Ask a sales leader what makes a good negotiator and the answer is usually a trait: tough, calm, persuasive, hard to rattle. Ask what the research measures and the answer is different. Negotiation skill shows up as a small set of learnable behaviors, and most of them have little to do with toughness. The best-documented ones are about what a seller pays attention to: the other side's interests rather than their feelings, the possibility that both sides want the same thing, and the way the buyer will feel about the deal after it is signed.

The earlier piece on the negotiation stage covered anchoring, packaged offers, and channel choice. This one asks what a seller actually gets better at when they get better at negotiating.

The short answer is that skilled negotiators keep two ledgers instead of one. They track the economic result, and they track the subjective result with equal care, because in B2B sales the subjective ledger decides what happens in the next negotiation, the renewal, and the reference call.

Most sellers are trained, explicitly or by their comp plan, to keep one ledger: price, discount, term, and margin. This framing feels rigorous, and it misses most of what the research finds.

Jared Curhan, Hillary Anger Elfenbein, and Heng Xu set out in 2006 to map what people value when they negotiate, publishing the results in the Journal of Personality and Social Psychology [2]. In their first study, 103 participants (MIT undergraduates, Boston community members, and union and management practitioners) listed what mattered to them in a negotiation. A striking 20% did not list a single factor describing the objective terms of the agreement. The authors went on to define "subjective value" as "the social, perceptual, and emotional consequences of a negotiation" and confirmed it has four parts: feelings about the instrumental outcome, feelings about the self, feelings about the process, and feelings about the relationship. They also found subjective value was a better predictor of future negotiation decisions than economic outcomes were.

Sales organizations tend to skip past that last finding. The follow-up research makes it hard to skip.

In 2009, Curhan, Elfenbein, and Gavin Kilduff followed MBA graduates who negotiated their first job offers, publishing in the Journal of Applied Psychology [3]. The analysis covered 70 graduates who had negotiated and accepted offers, surveyed once in March 2005 and again in May 2006. Subjective value from the negotiation predicted higher compensation satisfaction, higher job satisfaction, and lower turnover intention a year later. The economic value the negotiators had won, in the authors' words, "had no apparent effects on these factors." The two ledgers were also only weakly related to each other (r = .20, not significant): winning more money did not reliably make people feel better about the deal.

The sample is small and the setting is employment, not vendor selection, so treat this as directional. But the pattern carries over cleanly to a customer who negotiated a contract with you and now has to live with it for a year. How they felt about the negotiation is a better guide to how they will feel at renewal than the discount they extracted.

A 2010 study by Curhan, Elfenbein, and Noah Eisenkraft in the Journal of Applied Social Psychology tested the repeat-business question directly [4]. 174 master's-level business students negotiated an eight-issue employment contract twice with the same counterpart, a week apart, told that a year had passed in between (82 dyads remained in the final analysis). Negotiators who experienced more subjective value in round one achieved better individual and joint economic results in round two, even after controlling for their round-one economic outcomes. Subjective value in round one also predicted the desire to negotiate with the same counterpart again; round-one economic performance did not. The authors call the effect sizes "relatively modest," which is the right way to read them: subjective value is not magic, but it is a real input to future economic results with the same party.

For a seller with a land-and-expand motion, that is the whole game. You will negotiate with this customer again, often with the same people.

The evidence has a sharp edge, and a skill article that skips it would mislead. William Becker and Jared Curhan published a study in the Journal of Applied Psychology in 2018 titled "The Dark Side of Subjective Value in Sequential Negotiations" [5]. They looked at sequential negotiations, meaning several negotiations in a short window, each with a different counterpart. Across a multi-round lab study and a longitudinal field study of employees negotiating as part of their jobs, high subjective value from the first negotiation was negatively related to economic results in the next one. The mediator was pride: feeling good about the last deal made negotiators less effective in the following one. (The authors also tested anger as a mediator; it was not supported.)

So the subjective ledger has two entries that sellers must keep separate. The buyer's subjective value is an asset you build, because you will meet that buyer again. Your own subjective value from the last deal is a risk to manage, because it travels with you into the next deal with a different customer. A rep who closes a warm, satisfying deal on Tuesday and walks into a procurement call with a different company on Wednesday is, on this evidence, at elevated risk of giving away too much.

If the goal is to build the buyer's subjective value without giving away the economics, the first skill is knowing which kind of attention to pay. Adam Galinsky, William Maddux, Debra Gilin, and Judith White tested exactly this in a 2008 paper in Psychological Science [1]. They separated perspective-taking (understanding what the other side is thinking, their interests and purposes) from empathy (understanding and connecting with what they are feeling).

In their second study, 152 MBA students negotiated the sale of a gas station in a case where the buyer's maximum was below the seller's minimum, so a deal required discovering an interest beyond price. Buyers instructed to take the seller's perspective reached a deal 76% of the time, against 54% for buyers instructed to empathize and 39% for a control group. Perspective-taking beat control significantly; empathy did not. (The perspective-taking versus empathy gap was in the predicted direction but did not reach conventional significance, p = .089.)

The third study, a job negotiation among 146 MBA students with eight issues, is more relevant to complex B2B deals. Recruiters told to take the candidate's perspective produced higher joint gain than the control group, and 40% of their dyads found the maximum possible joint outcome, versus 22% in the empathy condition and 12% in control. The empathizers did create some extra value, but, as the authors put it, those gains "went mostly to the empathizers' opponents." The paper's closing advice is that it pays to "think for" rather than "feel for" the other side.

Empathy is not a flaw. Negotiators in the empathy condition left their counterparts more satisfied with how they were treated. The lesson is about where the analytic work happens. Understanding the buyer's interests is what finds the deal and protects your share of it. Warmth toward the buyer's feelings is what builds the subjective ledger. Collapsing the two into "be empathetic" produces a seller who is liked and underpaid.

