Four decisions determine whether a mystery shopping programme produces data you can act on: sample size, frequency, daypart coverage and reporting units. Each is settled before fieldwork or it is settled badly.
Reporting units come first
Decide what you want to compare before you decide how many visits to buy. A single national score, a comparison between regions, a branch league table and a format comparison are four different programmes with four different costs, and the difference between them is an order of magnitude rather than a margin.
The rule is simple: size the sample against the smallest group you intend to act on. If you will move budget between individual branches, the branch is the unit. If you will only ever brief the network as a whole, the network is the unit and you can buy a much smaller programme.
Sample size
For a single location, three visits in a wave is the practical floor. One visit measures one shift on one day with one staff member; two visits that disagree tell you nothing about which is representative. Three begins to distinguish a pattern from an incident.
Across a network, the useful figure is visits per location per wave rather than total visits. A thousand visits spread across four hundred locations is a national average with no location-level finding in it at all.
Frequency
Frequency should follow how fast the thing you are measuring changes. Service quality changes with the shift rota, which means a quarterly programme measures four rotas a year and reports findings that are already three months old when they arrive.
Monthly waves across varied dayparts produce a trend that supports operational decisions. Quarterly waves produce a document for a review meeting. Both are legitimate purchases; they are not the same purchase, and clients frequently buy the second while expecting the first.
Daypart and day-of-week coverage
This is the decision most often left to the field team and it should not be. Weekend and weekday performance in Qatari malls diverge sharply. A forecourt after 22:00 is a different operation from the same site at noon. A restaurant on a Friday evening bears little resemblance to the same restaurant on a Monday lunchtime.
A programme that samples whichever slot was convenient produces a comfortable number. Specify the distribution in the brief, monitor it during fieldwork, and report the achieved distribution alongside the score so that the reader knows what the number covers.
What to leave out
The commonest fault in a first-time scorecard is length. A ninety-criterion instrument for a two-minute counter transaction produces shopper fatigue, guessed answers and a score that averages away everything that mattered.
Cut every criterion you would not act on. If a branch fails it and nothing happens, it is not a criterion, it is a note. Twenty-five criteria that each have an owner beat ninety that have none.
Budget the second wave before the first
A single wave gives you a position. It cannot give you a trend, and a trend is what changes behaviour, because the second report is the one where a manager sees their own number move. Programmes commissioned as a one-off pilot and then not renewed are the most common way this work fails to deliver, and the cause is a budgeting decision rather than a research one.