Never Split the Difference

The book in one minute

Chris Voss spent 24 years in the FBI and finished as its lead international kidnapping negotiator. This is his attempt to hand the techniques that work when a life is on the table to people negotiating salaries, contracts, house prices and bedtimes.

The title is a provocation. Splitting the difference is what everybody does when a negotiation stalls, and Voss argues it is almost always the worst outcome available - not a solution but an agreement to stop thinking. His illustration: your partner wants you in black shoes, you want brown, so you split the difference and go out wearing one of each. Nobody wanted that.

The working version of that is tactical empathy. Note the adjective: this is not niceness, agreement or sympathy. It is the deliberate recognition of the other side's position and feelings, said out loud, because a person who feels understood becomes dramatically easier to move. Everything else in the book - mirroring, labelling, inviting "no", calibrated questions - is machinery for producing that feeling and then keeping the other side talking long enough to say something you did not know.

Why the rational model fails

Chapter one is an argument with the orthodoxy. For decades the dominant framework came from the Harvard Negotiation Project and Getting to Yes: separate the people from the problem, focus on interests rather than positions, know your BATNA, work toward mutual gain on objective criteria.

Voss's objection is that this treats the other party as a rational actor calmly weighing costs and benefits, and human beings are not that. He leans on Kahneman and Tversky:

• Loss aversion. A loss looms roughly twice as large as an equivalent gain, which is why deadline pressure extracts concessions that logic never would.
• Anchoring. The first number stated pulls the final one toward it, even when everybody knows it is arbitrary.
• Framing and fairness. People reject profitable deals that feel unfair, and accept worse ones that feel respectful.

So: you cannot separate the people from the problem, because in most negotiations the people are the problem - and the emotions are not noise obscuring the deal, they are the mechanism through which it happens.

The FBI's own model, developed at Quantico, is a staircase called the Behavioral Change Stairway. Active listening, then empathy, then rapport, then influence, and only then behavioural change. Most people try to start at influence. That is why most people fail.

The most popular ideas in the book

1. Tactical empathyRecognising and articulating the other side's position out loud. Not agreement, not sympathy - a deliberate instrument for making someone feel understood so they become movable.
2. MirroringRepeat the last one to three words they said, with a questioning inflection, then shut up for four seconds. It keeps them talking and builds affinity without asking a single question.
3. Labelling"It seems like...", "It sounds like...", "It looks like..." Naming an emotion defuses the negative ones and reinforces the positive ones. Never start with "I" - that makes it about you.
4. The accusation auditList every negative thing they could think about you, and say all of it yourself, first, at its worst. It sounds like handing over ammunition and does the opposite.
5. "No" is the start, not the endChasing "yes" makes people defensive. "No" makes them feel safe and in control, and a person who feels protected keeps talking. Ask questions they can safely refuse.
6. "That's right" vs "You're right""That's right" means they feel understood and is the turning point. "You're right" is a polite dismissal that means they want you to stop talking.
7. Calibrated questionsOpen questions starting with how or what that hand the other side the work while giving them the feeling of control. Never "why" - it reads as an accusation in almost any language.
8. "How am I supposed to do that?"The single most useful sentence in the book. A refusal, a request for help and an invitation to reconsider, all at once, and impossible to take offence at.
9. Never split the differenceCompromise is chosen because it is safe and defensible, not because it is good. An agreement neither side wants is a shared failure with better manners. No deal beats a bad deal.
10. Deadlines are usually softer than they lookTheir function is psychological, not contractual - they manufacture the fear of loss. Test them rather than obeying them, and never hide your own.
11. The Ackerman modelA structured concession pattern - 65%, 85%, 95%, 100% of target - whose decreasing increments make your final number read as a real wall.
12. The Black SwanEvery negotiation hides three to five pieces of information you do not know exist. Behaviour that looks irrational is a signal one is nearby. Keep them talking until it falls out.

The techniques, in detail

1. The three voices

Voss starts with the simplest possible tools, because under pressure you will not execute anything complicated. Most people only ever use one voice.

