Mystery shopping: measuring what your customer actually experiences

1. The problem: the gap between procedure and reality

Every organisation with direct public contact has standards: how the customer is greeted, how quickly the phone is answered, which information is given, how a complaint is handled. Managers know the standards, employees have learned them, and satisfaction surveys show acceptable scores. And yet complaints concentrate in the same locations, add-on sales do not materialise, and a customer walking into a branch experiences something other than what the manual says. Mystery shopping is the method built for this gap: it does not ask customers what they felt, it sends a trained observer to live the experience and score it criterion by criterion.

It is a structured observation method, not a survey, and that gives it both its strength and its limits. The strength: it measures concrete behaviour, not impressions, under controlled conditions comparable across locations. The limit: it does not say what real customers feel, only what happens to them; for perception, the instrument is the satisfaction study, and the two complement each other.

2. When mystery shopping is used

• Checking standards in networks with many points of contact: banks, retail, pharmacies, restaurants, petrol stations, public services. The question is "is the standard followed in practice, and where is it not?".
• Evaluating training: after a training programme, mystery shopping shows whether what was learned is applied at the counter.
• Cross-selling and advice: do employees offer the complementary product, ask the needs questions, explain the benefits? These are behaviours the customer cannot evaluate, because they do not know what they should have received.
• Comparison with competitors: the same scenarios applied in your own locations and in competitors' show where the experience differs.
• Testing a new process: a sales script, an account-opening flow, a returns procedure, before rolling them out across the network.
• Public services: waiting time, clarity of information, equal treatment of applicants, compliance with procedures in institutions with service counters.

Mystery shopping is not used to measure satisfaction, awareness or purchase intention, to evaluate a specific employee with a view to sanctions, or in contexts where the fictitious customer would consume scarce resources, for example emergency medical services.

3. The scenario: what the evaluator does

The scenario describes exactly the experience the evaluator goes through: which customer they are, what need they come with, what they ask, how they react to proposals, whether or not they complete the transaction, how long they stay. A good scenario is realistic, because employees detect customers who are "acting", standardised enough for visits to be comparable, and leaves room for the employee's behaviour: the evaluator does not lead the interaction, they receive it.

A programme usually has two to four scenarios covering the situations that matter to the organisation: the new customer asking for information, the existing customer with a problem, the customer hesitating between two products, the customer who calls. Scenarios are built together with the client from the existing standards and tested in a few pilot visits, to check realism and duration.

4. The evaluation grid: observable criteria

The evaluation grid is the measuring instrument and is built according to strict rules. Each criterion describes an observable behaviour, not an impression: "the employee greeted within the first 10 seconds" can be verified; "the employee was friendly" is an opinion. Criteria are worded as yes/no whenever possible, and on short scales of 3-5 points where degree matters, for example the clarity of an explanation. The grid is completed immediately after the visit, within 15-30 minutes at most, so that memory is accurate, and includes free-text fields for context.

Weights reflect the importance of each dimension for the customer experience and for commercial objectives, and are set before the first visit, not after the results appear. The final score of a visit is the weighted sum, and the score of a location is the average of its visits.

5. How many visits and when

A single visit measures one employee on one day, not a location. For the score to be a property of the location, 3-6 visits per location per wave are usually needed, spread across different days of the week, different time slots and, where relevant, different employees. For a chain with 40 locations, a wave means 150-250 visits. Waves are repeated quarterly or half-yearly, with the same scenarios and the same grid, so that change is measurable.

All locations are visited, not a sample, when the purpose is network management: every branch manager must receive their own score. Sampling locations is appropriate only for general diagnostic studies or for comparison with competitors.

6. Evaluators and objectivity

The quality of the programme depends on the evaluators: people selected to match the profile of the location's typical customer, trained on the scenarios and the grid, and checked. An evaluator who resembles the usual clientele is not detected; one who does not, for example a student in a private banking branch, compromises the visit. Each evaluator makes a limited number of visits to the same network, to avoid being recognised, and training includes calibration exercises: all evaluators score the same recorded interaction and the differences between them are discussed until they score alike.

Quality control of visits includes proof that the visit took place, a receipt, a photograph of the exterior, entry and exit times, a check of the grid's consistency with the free-text description and, where the law allows and the client decides, audio recording. Incomplete or suspicious visits are redone.

7. Results: from score to training

The report of a mystery shopping wave has three levels. The network score, by dimension and by criterion, with the change from the previous wave. The scores by location, ranked, identifying those below the threshold and the dimensions on which each loses points. And the qualitative observations: what concretely happens in the weak locations, what the good ones do differently, which phrasings work and which generate resistance.

The proper use of results is in training and processes: criteria with low scores across the network show what is missing from training or from the standard; locations with low scores receive coaching; the practices of good locations are documented and spread. Linking scores to bonuses is possible but risky: employees learn to recognise evaluators and perform for them, and the score stops describing the real customer's experience. A mystery shopping programme that produces sanctions is sabotaged from within in two waves.

8. Variants: physical, telephone, online, competitive

• Physical visits: the standard, for retail, banks, public services.
• Telephone calls: for call centres and support lines; they allow large volumes and full recording.
• Online interactions: forms, chat, e-mail, digital processes; they measure response time, accuracy and tone.
• Competitive mystery shopping: the same scenarios at competitors; results are reported in aggregate, without identifying employees, and show where the experience, not the price, can win.
• Accompanied shopping: a researcher accompanies a real customer and observes the decision in context; it is a complementary qualitative method, described on the monitoring and testing page.

9. What to ask the agency

• The scenarios and the grid, built from your standards and approved before the pilot.
• The criterion weights, set before the first visit.
• The number of visits per location, their distribution over time and the wave calendar.
• The evaluator profile, training and calibration exercises.
• Proof of visit and the procedure for rejected visits.
• Compliance with labour and data protection law in the treatment of observed employees.
• The report format by network, by location and qualitative, and its delivery to location managers.
• Training and process recommendations, not just scores.

Of all research methods, mystery shopping is the closest to operations: it measures what can be changed on Monday morning. Done properly, with realistic scenarios, an objective grid, enough visits and use for training, it is the most direct instrument for improving the customer experience. For choosing between it and other methods, see the guide Which type of research solves your problem.

See also: Mystery Shopping Monitoring and testing Customer satisfaction and NPS Quantitative research Decision support Which type of research solves your problem

The Perspicuus research team · 7 June 2026