Negotiation
Reaching decisions with people whose objectives differ — by working from interests rather than positions, and by making trades explicit.
Definition
Architects negotiate constantly: scope with product, timelines with delivery, controls with security, cost with finance, priorities between teams. Most of these are resolvable, and most are handled badly because both sides argue positions.
Positions versus interests
A position is what someone says they want: "we need it in six weeks". An interest is why: a customer commitment, a contractual date, a competitor launch, an internal expectation someone set.
Positions conflict; interests frequently do not. "Six weeks" and "this design needs ten" is a deadlock. "There is a customer commitment for the pilot in six weeks" opens options: a reduced scope for the pilot, a manual process for a component, a phased delivery.
Ask why before arguing about what. It is the single most useful move available.
The techniques that work
Make the trade explicit. "We can have it in six weeks without the reconciliation process; that means manual intervention for exceptions, roughly ten a week, until we add it in month three. Is that acceptable?" That converts a disagreement into a decision someone can take.
Offer options, not a verdict. Three choices with costs is far more likely to produce agreement than one recommendation with a justification.
Separate reversible from irreversible. Concede readily on reversible things; hold on the one-way doors. Knowing which is which is what makes you credible when you do hold.
State constraints as constraints. "This is not a preference — a data residency obligation means this cannot go in that region." Reserve that framing for things that are genuinely constraints, or it stops working.
Find the shared objective. Almost always there is one: a working system that ships.
What to avoid
- Winning. A decision the other side resents is implemented badly.
- Escalating early, which spends credibility and creates precedent that escalation is the mechanism.
- Treating every issue as equally important, so the important ones are not distinguished.
- Agreeing to something you know will not work, which defers the conflict to a worse moment.
Interview question
"Product wants a date that the design cannot meet. How do you approach the conversation?"