2. Mirroring

Not body language. Repeat the last one to three words the other person said, with a slight upward, questioning inflection - then stay silent for about four seconds. The silence feels much longer than it is, and the discomfort is doing the work.

It signals that you are listening, asks them to elaborate without asking a question, and quietly builds affinity, because humans are drawn to what resembles them. Voss cites a study of waiting staff in which those who repeated customers' orders back word for word earned substantially more in tips than those who responded with something positive and reassuring. Same content, entirely different result, purely from the repetition.

3. Labelling and the accusation audit

Labelling is naming the other person's emotion out loud - not guessing at it silently, not feeling it, not fixing it. All three forms begin the same way, and none of them contains the word "I".

The most counterintuitive extension is the accusation audit. Before you make your case, list every negative the other side could possibly think about you - and say all of it yourself, first, in the worst plausible phrasing.

4. Master "No"

Conventional technique chases "yes" - small agreements, momentum, close. Voss says this is backwards: people know exactly what you are doing and can feel the trap closing. He distinguishes three kinds of yes, and only one is worth anything.

So invite "no" instead. It gives people something they want more than agreement: safety and the feeling of control. A person who feels protected keeps talking.

5. "That's right" - the turn

When the other side says "that's right", unprompted, you have reached the moment the whole method aims at. It means they feel genuinely understood, and it is almost always where the negotiation turns.

The counterfeit is "you're right", and Voss is emphatic that it is the worst thing you can hear. It is what a person says to end a conversation - a polite dismissal meaning please stop talking. If you win someone over with a clever argument and they say "you're right", you have almost certainly lost them.

"That's right" cannot be extracted by argument. It is produced by summary, which Voss defines precisely as labelling plus paraphrasing: you describe their world back to them in your own words, name the emotions underneath it, and do it well enough that they recognise themselves in the description. If the summary is accurate the response comes back almost automatically.

6. Bending reality: deadlines, fairness, numbers

Deadlines are more flexible than they appear and frequently invented. Their real function is to manufacture the fear of loss. Never hide your own - concealing it produces worse outcomes than disclosing it - and treat theirs as information to be tested. Most movement in any negotiation happens right at the end, so whoever panics first pays for it.

The F-word is "fair", and it gets used three ways: as a weapon ("we've given you a fair offer", delivered wounded, to make you defensive), as a manipulation ("I just want what's fair"), and as the genuine article. Adopt the third early:

On numbers:

• Anchoring. Generally let them go first - their opening reveals how they see the deal - but arrive with a defence, and if their number is extreme do not counter with a number at all. Deflect with a calibrated question; countering legitimises their anchor.
• Anchor the emotion first. Before naming a figure, deflate the shock: "I'm going to give you a number you'll think is far too low, and I want to explain how I got there."
• Ranges, not points. Cite a range with your real target near the bottom. People hear a legitimate band rather than a demand.
• Non-monetary terms. When the number will not move, move something else - timelines, scope, credit, referrals, order of delivery.
• Odd, precise numbers. $37,893 reads as the output of a calculation. $40,000 reads as an opinion, and invites a counter-opinion.
• The surprise gift. An unexpected, modest concession offered unasked triggers reciprocity, which is among the strongest levers available.

7. Calibrated questions

The technique most readers actually adopt, because it survives contact with real life. Open questions beginning with how or what, containing no accusation and no demand. They give the other person the experience of control while handing you the agenda, because the person answering is the one doing the work.

Avoid why. In almost every language it sounds like an accusation. The one exception Voss allows is using "why" to make someone defend a position that favours you.

The same tool lets you say no without the word. Voss gives four escalating forms: "How am I supposed to do that?" → "Your offer is very generous, I'm sorry, that just doesn't work for me." → "I'm sorry, but I'm afraid I just can't do that." → "I'm sorry, no."

There is a second half people skip: after asking, actually listen to the answer. That is where the information you needed has been hiding.

8. Guaranteeing execution

You got the agreement, everyone shook hands, and nothing happened. Voss's principle is that "yes" is nothing without "how" - an agreement with no implementation plan is a wish. Ask the how questions in the room, so the other side articulates the plan in their own voice and therefore owns it. And find out who else has to agree, because deals collapse when the person you convinced was never the decision-maker.

