Mystery Shopping Program: How to Set One Up

How to run a simple mystery shopping program for your own business: what to test, how often, how to score it fairly, and how to involve your team.

By the Mystery Shopper AI editorial team

Published November 13, 2026

Short answer: a mystery shopping program is a regular, planned set of anonymous tests of how your business treats customers. For a small business it can be simple: pick the channels customers use most (email, web form, phone), decide what a good answer looks like, test each channel on a schedule, score every test the same way, and share the results with your team as coaching, not as a trap.

Why run a program instead of one test

One test is a snapshot: maybe it landed on a great day or a terrible one. A program shows patterns:

  • which channel or location is slowest;
  • which questions your team struggles to answer;
  • whether changes (a new template, a new inbox owner) actually stuck.

Step 1: Choose what to test

Start with the moments that win or lose customers:

  • First enquiry by email or web form: speed and quality of the first reply.
  • Phone calls: how long it rings, how the call is answered, whether the question gets answered.
  • Follow-up: does anyone follow up if the customer goes quiet?
  • In person (for shops, clinics, restaurants): welcome, knowledge, next step.

If you can only do one, test the channel most of your new customers use first.

Step 2: Define what "good" means

Write a short scorecard before the first test, so every test is judged the same way. For an email enquiry, for example:

Criterion What good looks like
Speed first human reply within your target time
Answered every question each question answered clearly
Next step a booking link, a time, or a question back
Personal uses the customer's name and situation
Specific prices, dates, details instead of "it depends"
Writing clear, friendly, no errors

Weight the criteria by what matters most to your customers. Our mystery shopper checklist has a fuller version.

Step 3: Write realistic scenarios

Each test should be a believable customer question your team should be able to answer: a price, availability, what's included, a special request. Avoid scenarios that create real work (fake bookings or orders), complaints, emergencies, or anything that asks staff for personal information.

Vary scenarios between tests so staff can't recognise a pattern.

Step 4: Decide on a schedule

  • Monthly is a good default for a small business: frequent enough to spot problems, rare enough not to be a burden.
  • Random timing within the month, including quieter times like Friday afternoons.
  • After any change: new staff, new booking system, new opening hours.

Step 5: Tell your team the program exists

Testing in secret and then using the results against people destroys trust. A fairer approach that still shows real behaviour:

  • tell staff that customer-service checks happen from time to time, without saying when;
  • explain what is scored and why;
  • use results for coaching and to fix processes (who owns the inbox, which template to use);
  • share good results too.

Monitoring staff can be subject to data protection and employment rules depending on where you are. In the UK, for instance, the Information Commissioner's Office publishes guidance on monitoring workers (its page notes the guidance is under review). Check the rules that apply to you.

Step 6: Review and act

After each test:

  1. record the score and the reply time;
  2. note one thing that went well and one thing to change;
  3. fix the process, not just the person (a missing price list, a template, a notification);
  4. re-test next month to see whether it worked.

An example quarter for a small business

  • Month 1: one email test (price and availability question) and one phone test at lunchtime. Result: email answered in a day with no price; phone rang out. Fixes: a price range staff may share, and a lunchtime call rota.
  • Month 2: the same channels with new scenarios, one tested on a Friday afternoon. Result: email answered in two hours with a price and a booking link; phone answered on the fourth ring.
  • Month 3: a follow-up test: the "customer" goes quiet after the first reply. Result: nobody followed up. Fix: a reminder to follow up after two days, and a template.

Three months, six tests, and three process fixes that every customer benefits from. (An illustrative plan, not a real business.)

Common mistakes

  • No scorecard, so results depend on who reads the reply.
  • The same scenario every time.
  • Testing only on busy days or only on quiet days.
  • Punishing individuals for process problems.
  • Stopping after one good result.

Run the email part automatically

Mystery Shopper AI runs the email side of a program for your own business: each test enquiry is written from your website, sent at a random time from a fictional customer, and once you paste the reply it's timed and scored on the scorecard above, with tips and a stronger example reply. Pro adds a new automatic test every month. Mystery calls are in early access.

FAQ

How much does a mystery shopping program cost?

It ranges from free (doing it yourself) to agency programs priced per visit or call. Start small and expand what proves useful.

Should managers be tested too?

If they answer customers, yes. The program measures the customer experience, whoever delivers it.

How many tests do we need?

Enough to see a pattern: monthly tests over a quarter give a much better picture than one test.

How fast does your team really reply?

An AI mystery shopper emails your business a realistic enquiry at a random time, then times and scores the reply.

Test your response

Sources

Explore more guides