Ten beauty salons in Almaty have signed the terms of a pilot: for four weeks a future booking service sends them new clients through a page of free slots, and the salon pays 1,000 ₸ for each one who turns up. A developer is ready to build a catalogue, a calendar and apps for iPhone and Android, but that will take four months. The founder decides to test the idea in four weeks and without a developer. The numbers are made up.
What an MVP is
In a 2009 blog post Eric Ries defines the minimum viable product, or MVP, as the version of a new product that lets a team collect the maximum amount of validated learning about customers with the least effort. He repeated the definition in "The Lean Startup" (2011), and Lean Startup Co. quotes it too. The key word is learning. You build an MVP to answer a question, not to ship a cut-down first version.
Two things follow. How minimal the product is depends on the question, not on a feature list: a page with a form answers one question, another needs a working calendar. And viable means the customer gets what was promised. If a booking gets lost, the test has measured your sloppiness, not the idea.
Which test answers which question
| Question | Test | What it shows | What it does not show |
|---|---|---|---|
| How people deal with the job today | A conversation about the last time | How the person actually chose, what they used, what annoyed them | How often others do the same, and whether they will pay |
| Whether people respond to the promise | A page with the offer and a form | How many people left a contact in this channel | Whether they will use the product and pay for it |
| Whether they understand the flow | A clickable mockup | Which screen people get stuck on | Whether the product is needed and whether anyone will pay |
| Whether they will commit | Signed terms or a prepayment | Who is ready to commit | Whether they will use it and renew |
| Whether they pay for the result | A paid pilot where a person does the work and the customer knows it (concierge) | Who pays, and what it costs to keep the promise | Retention and economics once there are a hundred times more customers |
| How people behave with a finished product | The interface looks finished, but a person works behind it (Wizard of Oz) | What people do with a product that does not exist yet | What real automation will cost |
The difference between the last two rows is whether the customer knows. In a concierge test they do: the value is delivered by hand, as a personal service, to a small number of real customers. In a Wizard of Oz test the person believes the product is automated. If you charge for such a test, the Kromatic handbook advises refunding people for what is not built yet and telling participants afterwards how it really worked.
The rows run from cheap tests to expensive ones, but you do not have to go through them in order. The booking service's question is the one in the fifth row: whether salons will pay for new clients. So it needs a paid pilot with the work done by hand. A team poll on which screen looks more convincing will not answer it: the team is not the one paying. A full platform will not answer it better either, only four months later and at a higher cost.
The picture shows the same move on a different service. To find out whether buyers will pay for hand-picked suppliers, the first MVP is a limited paid pilot: a person builds the shortlist for each request, the customer pays for the result, and automation gets built once it is clear that people pay for the result.
How the pilot was set up
The page took two days on a website builder: ten salons, each with its free slots for the week ahead and a booking form. The founder updated the slots by hand every morning and evening from the receptionists' messages. A request from the form landed in his WhatsApp; he called the salon, confirmed the time and messaged the client. Under the form it said that a person confirms the booking within 15 minutes. Clients knew how it worked, and the salons certainly did.
Each salon also had a page of its own, with only that salon's slots. It was made for ads about a single salon, and nobody expected much from it.
Instagram ads for 80,000 ₸ pointed at the page, and the 260 people on the waiting list got the link on WhatsApp, as promised.
In the second week two clients turned up for the same slot: the founder had not closed it in time that evening. After that he closed a slot as soon as he had called the salon, before replying to the client. That is what "viable" means in practice. One botched booking and the salon drops out of the pilot, and you are left guessing whether the problem was the idea or the double booking.
Write the success threshold down before you start
| What we count | Threshold |
|---|---|
| How many salons paid the invoice | At least 6 out of 10 |
| How many new clients turned up | At least 60, six per salon |
| What one client who turned up cost in ads | No more than 1,000 ₸, otherwise every booking loses money |
| How many hours went into manual work | Logged daily, no threshold |
Six out of ten is not a standard; it is the bar below which the founder decided not to continue. With ten salons you cannot be more precise. The last row gets counted even without a threshold: manual work in a pilot is the future cost of the product, and you need to know it before anyone writes code.
What happened
| Metric | Threshold | Actual |
|---|---|---|
| Booking requests from new clients | - | 41 |
| New clients who turned up | 60 | 26 |
| Ad cost per client who turned up | Up to 1,000 ₸ | 3,077 ₸ |
| Salons that paid the invoice | 6 out of 10 | 8 out of 10 |
| Service revenue | - | 23,000 ₸ |
| Bookings by regulars through the salons' own pages | Not counted | 340 |
| Founder's hours | - | About 70 |
Eight salons out of ten paid, so that threshold was passed. But there was hardly anything to pay for: invoices came to 26,000 ₸ in total, and two salons refused to pay because they recognised three of their "new" clients as old ones. The service earned 23,000 ₸ for the month, under 3,000 ₸ per paying salon.
Twenty-six new clients turned up instead of 60, and each cost three times more in ads than the salon paid for her. Even if you put all 26 down to the ads, including the ones who came from the waiting-list link, that is 80,000 / 26 = 3,077 ₸ against 1,000 ₸ of revenue.
The biggest number in the table is the one nobody planned. Receptionists started sending the link to their own salon's page to regular clients and put it in the salon's Instagram bio. Those pages took 340 bookings, against 41 requests from new clients: eight times as many. The service earned nothing from them: under the terms, salons only paid for new clients. In the last week seven receptionists out of ten asked whether they could keep the page after the pilot.
And the hours. About 70 hours over four weeks, or 17-18 hours a week, and more than half of it went on calls about regulars' bookings. For ten salons that is almost half of one person's working week. For a hundred salons it would not work at all.
How to read the result
The pilot answered the question it was run for: the "salon pays for a new client" model does not add up in this form. Spending four months on an app for it would mean building a product that brings in 1,000 ₸ per client and spends more than 3,000 ₸ on ads to get her.
But the MVP produced an observation nobody had planned: it looks as if salons need online booking for their own clients. Money has not confirmed it yet, since nobody paid for those 340 bookings. It is the next hypothesis, and it has to be tested the same way: a paid pilot with a threshold written down in advance. Keeping the booking page and closing the catalogue is exactly what a pivot is: the learning from the pilot stays, and one main assumption changes.
The classic mistake at this point is to declare the pilot a failure and shut the project down, or the opposite, to declare it a success because eight out of ten paid. The table contains both, which is why thresholds are written down before the start: they stop you from picking the convenient row.
Exercise: an MVP for your own idea
A designer shows an app mockup to the owners of 30 salons, and 25 of them say they would sign up. Your partner wants to start development. What do you tell him?
"I would sign up" is an opinion about a picture, and it costs nothing to give. Even signed pilot terms did not fully answer the money question: eight salons out of ten paid, but on average less than 3,000 ₸, and the real answer came from the number of new clients and what each cost in ads. Before development you need a limited paid pilot: a few salons, manual work instead of an app, a price and a threshold written down in advance. Invite those 25 into it. How many of them sign the terms and pay is your answer.
Video: the minimum viable product
MVPs explained through examples of real products.