Then the detection tools:

• The rule of three. Get agreement to the same thing three times in one conversation, in different forms - the commitment, then a summary that draws "that's right", then a calibrated question about implementation. It is hard to repeat a lie three times comfortably.
• 7-38-55. Voss's reading of Mehrabian: 7% of the message is words, 38% tone, 55% body language. His practical use is narrow and sensible - when words and tone disagree, believe the tone. (See the critique: this statistic is badly overextended in general use.)
• The Pinocchio effect. Liars tend to use more words, more third-person pronouns and more complex sentences, because they are working harder to be believed. An unexpectedly elaborate answer to a simple question is worth noticing.
• Pronoun watching. The more important someone is, the less they use "I". Heavy "I" and "me" often signals less authority than is being implied.

9. The three negotiator types

There is no correct style, only a style you must identify. Voss's warning is that most people assume the other side negotiates the way they do, and the mismatch is where deals go wrong.

10. The Ackerman model

The concrete bargaining system, developed at the FBI. A structured concession pattern rather than a haggle, designed to make your final number feel like a genuine limit.

1. Set your target price first.
2. Open at 65% of it.
3. Plan three raises in decreasing increments: 85%, 95%, 100%.
4. Use empathy and varied ways of saying no between each, so each move has to be earned.
5. Make the final figure precise and non-round, so it reads as computed rather than chosen.
6. At the final number, add a non-monetary item they probably do not want - it signals you have genuinely run out of money.

11. The Black Swan

A black swan is information you did not know existed, which reshapes the whole negotiation once it surfaces. Not an unknown - an unknown unknown. Voss claims every negotiation contains three to five, and that most are lost by people who prepared thoroughly for the wrong deal.

Which reframes everything above: the point of the techniques is not to persuade. It is to keep the other side talking long enough for a black swan to fall out.

Chapter-by-chapter summary

The appendix: prepare a negotiation one sheet

The most immediately usable page in the book. Everything compressed onto a single sheet you fill in before any negotiation.

Where the book is weakest

• 7-38-55 is the weakest thing in it. Albert Mehrabian's original studies were narrow: how people judge a speaker's feelings and attitudes when words and delivery contradict each other, using single words in a lab. Mehrabian has publicly objected to how the numbers get generalised. As a claim about communication in general the ratio is not supported, and it should not be quoted as though it were.
• The transfer from hostage work to business is asserted, not demonstrated. Voss is upfront that his world had no walk-away, no repeat business and nothing to divide, and argues this makes it a harder case. The other reading is that it makes it a different one. Commercial negotiation is usually repeated, usually has alternatives on both sides, and usually involves a genuinely divisible surplus.
• The anti-compromise position is rhetorically excellent and situationally risky. In a one-off transaction with a stranger, holding firm is often right. In a long-running relationship with a colleague, supplier or spouse, an occasional visible willingness to meet halfway is not weakness - it is maintenance. The framing leaves little room for that.
• The evidence is anecdotal. The cases are gripping and told by the man who was in the room. There are no controlled comparisons and the counterfactual is unknowable. We hear about the negotiations that worked.
• The manipulation question is raised and not fully resolved. Tactical empathy is explicitly instrumental - empathy deployed in order to get something. In the hands of someone who does not much care about the other party, this is a toolkit for extracting agreement from people who feel understood and are not. Voss would argue genuine understanding produces better deals for both sides, and there is a good case for that, but the word "tactical" is doing real work.

What survives is a great deal. The critique of the purely rational model is correct and well supported by the behavioural literature. Mirroring, labelling, the accusation audit, calibrated questions and the "that's right" / "you're right" distinction cost nothing, and can be practised tomorrow in conversations where nothing is at stake - which is exactly what Voss recommends.

Frequently asked questions

About this summary. This is an original summary and commentary on Never Split the Difference: Negotiating As If Your Life Depended On It by Chris Voss with Tahl Raz (2016), written and narrated for the AIEIO book library. It paraphrases the book's ideas and contains no reproduced text. If the techniques were useful to you, buy the book - the case studies carry most of its force and no summary can reproduce them.