The second skill is resisting a specific perceptual error. Leigh Thompson and Reid Hastie's 1990 study in Organizational Behavior and Human Decision Processes found that "most negotiators enter negotiation expecting the other party's interests to be completely opposed to their own," and that "a substantial number of negotiators fail to realize when they have interests that are completely compatible" [6]. They named these the fixed-sum error and the incompatibility error. They also found that most learning about the other side's interests happens in the first few minutes, and that negotiators who learned early earned higher payoffs than those who learned late.

Max Bazerman, summarizing Thompson's work for the Program on Negotiation's newsletter, describes an eight-issue simulation in which two issues were fully compatible, meaning both sides preferred the same outcome. Even so, 39% of negotiating pairs failed to agree on the shared preferred outcome for at least one of them [7]. They settled on something worse for both sides because each assumed the other wanted the opposite.

In a software deal this looks like a seller defending a long contract term the buyer would also prefer (for budget predictability), or conceding on implementation speed when the buyer's team cannot absorb a faster rollout anyway. The skill is cheap: early in any negotiation, ask what the buyer would rank first and last among the open issues, and treat every issue as possibly compatible until the answer says otherwise.

The newest large-scale evidence on negotiation style comes, unusually, from AI agents. In a study published in the Proceedings of the National Academy of Sciences in 2026, Michelle Vaccaro, Michael Caosun, Harang Ju, Sinan Aral, and Jared Curhan ran an international competition in which 286 participants from more than 40 countries designed AI negotiating agents that then negotiated with each other 182,812 times across distributive and integrative scenarios [8]. Each agent was scored for warmth and dominance and measured on value claimed, value created, deals reached, and the counterpart's subjective value.

Warm agents reached deals at significantly higher rates, created more value, fostered higher counterpart subjective value, and earned more points overall. Dominant agents claimed more value in the deals they did close, but their results were muted by a tendency to drive negotiations to impasse. Question-asking, positivity, and gratitude (all associated with warmth) were strongly associated with reaching deals and creating both objective and subjective value. Longer conversations, associated with dominance, were associated with impasses. The winning agent used structured, step-by-step preparation and earned the most individual points across the integrative negotiations, with value creation in the 90th percentile and counterpart subjective value that was second highest of all agents.

These are AI agents negotiating with each other, not humans negotiating enterprise contracts. Still, the pattern matches three decades of human research: warmth gets you to a deal and grows the pie, a firm position on your own priorities claims your share of it, and toughness that crowds out questions costs more in lost deals than it earns in margin.

Pulling the evidence together, here is a working definition sellers can use.

The Two-Ledger Rule: a negotiation is skilled when the buyer leaves with high subjective value and you leave with your priority terms intact, and neither was traded for the other.

In practice, each ledger has its own entries.

The economic ledger tracks your priorities ranked in advance, your walk-away terms, and each issue tagged as distributive (you gain only if they lose), integrative (you value it differently, so you trade), or compatible (you both want the same thing). Perspective-taking is the tool for filling in this ledger, because it tells you which issues are which.

The subjective ledger tracks the buyer's four entries from Curhan's research: do they feel the outcome is fair, do they feel competent and not embarrassed, did the process feel respectful and orderly, and is the relationship stronger or weaker than when you started. Warmth, questions, and gratitude are the tools for this ledger, and they cost nothing on the economic one.

The rule also includes a self-check from Becker and Curhan: before the next negotiation with a different counterpart, close out your own subjective ledger from the last one. Pride from a good deal is not information about this deal.

Sales leaders reasonably ask whether any of this transfers from a paper to a rep. The evidence says yes, with conditions. A meta-analysis presented at the International Association for Conflict Management in 2020 by Dominik Sondern, Christoph Nohe, and Guido Hertel of the University of Münster pooled 17 studies (20 samples, 44 effect sizes) and found a large positive training effect (Hedges' g = .72), with considerable variation between studies [9]. Longer trainings were more effective. Neither feedback nor whether the training was virtual made a significant difference. This is a conference abstract, not a journal article, so treat the exact size with caution.

Hal Movius's 2008 review in Negotiation Journal adds detail on what kind of training works [10]. Case studies help most when trainees compare multiple cases to extract the shared principle; studying cases one at a time does not reliably improve later problem solving. Didactic instruction alone did no better than a control group in one cited study, while observational and analogical learning did better. And trainee confidence is a poor guide: Movius cites a study in which executives felt sure they had learned new skills, yet 90% still left money on the table at the end of the day.

For an individual rep, the Two-Ledger Rule turns into a short pre-call routine. Rank your open issues and tag each as distributive, integrative, or possibly compatible. Write down two or three hypotheses about what the buyer is thinking and trying to protect (their interests, not their mood), and plan the questions that would test each one in the first ten minutes. Decide what you will hold on your top priority before the call, so warmth in the room does not slide into concessions.

During the call, ask about the buyer's priorities across issues early, before positions harden. Express appreciation for specific moves the buyer makes. Keep the conversation focused: in the AI study, long conversations were a marker of impasse, not diligence.

After the call, score both ledgers. Did you keep your priority terms? Would the buyer describe the outcome as fair, the process as respectful, and the relationship as stronger? If you scored high on the second and low on the first, you were empathetic but not perspective-taking. If you scored high on the first and low on the second, expect to pay for it at renewal.

For sales leaders, the training evidence points away from one-off workshops and toward longer programs built on comparing real deals: two won negotiations and one lost one, examined side by side for the principle that separated them. Deal reviews that ask only "what discount did we give?" are training reps to keep one ledger. Adding "how will this buyer feel about this negotiation in twelve months?" trains them to keep two.