Skip to main content
The concierge test is beautifully simple: instead of building a product that does something automatically, you do it manually for a small number of customers first. You become the product. You’re the algorithm, the automation, the interface - all of it, by hand. It’s one of the cheapest and most powerful validation techniques in the discovery toolkit 🛠️

​
Where it comes from

The term comes from the lean startup world - Eric Ries discusses the principle throughout The Lean Startup as part of the build-measure-learn loop. The concierge test takes the “build the minimum to learn” idea to its logical extreme: what if the minimum is a human doing the job, not software at all? The classic example is Food on the Table, an early startup that built a meal planning and grocery app. Before writing any code, the founders manually created personalised meal plans and grocery lists for one family each week. They delivered real value. They learned what mattered. Then they built.

​
What you’re testing

The concierge test answers one question: if we delivered this value manually, would customers find it valuable enough to keep using it? It doesn’t test scalability, automation, or UX. It tests the core value proposition in its rawest form. If customers don’t find it valuable when a human delivers it personally and attentively, software won’t fix that 😬

​
When to use it

Concierge tests work best when:
  • The product involves a service or recommendation (curation, matching, personalisation)
  • You’re unsure whether customers actually want the outcome you’re promising
  • You want to learn deeply about the job before automating it
  • You need to validate willingness to pay before committing to a build
It’s less useful for products where the value is inherently in the software experience itself - a design tool, a game, a real-time dashboard.

​
How to run one

  1. Pick one customer - ideally someone who has the problem acutely. Don’t try to scale this yet.
  2. Deliver the value manually - do whatever the product would eventually do, by hand. Email them results. Message them on WhatsApp. Use a spreadsheet.
  3. Observe and ask - Did they use what you sent? Did it change their behaviour? What was missing? What did they ignore?
  4. Decide - Is the value real? Is there a path to automating this in a way customers would still find valuable?
The concierge test pairs naturally with pretotyping - both validate demand before building. The difference is that pretotyping tests whether people will take action to get the solution; the concierge test actually delivers it and measures whether they value the result. Lesson learned: the most surprising thing about concierge tests is how often the thing customers say they want isn’t the thing they actually use when you deliver it 👀 That gap is gold.