and no client behind any of it. It is here to show the joins between the blocks, and the numbers in it are made up for that purpose.

Block one, the open. "Morning, my name's Olga, I'm with Technoservice — we service office equipment. I'll need two minutes, and I'll tell you at the end whether there's any point in a third." Name, company, subject, a stated length, and a reason to stay that is not a benefit claim.

Block two, the right person. "Is the equipment your side, or does someone else look after it?" If it is theirs, carry on. If it is not, you get a name and a route, and you ask one more thing before you go: what usually prompts them to call somebody about it. That answer travels with you to the colleague.

Block three, qualification. "How many machines are we talking about — printers, copiers, the lot?" Sixteen. "And when one of them goes down, who fixes it today?" Nobody on contract; they call whoever is available. "How long is it usually out for?" Up to a week, sometimes.

Block four, the offer, mirrored first. "So: sixteen machines, no contractor, repairs when something breaks, and up to a week without a machine. The part that costs you is the week, not the repair. What we do is a fixed monthly contract with a four-hour response and a loan unit if the fix takes longer — so the week becomes an afternoon. Want me to price it against what you spent on repairs last year?"

Block five, the objection, which arrives on cue. "A contract is going to be more expensive than calling someone when it breaks." Do not deny it. "It might be, on the invoice. Let's compare it against what you actually spent: give me last year's repair total and how many days you were down, and I'll put both numbers on one page. If the contract loses, you'll know in writing."

Block six, the next step, with dates in it. "I'll send the comparison by six this evening. I'll call you Thursday at eleven to go through it — does Thursday at eleven work, or is later in the day better?" Two dates, one of them a question, and nothing about the call is left to "we'll be in touch".

Read it back and the only thing worth noticing is structural: every block ends with either a question or a date. That is the entire mechanism. A rep who never ends a block flat cannot lose control of a call, and a rep who ends three blocks flat has already lost it.

The mistakes that show up in almost every script

From reviewing other people's, ours included:

And then the quiet one: nobody knows whether the script is being followed. In our own operations, under 5% of recorded calls are ever reviewed by a human, so "the team follows the script" is not a finding, it is an impression formed from a handful of calls a month.

Why a neuro-agent does not read the script either

This is the distinction that matters when someone shows you a demo. An auto-dialler with a recording plays a track. A neuro-agent takes the same script you just built, listens to the answer, picks the branch, and composes the next line — which is why it can handle the prospect who answers question three before you have asked question two.

What it does differently from a person is narrower and duller than the sales pitch suggests. It does not sound better than your best rep. It never varies with mood or with the hour, it never forgets the qualification block because the prospect was friendly, and it never ends a call without a date. On a floor, those three omissions are most of the gap between the top tier and the bottom.

Which script it follows is yours, decomposed into branches and fitted to your product — not a generic one shipped with the platform. The published rates, ex VAT per minute, put the choice at:

Script adherence stopped being an opinion

Locator listens to 100% of calls and checks 30+ parameters, script adherence among them, against the under 5% a human reviewer gets through. That changes what the script is for. It stops being a training document nobody audits and becomes a spec you can report against — which is also the moment a script starts improving, because you can finally see which block people skip.

What to do this week

  1. Pull 20-30 recordings from different reps and write down the four things above. Half a day, and it replaces every opinion in the room.
  2. Write the seven blocks on one page. If it does not fit on one page it is not a script, it is a manual.
  3. Rewrite every question that can be answered with one word.
  4. Put a date in the last block. Not a promise to follow up — a date, and a second option beside it.
  5. Decide how you will know it is being followed. If the answer is "we listen to a few", you have decided not to know.

A script does not make anyone persuasive. It makes the floor's worst call better and its best call repeatable, and on any floor of any size that is where the money is — not in the phrasing of the pitch, but in the calls that stopped ending on "I'll get back to you".

Frequently asked questions about phone sales scripts

Why write a script for experienced sales reps?
Not to constrain them. A script levels the floor — a new hire's first week sounds like the frame rather than like a first week — and it gives the team one place to put a discovery, so a question that works spreads instead of leaving with the person who found it. It also makes a change in strategy possible: you cannot deliberately change how a floor sells if there is no written version of how it sells now.
We have no script at all. Where do we start?
With your own recordings, not a template. Pull 20-30 calls from different reps, deliberately including weak ones, and extract four things: openings that got past the first ten seconds, questions that changed what got offered, objections that recur, and what the people who closed said at the end. That is a first version, and it is already fitted to your product.
Should reps read the script word for word?
No, and the ones who do are audible. Treat it as a frame: the reason for the call, the qualification questions and a dated next step are obligatory, the wording is not. What has to be identical across the floor is the order and the omissions, never the phrasing.
How do we know the script is actually being followed?
Not by listening to a few calls a month — in our own operations under 5% of recordings are ever reviewed by a human, which is not a sample you can conclude anything from. Speech analytics scores every call instead: Locator checks 100% of them against 30+ parameters, script adherence among them, at 0.10 per minute ex VAT.
Does an AI voice agent follow our script or its own?
Yours, decomposed into branches and fitted to your product. The difference from a recorded dialler is that it listens to the answer and picks the branch rather than playing a track — and the difference from a person is that it never skips the qualification block and never ends a call without a date.
How long should a cold call script be?
Short enough to fit on one page, and timed rather than counted: about 10-15 seconds to say who you are and why, roughly 30 seconds for the first three blocks, 1-1.5 minutes of qualification, and under 4 minutes for the whole call. A cold call running past ten is a rep talking, not a prospect interested.