How Many Customer Interviews to Conduct Before Building

You need exactly 12 customer interviews before writing a single line of code. Stop at 12. If you have not found a burning problem after talking to a dozen people in your target market, the problem does not exist. If you have found a problem, talking to a thirteenth person is a procrastination tactic to avoid opening your code editor.

Why Twelve is the Magic Number

Let us look at the mechanics of qualitative research. In usability testing, Jakob Nielsen proved that five users uncover 85 percent of usability problems. Customer discovery follows a similar curve of diminishing returns. By your sixth interview, you will start hearing the exact same phrases. By your tenth, you can finish their sentences. Twelve gives you a buffer for two complete duds. It provides a sample size large enough to spot a trend but small enough that you can conduct them all in a single week. If you interview 12 freelance graphic designers about how they collect unpaid invoices and nine of them describe the exact same manual spreadsheet process, you have a product. If all 12 give you different answers, your niche is too broad. You do not need more interviews. You need a tighter target audience.

The Anatomy of a Useless Interview

Most founders conduct 50 interviews and learn absolutely nothing. They fail because they ask about the future. The future is a polite fiction. If you ask someone if they would use a product, they will say yes to avoid hurting your feelings. A useless interview is filled with toxic, hypothetical questions. Would you pay fifty dollars a month for this? Do you think this is a good idea? If we built a feature that did this, would that save you time? These questions force the user to guess their future behavior. Humans are terrible at predicting their future behavior. We all think we will go to the gym tomorrow. We all think we will buy the productivity software next week. The customer wants to be helpful. They want to be the kind of person who buys your smart, efficient tool. But wanting to be that person and actually entering a credit card number are entirely different states of being. A useless interview leaves you feeling highly motivated and completely devoid of actionable data. The founder just validates their own ego. The customer has committed nothing, spent nothing, and revealed nothing about their actual workflow. Hypothetical questions breed hypothetical businesses.

The Three Questions That Actually Matter

You are not looking for validation. You are looking for past behavior. People lie about what they will do. They cannot lie about what they have already done. Question one is about recency. Ask when they last dealt with this problem. If the answer is not within the last month, the problem is not acute enough to monetize. Question two is about friction. Ask them to show you exactly how they solve this right now. This is where you find your feature list. If they do not have a workaround, they do not have a problem. A workaround might be a messy Excel file, a physical notebook, or a string of Zapier automations. The uglier the workaround, the stronger the demand. Question three is about budget. Ask what else they have tried and why it failed. This maps your graveyard of competitors. If they tried an existing software tool and churned, you need to know why. If they say it was too expensive, you are in trouble. Competing on price is a race to the bottom. If they say it was too complex, you have a wedge.

Sourcing the Twelve

Finding twelve people is a mechanical process. Do not interview your friends. Do not interview your former coworkers. You need strangers who match your exact ideal customer profile. Start by looking for demand signals across the web. Go to Reddit, Hacker News, or specialized forums. Look for people actively asking for solutions. Once you find someone actively experiencing the problem you want to solve, send a cold message. Keep the outreach short. State that you are researching the space, not selling a product. Ask for 15 minutes. Offer a twenty dollar coffee gift card in exchange for their time. Do not ask them to book a time on your calendar yet. Just ask if they are open to it. If you send 100 personalized messages to people actively experiencing the problem, you will get about 25 positive replies. From those 25, you will successfully book and complete 12 calls. If you cannot book 12 calls out of 100 highly targeted messages, the problem is not painful enough to warrant a product. The market is telling you to move on before you write a line of code.

The Arithmetic of Validation

Let us do the math on what a successful batch of 12 interviews looks like. You need a strict threshold for moving forward. Out of 12 interviews, you need at least eight people to confirm they experience the problem regularly. Of those eight, you need at least six to show you an active, painful workaround. Of those six, you need at least four to have spent money trying to solve it in the past. If you hit those numbers, you have eight experiencing it, six hacking a solution, and four paying for a solution. You have a validated idea. Build it. If you get 12 people who say the problem is annoying but none of them have spent a dime or an hour trying to fix it, stop. You have found a mild inconvenience. Mild inconveniences do not sustain software businesses.

Running the Room

When you get on the call, shut up. The founder should be speaking for less than 10 percent of the interview. Record the call using a transcription tool. Taking notes manually forces you to break eye contact and interrupts the flow of conversation. Start with a broad prompt about their workflow. Ask them to walk you through their process for onboarding a new client. Listen for the friction points. When they mention a frustration, drill down. Ask them to tell you more about that specific step. Ask why that step is so manual. If they say they spend three hours every Friday reconciling invoices, do not immediately offer your automated invoicing tool. Ask them to share their screen and show you the exact process. Watch them open the three different browser tabs. Watch them copy and paste data between Stripe and Excel. The exact sequence of their clicks is the blueprint for your product. Never pitch your solution. If you mention your idea before the last two minutes of the call, you have corrupted the data. The moment you introduce your product, the customer switches from describing their reality to evaluating your idea. Keep them in their reality.

Documenting the Reality

Once you finish the 12 interviews, synthesis is mandatory. Do not rely on your memory. Your brain will naturally amplify the one interview that agreed with your preconceived notion and forget the eleven that told you the idea was useless. Create a grid. Put your 12 subjects in the rows. Put your key criteria in the columns. The columns should be frequency of problem, current workaround, money spent, and exact quotes describing the pain. Fill this grid out immediately after each call. When you have completed all 12, look at the board. The patterns will be blindingly obvious. You will see the exact same phrase repeated across multiple rows. That phrase is your landing page headline. You will see the exact same messy spreadsheet mentioned three times. That spreadsheet is your core feature.

The Cost of Ignoring the Limit

Why not do 20 interviews? Or 50? Because momentum is the most fragile asset a founder has. Customer discovery is comfortable. Writing code that might fail is terrifying. Founders use endless customer interviews as a shield against the market. As long as you are doing research, you cannot be rejected. But research does not generate revenue. Twelve interviews take exactly one week to schedule and conduct. One week of research is responsible. Two months of research is cowardice. You will learn more from launching a terrible version one to those 12 people than you will from conducting 40 more interviews. Get your 12 calls. Ask about the past. Look for the workarounds. Count the money spent. If the math works out, close your notebook and open your terminal.