Test–retest reliability describes how consistently a test measures people on separate occasions when the characteristic being measured is expected to be stable. For fitness assessment, it helps you judge how much variation to expect before interpreting a later result as progress or decline.
A useful retest needs more than the same test name. You need a repeatable procedure, comparable conditions and an estimate of measurement error that fits your test and participants. This guide includes a copyable retest plan, a worked change example and a decision checklist.
First decide why you are repeating the test
Checking the repeatability of your method
In a reliability study, you repeat measurements in a group whose underlying performance is expected to remain reasonably stable. The question is: how much do scores vary when there should be little genuine change? Familiarisation, recovery and the time between sessions need to suit that question.
Monitoring a participant’s progress
In progress monitoring, genuine change may be expected after training or another intervention. You compare the new result with the baseline and interpret the difference using relevant measurement-error evidence. A before-and-after improvement alone does not establish the reliability of the test or prove that a training programme caused the improvement.
For the broader concepts of reliability, validity and standardisation, start with our fitness-testing reliability guide. This page focuses on planning and interpreting the next test.
How long should you wait before retesting?
There is no single interval that suits every fitness test or purpose. Choose the interval from the test protocol, relevant reliability research and the decision you need to make. Record why you chose it.
- For a reliability study: allow suitable recovery and address learning effects, while avoiding an interval in which meaningful underlying change is likely. A same-day repeat and a repeat on another day answer different questions.
- For progress monitoring: choose a review point when the result can inform a decision and plausible change could be distinguishable from measurement variation. Follow any programme-specific assessment requirements.
- When conditions have changed: review illness, injury, unusual fatigue, preparation, equipment faults or a protocol change before treating the visit as comparable. Apply the relevant screening and test procedures.
Do not repeat a demanding test immediately just because a result is disappointing. Use the protocol’s recovery, stopping and repeat-attempt rules. Record a postponed or non-comparable session rather than hiding the reason.
What should stay consistent?
- Protocol: exact version, instructions, warm-up, test order, rest, number of attempts and rules for invalid trials.
- Scoring: best attempt, mean or another defined score; units, rounding, equations and any normative reference.
- Equipment: device identity, setup, settings, software version and required checks. Record any replacement or update.
- Assessor: use the planned assessor arrangement and consistent instructions. Record changes in operator or technique.
- Participant preparation: familiarisation and protocol-relevant preparation, including recent activity and recovery.
- Environment: relevant surface, location, conditions and time of day. Document material differences.
Use the fitness test session planner to organise the visit and the equipment-check log to record checks. Standardisation reduces avoidable differences; it does not eliminate all measurement error.
Copyable retest plan
Copy these fields into your assessment record before the baseline visit. Use a participant or study ID rather than unnecessary personal details. This is a planning template; it does not replace screening, consent or the test protocol.
Purpose and timing
Participant or study ID: [ ]
Purpose: [method repeatability / progress review / other]
Decision the result will inform: [ ]
Baseline and planned retest dates: [ ]
Reason for interval and supporting protocol/source: [ ]
Familiarisation and recovery plan: [ ]
Method to repeat
Test and protocol version: [ ]
Outcome and units: [ ]
Measured or predicted outcome: [ ]
Attempts, rest and scoring rule: [ ]
Assessor, device ID, settings and software: [ ]
Required equipment checks and record reference: [ ]
Preparation, warm-up, order and environment: [ ]
Interpretation agreed in advance
Relevant error estimate and source: [value, units, method, population and protocol]
Why that evidence fits this use: [ ]
Meaningful-change criterion, if supported: [value, rationale and source / not established]
Reasons to postpone or flag a comparison: [ ]
Review owner and planned next action: [ ]
If suitable error evidence is unavailable, write “not established”. Do not turn a convenient percentage into a validated threshold. Use the assessment record and participant report templates to retain the raw results, deviations and interpretation.
What do reliability statistics tell you?
Pearson correlation describes association, not agreement. In an invented example, first-session scores of 20, 30 and 40 become 25, 35 and 45. Pearson’s r is 1.00, yet every score has increased by five units. A perfect correlation has not ruled out a systematic difference between sessions.
An intraclass correlation coefficient (ICC) needs its full specification. Report the model, single-measurement or average-measurement type, consistency or absolute-agreement definition, and confidence interval. The study design must fit your use: reliability for an average of several trials is not automatically reliability for one trial. Koo and Li’s reporting guide explains these choices.
A high ICC is not a universal pass mark for individual decisions. Consider the population, uncertainty in the estimate and the size of the error in useful units. It also does not establish validity or fitness to participate.
Absolute error and systematic differences need attention alongside ICC. Depending on the design, a study may report a standard error of measurement (SEM), typical error, coefficient of variation or limits of agreement. Check how each was calculated and whether error changes with the size of the score. These quantities are not interchangeable.
Worked example: is a two-centimetre change convincing?
This is an illustrative calculation, not a jump-test norm or a claim about a particular device.
- Baseline jump height: 30.0 cm.
- Retest jump height: 32.0 cm.
- Observed difference: +2.0 cm, or about 6.7% of baseline.
- Hypothetical, applicable SEM for each session score: 1.0 cm.
Under a simple model with independent, approximately normal errors of equal variance and no unaccounted systematic shift, a commonly used individual minimal detectable change at 95% confidence is:
MDC95 = 1.96 × √2 × SEM = 1.96 × √2 × 1.0 ≈ 2.77 cm.
The observed 2.0 cm increase does not exceed that illustrative threshold. You can report the higher recorded score, but this model would not classify the difference as exceeding expected measurement error at that level. It does not prove that no change occurred.
Exceeding an MDC would still not establish that the change matters to the participant or that training caused it. De Vet and colleagues distinguish detectable from important change; their paper concerns health-status questionnaires, so it supplies the distinction, not a jump-specific cutoff.
Explore calculations and their assumptions with our measurement-error guide and calculator. Use an error estimate for the actual protocol and scoring method, not the illustrative 1.0 cm above.
A decision checklist for the retest visit
- Confirm whether to proceed. Follow screening, preparation and protocol requirements. Record reasons for postponement or stopping.
- Check comparability. Review the original plan, device checks and any changes in preparation, method or conditions.
- Retain the observations. Save individual trials, invalid attempts and reasons, then apply the agreed scoring rule.
- Report the difference. Include both scores, dates and units. Label predictions as predictions and explain material deviations.
- Judge uncertainty and importance separately. State the applicable error evidence and the basis for any meaningful-change criterion. Flag missing evidence.
- Agree the next action. Record the interpretation, its limitations and the reason for the next review date.
A useful report might say: “The recorded jump score increased from 30.0 to 32.0 cm using the same scoring rule. The 2.0 cm difference is below the illustrative MDC95 of 2.77 cm. We cannot confidently distinguish it from measurement variation using this estimate.”
If you are running your own reliability study
Two visits from one participant cannot establish a general reliability coefficient. Define the target population, sample-size rationale, repeated-measurement design, stability assumptions and analysis before collecting data. Record missing or excluded measurements and any departures from the plan. Use the GRRAS reporting guidance and appropriate statistical support for a formal study.
Put the retest plan into your assessment workflow
For a Chester Step Test follow-up, use the specific test protocol and results interpretation guide. The CST2 Chester Step Test Calculator supports that test’s calculation and reporting workflow; software cannot make two different protocols comparable.
For equipment planning, discuss your testing setup. For broader assessment software, request Cartwright Assess details and confirm the functions currently available for your workflow.
