Section 012 · Chapter 2, From Tests to Release Evidence

Pairwise Comparison

When absolute scoring is hard, asking which output is better can produce useful evidence.

pairwise comparison

What to do

  1. Review disagreement cases, especially when the judge chooses a fluent but factually weaker answer.
  2. Version B adds a deterministic fake clock and asserts that exactly three retries happen before the checkout flow gives up.
  3. Version A hides the timing problem and makes the suite slower.
  4. Version B explains the behavior and makes the test more stable.

Evidence to preserve

  • Preserve the inputs, versions, configurations, raw outcomes, and results for pairwise comparison needed to reproduce work on Pairwise Comparison.
  • Report results for pairwise comparison by relevant slice, separate blocker failures from averages, state uncertainty and blind spots, and connect the result to a release decision.

Expert note

Blind the version labels, randomize side order, allow ties, and analyze win rate by category. Pairwise wins do not replace absolute gates because the better of two bad outputs can still be unacceptable.

Continue the conversation

Apply this to your context.

Save your product context once, then open a focused conversation that combines it with this concept.

Cite this page

Jason Arbon. "Pairwise Comparison." Testing AI Knowledge Edition, section 12.

https://jarbon.ai/testing-ai/knowledge/ch012-pairwise-comparison.html

Shared across the Knowledge Edition

Adapt every concept to your world.

Describe your product, role, users, risks, constraints, or current quality problem. This stays in this browser until you choose to send it to ChatGPT.

Saved only in this browser.0 / 2400