How to Run a Concierge MVP Before Writing Code

A concierge MVP is a highly manual service you sell to prove people will pay for a specific result before you write a single line of code. You are the algorithm. You do the work by hand. You deliver the outcome directly to the customer. If they will not pay you to solve the problem manually, they will not pay for your software to do it either.

Concierge versus Wizard of Oz

People confuse a concierge test with a Wizard of Oz test. They are different tools for different stages of a business. A Wizard of Oz test hides the human. The user interacts with what looks like a finished app, but you are behind the curtain frantically pulling levers to make it work. A concierge MVP is completely transparent. The customer knows they are hiring a human to do a job.

Wizard of Oz is useful when testing user experience or a specific interface interaction. Concierge is mandatory when you are still trying to figure out if the core problem is worth solving. Transparency builds trust. When the customer knows you are doing the work by hand, they will talk to you. They will explain their edge cases. They will tell you exactly what they hate about their current process. You cannot get that level of detail from a user who thinks they are clicking buttons in an automated web application.

Consider a service that categorizes business expenses. A Wizard of Oz test requires building a landing page, a fake dashboard, an upload button and a progress bar. You sit at your computer waiting for a file upload, manually categorize the rows, and push a button to update the fake dashboard. A concierge test requires an email address and a Stripe link. You tell the customer to email you their CSV file, you categorize it in Excel, and you email it back. The concierge test gets you to the truth faster with zero interface development.

You must charge from day one

Free users lie. They tell you your idea is great because being polite is cheaper than being honest. The only validation that matters is a credit card transaction. If you run a concierge MVP for free, you are running a charity. You are not running a business experiment.

Charge a premium. If your eventual software will cost 50 dollars a month, charge 500 dollars for the manual concierge version. You are providing a bespoke, manual service. It takes your time and your manual labor. High prices filter out the tire kickers and leave you with customers experiencing acute pain. Those are the exact people you want to learn from.

If you cannot find anyone willing to pay a premium for the manual solution, you have a weak problem. Software will not fix a weak problem. It will just make a useless solution scale faster. A high price point forces the customer to evaluate the real value of the outcome you provide.

Finding the right problem to solve

You do not start a concierge MVP with an app idea. You start with a problem people are actively trying to solve right now. You look for demand signals across forums, review sites and social media platforms. People asking for workarounds, hiring virtual assistants for specific data tasks or hacking together complex spreadsheets are all broadcasting demand signals.

If you want to skip the manual scraping of forum posts, Unthinkable surfaces these demand signals from across the web. You look at the raw demand, identify the manual workaround people are currently using and offer to do that exact workaround for them as a paid service. The goal is to insert yourself into an existing flow of money and frustration.

Serve exactly ten customers before writing code

Do not automate after one successful transaction. One customer is a fluke. Two is a coincidence. Ten paying customers is a repeatable business process.

Ten is the exact number of customers you need to map a reliable workflow. The first three customers will show you the happy path. Customers four through seven will introduce edge cases you never considered. The final three will prove you have stabilized the process.

If you write code before serving ten customers, you will build the wrong thing. You will automate the happy path and your software will break the moment a user does something unexpected. By doing the work manually ten times, you experience every edge case firsthand.

Let us do the math on a B2B lead generation service. You charge 500 dollars to deliver a list of 100 qualified leads. Ten customers generate 5000 dollars. You now have capital to fund your development time, a list of guaranteed beta testers and a perfectly documented workflow to automate. You also know exactly which data sources are reliable and which ones are full of dead links, because you checked them yourself.

How to map the manual workflow

While you deliver the service, you must document every click, every decision and every data transfer. This document becomes your product specification. You cannot automate a process you cannot describe.

Open a blank text file. Write down the trigger that starts the job. Write down the exact inputs you receive from the customer. Then, list every single action you take to transform those inputs into the final output.

Be painfully specific. Do not write a step that says "clean the data". Write "remove rows where the email column is blank, then lowercase all company names, then strip out legal suffixes like LLC and Inc".

When you finish serving your ten customers, look at your text file. Group the steps into three categories.

1. Steps that require human judgment.
2. Steps that move data from one place to another.
3. Steps that transform data based on simple rules.

You will build software to automate the second and third categories. The first category remains manual, or you restrict the scope of your software to exclude it. If a step requires you to read a paragraph and determine the emotional tone, you either drop that feature or prepare to integrate a language model. If a step requires copying a name from an email and pasting it into a spreadsheet, you write a script.

The signals that say it is time to build

You run the concierge MVP until the manual work becomes unbearable. You do not transition to software because you are bored. You transition because the manual process is breaking under the weight of customer demand.

Look for three specific signals.

1. Dropping balls. You are missing deadlines, forgetting to send update emails or making typos in the final deliverables. Your manual system can no longer handle the volume of requests.
2. A static workflow. The last three customers asked for the exact same outcome. You did not have to invent any new steps, rewrite your text file or handle any weird edge cases.
3. Negative margin on your time. Calculate your effective hourly rate. When the manual delivery takes so long that your effective hourly rate drops below what you could make at a standard day job, it is time to write code.

Building the software in phases

Do not build the entire system at once. Automate the most painful manual step first. This creates immediate leverage.

If your concierge service involves scraping a real estate website, formatting the pricing data in Excel and emailing a PDF report to investors, do not build a fully automated SaaS platform on day one. Build a Python script that does the scraping. You still do the Excel formatting and the emailing by hand.

Next week, write a script to format the Excel file automatically. You still send the email by hand.

Finally, automate the email delivery using an API provider like SendGrid or Postmark.

By replacing yourself one piece at a time, you ensure the customer never experiences a drop in quality. You maintain the revenue stream while gradually increasing your profit margins. The concierge MVP transitions smoothly into a software product, funded entirely by the customers who needed it most. You end up with a product that does exactly what the market wants, because the market paid you to design it.

Managing customer expectations

You must tell the customer exactly how the service works. Do not pretend you have a massive team or a proprietary artificial intelligence engine. Tell them they are working directly with the founder.

Position the manual nature of the service as an exclusive benefit. You are providing a white-glove experience. If they find an error, they do not have to submit a support ticket to a faceless help desk. They reply directly to your email, and you fix it immediately.

This positioning turns early adopters into advocates. They forgive manual delays because they value the direct access to you. Set clear turnaround times. If the manual work takes you four hours, promise delivery in forty-eight hours. This buffer protects you when life gets in the way or when a specific job turns out to be harder than expected. Over-delivering on a forty-eight hour promise is always better than missing a four-hour deadline.