# The concierge MVP: do it by hand before you write code

> Why it pays to deliver the service manually first, what that teaches you, and how to tell which part of the process is ready to automate.

Published: 2025-07-29 · Source: https://subxcat.com/articles/concierge-mvp · Author: SUBXCAT

The most expensive way to test an idea is to build the whole product and see whether anyone needs it. The cheapest is to deliver what the product promises, by hand, to a few customers. That's the concierge MVP: the customer gets the result, and behind it there's no system, just you with a spreadsheet, email and a phone.

## Concierge versus Wizard of Oz

There are two similar approaches, and it helps to tell them apart.

In the concierge version, the customer knows a person is doing the work. You say it plainly: we're doing this by hand for now, to learn how to do it well. The customer talks to you and sees the process, and you see their reaction at every step.

In the second approach, everything looks like a product from the outside, a form, a button, a result, while inside a person does by hand what code will do later. That helps when testing how people use an interface. The concierge is more useful at the very start, when it isn't clear yet what needs doing.

## What to do by hand

Take the product's promise and deliver it any way you can. A matching service: do the matching yourself. Analytics for online stores: export the data, crunch it in a spreadsheet, email the report. A scheduling tool: build the schedules and send the reminders manually.

A few rules:

- **Charge money.** Not much, but not zero. Many will say yes to free, few will pay, and only the payers tell the truth about value.
- **Cap the number of customers.** Five to ten is enough to see what repeats, and few enough not to drown.
- **Write down every step.** What you did, how long it took, where you had to think and where you followed a routine.
- **Don't build something unique for each one.** If every customer needs something different, that's an agency, not a future product. Look for the common part.

## What it teaches you

Manual work shows you things interviews don't.

Which steps are actually needed. On paper the process has five steps. In practice it has nine, and three of them deal with exceptions you never thought of.

What the customer values. Often not what you thought mattered most. You were planning a clever matching algorithm, and what people care about most is getting a reminder on time.

What it costs you. If the manual version takes an hour per customer, you know how much automation has to save and how much it's sensible to spend on it.

How customers talk. Your email threads with them are ready-made copy for the future site and interface.

And above all, whether they come back. A customer who returns a second and third time for the same manual service is a strong signal. One who tried it and vanished is a signal too, just a different one.

## When to automate

Not everything at once. Go through your notes and find the step you repeat identically, over and over, that eats the most time. Automate that one first. Keep doing the rest by hand for now.

Signs a step is ready for automation:

- you follow a set procedure without making decisions;
- its exceptions are known and written down;
- customers don't notice who does it.

Steps where you have to think every time are too early to automate: you don't yet know what rules to build in.

## Traps

The main trap is getting stuck. Manual work is comfortable: customers are happy, money comes in, and it feels like everything works. But if six months later you're doing by hand exactly what you did in month one, it's no longer an experiment. It's a service that doesn't scale. Give yourself a deadline and a question the concierge has to answer by then.

The second trap is the opposite: automating too early because manual work is boring. Any process turned into code before it settles will have to be rewritten.

SUBXCAT can help see this through: work out which steps of the manual version should become product, and build them to hold up as you grow.