NNaval
← All frameworks
Leadership

The Poverty of Compromise

Run one side's real idea rather than a middle option nobody believed in.

Difficulty
Moderate
Time to result
~weeks to results
Steps
6
Confidence
76%

Person A holds idea X, person B holds idea Y, and the standard move is to land somewhere between them on idea Z. But nobody ever believed Z was the best idea. When Z fails — and it usually does — A goes back to insisting X was right and B goes back to insisting Y was right. You have refuted a theory nobody held, so the disagreement is exactly where it started. The alternative is to keep both theories intact, choose one to actually run with a named owner, agree beforehand what outcome would count as a refutation, and let reality decide. That converts a stalemate into an experiment. Compromise remains preferable to a fight when no test is possible, but it should be the fallback, not the default. The posture that makes this work is strong opinions loosely held.

Origin

Discussed on Naval's The Beginning of Infinity series as the Popperian answer to decision-making at loggerheads, paired with Mark Andreessen's formulation of strong opinions loosely held.

Core principles

  • 01A blended option is a theory nobody actually holds.
  • 02When a compromise fails, both camps learn nothing and return to their original positions.
  • 03Compromise is only preferable to violence, not to testing a real idea.
  • 04Strong opinions loosely held beats weak opinions loosely held.
  • 05Someone has to own the idea being tested for the failure to teach anything.

How to run it

  1. 1

    Write the two theories down separately

    Get X and Y stated in the words of the people who hold them, without merging language. Most compromises begin with a summary that has already blurred the two positions.

    Pro tip Have each side write the other's theory and get it approved before you go further.

  2. 2

    Extract the prediction from each

    Ask each side what observable result their idea produces and by when. A theory that makes no prediction cannot be tested and does not belong in this process.

    Pro tip Force a number and a date; vague predictions survive any outcome.

    Watch out If both theories predict the same result, the disagreement is about something else — find out what.

  3. 3

    Refuse the blend

    Explicitly rule out the option that sits between X and Y unless it is a theory someone genuinely argues for on its own merits. Say out loud why: if Z fails, nobody learns anything.

    Pro tip Name the pattern in the room — it is far easier to reject a blend once everyone can see the trap.

    Watch out Blends are politically comfortable, which is exactly why they get chosen and why they teach nothing.

  4. 4

    Pick one and give it an owner

    Choose X or Y — by argument, by cost, or by who will be accountable — and let its holder run it properly rather than in a hedged half-version.

    Pro tip Let the person who believes it run it; a sceptic executing an idea half-heartedly produces an uninterpretable result.

    Watch out A hedged implementation is a compromise wearing a different hat.

  5. 5

    Agree the refutation condition in advance

    Before launch, write down the result that would mean the chosen theory is wrong, and the date you will check. This is what turns a decision into an error-correcting trial.

    Pro tip Put the condition in writing where both sides can see it, so nobody relitigates the criterion afterwards.

    Watch out Without a pre-agreed failure signal, a losing theory gets defended by moving the goalposts.

  6. 6

    Read the result and switch cleanly

    If the theory is refuted, discard it and run the rival next. Credit the correction, not the person who was right, so the next disagreement stays cheap.

    Pro tip Log what was learned so the same argument does not restart in six months.

    Watch out Punishing the owner of a refuted theory guarantees that the next disagreement ends in a compromise.

In the wild

The policy nobody believed in

Person A argues for idea X, person B argues for idea Y, and the group settles on Z somewhere in the middle. Neither A nor B ever thought Z was the best idea. When Z proves not to work, nobody is surprised and nobody updates: A says he always told them X was correct, B says she always told them Y was correct. The group has demonstrated only that Z is wrong — and no one ever thought Z was right in the first place. The disagreement survives fully intact, and the same argument resumes at the next meeting.

A cycle of failed compromises that produces no learning and no progress on the underlying question.

Two go-to-market theories, one quarter

A sales lead believes the company should move upmarket to enterprise; the founder believes self-serve is the growth engine. The compromise on the table is a hybrid motion that half-serves both. Instead the team writes each theory's prediction — enterprise: three signed pilots by quarter end; self-serve: a named activation rate — picks the self-serve theory because it is cheaper to refute, gives it to the founder to run undiluted, and fixes the failure threshold in writing.

The self-serve theory misses its threshold, is discarded without argument, and the enterprise theory runs next quarter with a real owner.

Common mistakes

Averaging two roadmaps into one

The merged plan inherits the costs of both theories and the conviction of neither, so when it underperforms nobody can say which assumption was wrong.

Strong opinions strongly held

Refusing to update once the refutation arrives is the opposite failure — it produces an intolerant minority who never error-correct.

Running the test with no failure signal

Without a pre-agreed condition for being wrong, every result is reinterpreted as partial success and the losing theory is never actually discarded.

Is it for you?

Best for

Leaders, founders and boards resolving genuine disagreements where both sides hold real theories about what will work.

Not ideal for

Disputes over values or allocation where there is no testable prediction, and situations where a real trial would be unsafe or irreversible.

From the transcript

So this is the poverty of compromise.

They've shown that Z is wrong, but no one ever thought that Z was correct in the first place.

Mark Andreessen summarizes this nicely as strong opinions loosely held.

From the episode

The Beginning of Infinity, Part 2