Better agreements.
Clearer decisions.

The Nash bargaining solution asks which agreement to select from feasible outcomes when parties can cooperate and know what happens without agreement. A Nash equilibrium asks which strategies are stable: no party benefits by changing its strategy alone while others keep theirs. Stability can coexist with inefficiency; calling an outcome “Nash” does not make it a good deal.

Use bargaining solutions to compare agreements under explicit assumptions. Use a strategic model when successive offers, private information or response rules matter. The tools are complementary: one selects an outcome; the other studies incentives within a specified game.

References: [nash1950] [nash1953]

Represent each agreement x by utilities uA(x) and uB(x). The disagreement point d = (dA, dB) records utilities without agreement. The symmetric solution maximizes (uA − dA)(uB − dB) over feasible agreements that do not worsen those alternatives. If no strict improvement for both parties exists, the model has no mutually positive surplus to share.

What matters is gains over disagreement, not the product of raw utilities. The weighted version maximizes (uA − dA)^α (uB − dB)^(1−α), with 0 < α < 1. A weight needs justification from the model or an agreed principle; a slider does not discover a person’s actual power. The formula does not automatically make a division fair.

References: [nash1950] [nash1953] [serrano]

In the classical framework, with a compact convex feasible set and a strict improvement over disagreement, the characterization combines Pareto efficiency, symmetry, invariance to positive affine transformations of each utility and independence of irrelevant alternatives. Symmetry means equal treatment of a symmetric problem, not always splitting the price in half.

Independence requires retaining the selected outcome when the option set shrinks but that outcome remains available, with the same disagreement point. This is a debatable assumption rather than a psychological law. Convexity may represent lotteries over agreements: when mixtures cannot be implemented or accepted, a discrete set needs careful interpretation.

References: [nash1950] [serrano]

Your BATNA is what you can actually do if agreement fails. A reservation value translates that alternative into the terms of this negotiation, accounting for costs, risks and timing. The ZOPA is the set of agreements acceptable to both parties. An ambitious aspiration is not a reservation value, and an implausible threat does not improve an alternative.

In the simulator, the seller requires at least C + dA and the buyer pays at most V − dB. The interval exists when C + dA ≤ V − dB. Outside that case, insisting on a price does not create value: costs, scope, financing or alternatives would need to change through feasible voluntary action.

References: [batna]

Write one page covering your objective, executable alternative, reservation value, aspiration, negotiable issues and constraints. Separate facts from estimates about the other party. Check who can decide, who must approve and what information can be shared. An agreed agenda and confidentiality rules help avoid negotiating terms that nobody can authorize.

Rehearse an expected, an adverse and a favourable scenario. Decide what evidence you would request in each and when you would pause to consult. Bring open questions: “What makes this schedule difficult for you?” reveals more than “Do you accept, yes or no?” Preparation does not require predicting every answer.

References: [preparation]

Break the agreement into attributes such as price, timing, scope, support, risk and flexibility. Identify non-compensable constraints before scoring. Use understandable scales for each party and compare complete packages. An additive model can be an approximation when its preference-independence assumptions are appropriate.

Do not directly multiply ordinal ranks such as first, second and third. The Nash product requires coherent cardinal utilities; arbitrary nonlinear transformations can change its solution. Use intervals for uncertain preferences, inspect sensitivity and ask which concrete trade-off would justify a score difference.

References: [serrano]

When only price is negotiated, one party’s improvement often costs the other. Adding issues may enable trades: flexible timing could matter greatly to a supplier and cost a buyer little. Explore different priorities without assuming that everything can be compensated with money.

Prepare packages of similar value to you that differ in their terms. Ask which fits best and why. Do not describe unequal offers as equivalent. In the example table, D dominates B because it improves both utilities: finding such improvements differs from deciding how to divide an already efficient frontier.

References: [meso]

Distinguish a position (“I need delivery on Monday”) from the interest behind it (“my team starts then”). Summarize your understanding and allow correction. Acknowledging a concern does not mean accepting a demand. When tension rises, return to a concrete issue and avoid assigning intentions you cannot verify.

Share information gradually and with authorization. You can explain what matters without necessarily revealing every limit. Ask about constraints, comparisons and consequences. An external criterion helps when both parties understand its relevance and provenance, rather than treating a number as beyond discussion.

References: [principled]

Connect each offer to a reason and a complete package. A conditional concession (“if timing changes, we can extend support”) clarifies the exchange. Avoid conceding repeatedly without understanding what the other party is learning. An initial number can frame the conversation, so prepare independent reference points.

An extreme offer is not automatically a sound strategy. After each round, update your understanding of preferences and limits. If a proposal is rejected, ask which component blocks progress. State its validity period and pending conditions without manufacturing urgency or nonexistent alternatives.

References: [principled]

In an alternating-offer model, rules and the cost of waiting affect the division. Rubinstein studies a precise setting with complete information and time discounting; this does not prove that the most patient person always wins in every real conversation. Offer order, commitments and exit options also matter.

In practice, record genuine deadlines and who bears each delay. In repeated relationships, reliable delivery and verification may matter more than a small advantage today. Check implementation incentives: an attractive written agreement can fail when a party benefits from breaking it and monitoring is absent.

References: [rubinstein]

When parties disagree about the future, distinguish different forecasts from different interests. An “if X occurs, Y applies” clause can link actions or payments to a verifiable outcome. Define measurement, who performs it and how disagreements are resolved before committing.

Compare scenarios and check whether the proposal still beats alternatives under reasonable assumptions. Do not assign probabilities merely to complete a spreadsheet. If an outcome cannot be verified or the clause changes incentives affecting it, the design needs review. Calculating an average does not remove uncertainty.

For several parties, a weighted extension uses the product of all parties’ gains with positive weights. You still need to describe participation, coalitions, vetoes and what disagreement means for each party. An affected person who is absent is not automatically represented by the formula.

The Kalai–Smorodinsky solution follows a different axiomatic approach and compares proportional gains relative to ideal aspirations. A utilitarian criterion maximizes a sum, requiring justification of scales and comparisons. No criterion replaces discussion of rights, representation and constraints. Choose and disclose the criterion before relying on its results.

References: [kalai]

Review the whole package: deliverables, quantities, responsibilities, dates, dependencies and acceptance criteria. Distinguish exploratory discussion from valid approval. Summarize terms in writing and confirm that the accepting person has authority. Disagreement over a word can conceal a substantial scope difference.

Plan follow-up, change management and a route for resolving problems. Define what happens if a dependency fails and when the agreement will be reviewed. Include appropriate professional review when the subject requires it. Agreement quality is also tested after signing, when delivery begins.

Before a conversation, choose a practice objective: asking better questions, summarizing without judgment or proposing packages. Afterwards, record what you assumed, what you learned and what remains untested. Request feedback about observable behaviour rather than simply asking whether you “negotiate well”.

Compare preparation with results: did you beat your alternative, what value went unexplored, were commitments delivered and how did the relationship change? Repeat the case with a different constraint and observe the consequences. Improvement means refining process and information rather than crediting the model for every personal success.

Models depend on the feasibility, preferences and alternatives you enter. A weight does not legitimize coercion, discrimination or deception. Check consent, decision-making capacity and constraints before optimizing. If information, timing or access to advice is severely unequal, make that visible and seek an appropriate process.

This guide is educational. Examples and photographs are illustrative, and figures do not describe an actual negotiation. The simulator assumes linear utilities, known information and two parties. It does not estimate hidden preferences, guarantee agreement or decide for participants.