A mystery shopping benchmark is useful only when the reader knows what is being compared and whether the underlying assessments are sufficiently alike. A percentage without its questions, sample and scoring rules is not a reliable performance reference.
This guide explains comparison methods. It does not publish national averages, sector league tables or confidential client results.
Identify the type of reference
| Reference | What it can help answer | What to disclose |
|---|---|---|
| Previous performance | Has a comparable assessed behaviour changed over time? | Periods, scenario and instrument changes, sample composition and visit counts. |
| Internal peers | Where do comparable teams or sites differ? | Customer task, channel, operating context and eligible observations. |
| Program target | Did performance meet the agreed expectation? | Who set the target, why it matters and whether it is a policy expectation or observed result. |
| External or sector reference | How does a suitable disclosed cohort compare? | Dataset provenance, period, population, permission, definitions and methodological differences. |
A target of 90% is an expectation, not evidence that an industry averages 90%. A provider’s own program results are not automatically representative of all Australian businesses.
Check comparability before calculating a gap
| Check | Question to answer |
|---|---|
| Construct | Do both measures assess the same behaviour or customer outcome? |
| Scenario and opportunity | Did the customer task create a comparable chance to demonstrate it? |
| Channel and context | Are phone, email and in-person results separated where their conditions differ? |
| Period | Are seasonality, changes in operating conditions and the timing of interventions visible? |
| Sample | How many relevant visits were completed, where and when? Were they selected for a particular risk? |
| Scoring | Are anchors, weights, critical rules, exclusions and rounding consistent? |
| Evidence quality | Were narratives and material factual checks reviewed to a comparable standard? |
Worked example: the same behaviours produce different totals
Fictional calculation: three criteria receive Green, Amber and Red. In this example those ratings are mapped to 2, 1 and 0 points. With equal importance, the total is 3/6 = 50%.
| Criterion | Rating | Normalised result | Alternative weight | Weighted contribution |
|---|---|---|---|---|
| Acknowledgement | Green | 100% | 20% | 20 percentage points |
| Needs discovery | Amber | 50% | 30% | 15 percentage points |
| Next step | Red | 0% | 50% | 0 percentage points |
| Total | Same observed behaviours | Equal-weight result: 50% | 100% | Alternative weighted result: 35% |
The change from 50% to 35% comes entirely from weighting. It is not a decline in the observed behaviours. When an instrument changes, show criterion-level results on a common basis or mark the break in the series rather than treating the totals as directly comparable.
Show the denominator and distribution
Four successful observations from five opportunities and forty from fifty both equal 80%, but the amount of evidence differs. Neither ratio alone establishes a stable difference between sites or representative population performance. Show visit counts, eligible opportunities and the spread of results, not only an average.
Document authorised N/A exclusions. A non-purchase brief has no checkout opportunity; a missed next-step explanation in a relevant enquiry is a different situation. Excluding the latter after fieldwork can inflate the result and defeat comparison.
Keep serious exceptions and selection effects visible
A high average can coexist with an important failure. Report critical items separately under the approved rules. Likewise, visits targeted at known concerns answer a diagnostic question and should not be presented as a random national sample.
Repeated observations strengthen a practical pattern, but they do not automatically justify a statistical confidence claim. Use an appropriate research design and qualified analysis before making population estimates or assigning precision to a purposively selected set of visits.
Use the comparison to choose an investigation
Ask which behaviour explains the difference, whether the difference survives a like-for-like comparison and what other evidence could confirm the explanation. Assign a proportionate action and retest after the operating change. A before-and-after rise alone does not prove the program caused the improvement.
For individual assessment controls see methodology and quality assurance. For presenting measures to management see business intelligence.
Sources and scope
Prepared by Secret Customer Australia · Published 13 November 2026. Adapts benchmark-governance and longitudinal-comparison principles from SC-RD-033 and SC-RD-034, scoring guidance from SC-M09 and selected benchmarking handbook concepts. Internal client series and confidential comparison datasets are excluded. All calculations above are fictional educational examples, not measured sector results.
Browse the Secret Customer Academy or discuss your objectives through our mystery shopping services.
