| name | batna-zopa |
| description | Activate when: user says 'walk-away point', 'reservation price', 'should I take this deal', 'what's my leverage', 'they won't budge', or any salary/term-sheet/vendor/M&A negotiation where stakes are non-trivial. Do NOT activate when: the situation is a purely emotional relationship dispute with no transactional dimension; cultural context makes explicit BATNA discussion taboo. More: deciqai.com/s/batna-zopa |
BATNA & ZOPA
Overview
BATNA (Best Alternative to a Negotiated Agreement) is the value of your best fallback if no deal happens — the floor below which you walk away. A negotiator without a clear BATNA negotiates from fear; with one, from position.
ZOPA (Zone of Possible Agreement) is the overlap between what each side will accept. If it exists, a deal is possible; if not, the question is structural change — not harder negotiating.
Composes with anchoring, signaling-games, repeated-games-reputation, and prisoners-dilemma.
When to Use
- Any non-trivial negotiation: term sheets, salary, partnership, M&A, vendor contracts, customer pricing
- You feel pressure to take a deal that "feels bad" — usually a sign your BATNA is weak or not modeled
- Assess whether a deal can happen before investing time in tactics
- Raising in the 2024–2026 AI market (term sheets, SAFEs) or committing to a GPU-cloud / compute contract, where FOMO and "capacity is scarce" pressure obscure your real alternative
Not when: purely emotional/non-transactional disputes; cultures where direct BATNA discussion is taboo.
Coaching Novices (Adaptive Front Door)
- Engine mode: user has a concrete deal → run The Process directly.
- Coach mode: user is unfamiliar → guide step by step.
In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.
- One-line: BATNA = your walk-away value; ZOPA = the range where both sides say yes.
- Check fit: relationship-only situations → point elsewhere.
- "What negotiation are you in right now?"
[WAIT — do not advance until user responds]
- One question at a time: What's your BATNA? What's theirs? Where's the overlap?
[WAIT — do not advance until user responds]
- Name the specific opening offer and walk-away point they've identified.
[WAIT — do not advance until user responds]
The Process
Step 1 — Compute your BATNA: specific alternative + dollar value + timeline + probability → risk-adjusted value. "Going elsewhere" is not a BATNA. Strengthen before negotiating (parallel processes, extended runway).
Step 2 — Set reservation & aspiration: Reservation > BATNA (with buffer). Opening offer > aspiration, but defensible.
Step 3 — Estimate their BATNA: Ask directly; research alternatives; observe behavior (patient = strong, pressured = weak). Their reservation caps your ceiling.
Step 4 — Find the ZOPA: Overlap of reservation prices. If none exists → structural move (add issues, extend time, find new counterparty). Not "negotiate harder."
Step 5 — Distribute within ZOPA: Open at your favorable boundary (anchoring). Make small concessions — don't split. Trade variables when single-issue stalls. Walk at reservation, not earlier or later.
Output: BATNA & ZOPA Analysis
# BATNA & ZOPA Analysis: <deal>
## My side: BATNA <value> | Reservation $__ | Aspiration $__ | Opening $__
## Their side: Est. BATNA <…> | Est. reservation $__
## ZOPA: $__ to $__ | Exists? yes/no
## Strategy: [if ZOPA] open at __; walk at __; concede via __
## [if no ZOPA] structural move: __
## Falsifier: observation that would tell me my estimates are wrong: __
→ Method in Action: Fisher and Ury, 1981 — The Harvard Negotiation Project · Camp David Accords, 1978
→ 2026 lens: An AI startup's term sheet & compute contract (2024–2026) — build your BATNA before you negotiate the raise.
Pack: BATNA & ZOPA Patterns
| Negotiation | Your BATNA | Their BATNA |
|---|
| Term sheet | Other investors / bridge / status quo | Other pipeline deals |
| Salary | Other offers, current job, savings runway | Next candidate, leave role open |
| Vendor contract | Other vendors, build in-house, defer | Other customers, lower revenue |
| M&A | Continue independent, alternative acquirer | Build in-house, acquire competitor |
Applying It Well
- Compute BATNA before any negotiation session — mid-session computation is slow and biased
- Model their BATNA with the same rigor as your own — it sets your ceiling
- Multi-issue deals: identify what each side values most and trade across variables
→ Primary sources: references/sources.md
Common Rationalizations
[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.
| Fake move | Reality |
|---|
| [D] "My BATNA is strong" without specifics | Where exactly, with what offer, by when? If you can't answer, BATNA is weak. |
| [D] No estimate of their BATNA | Their BATNA caps your aspiration — without it you over-ask or under-ask. |
| [D] ZOPA non-existence = "they're being difficult" | Sometimes no deal exists. Structural change, not harder negotiation. |
| [D] Splitting the difference | Signals you have room. Make small concessions instead. |
| [D] Walking at aspiration, not reservation | Walking early leaves money on the table. |
| [D] Staying past reservation (sunk costs) | Negotiation effort is irrelevant. If deal < BATNA, walk. |
| [D] Bluffing about BATNA | If discovered, credibility collapses. |
| → Add [O] entries here after each real use — paste the actual failure pattern | What went wrong and why |
Red Flags
- "My BATNA is going elsewhere" without specifics — No estimate of their BATNA
- Pressure to close before BATNA computation is done — ZOPA not explicitly checked
- Reservation price not pre-committed before negotiating
Verification
Part of deciqAI Knowledge Skills — 228 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/s/batna-zopa · Built by deciqAI · github.com/deciqAI · Contributions welcome.
Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/batna-zopa.json
