OPS.21:4.3 - Choose the comparison before choosing the number

Use the comparison whose meaning fits the aim and available evidence.

Comparison	When it helps	What must remain visible
An absolute result by a relevant date	A known population, recipient need or required service gives the number a useful meaning.	Eligibility, quality, timing and the work needed to reach the result.
A range or a qualified milestone	Precision would exceed the available knowledge, or progress is better demonstrated through a usable contribution.	What counts as adequate progress, what uncertainty remains and what decision follows each material outcome.
A ratio such as accepted work per usable hour	The relation between input and output answers the operating question.	Definitions of both quantities, omitted quality and shifted work; a ratio is not a comparison with peers.
Comparison with other operations	Their work, constraints and receiving results are sufficiently comparable for the intended inference.	Differences in mix, assets, starting conditions, shared support and observation coverage that could reverse that inference.
Improvement from a relevant earlier condition	Absolute peer levels would conceal important structural differences.	The original basis and changes in it. A low starting point can make improvement easier; percentages alone do not equalize opportunities.
Observation without a numerical improvement target	A condition is already adequate, the work is exploratory, or a quota would distort the useful contribution.	The condition or contribution to examine and the event that calls for a response.

Challenge a proposed comparison with an affected participant and one plausible unlike case. If a league table cannot distinguish superior work from easier conditions, use it to ask a learning question or narrow the comparison; do not use it to allocate credit. A complex adjustment is worthwhile only if its grounds are adequate and it can change the decision. An unexplained adjustment can hide the same unfairness as the raw number.