Voice of customer research: Start with the decision, then work through the evidence

Voice of customer research works backwards from a decision. Write down the decision you have to make, pick the sources that can answer it, review the evidence you already have before booking new customer interviews, and stop when new sources stop changing your mind. Most of the answer is already sitting in recordings and tickets you own.

You have just bought a company, or you sit on the platform team of the fund that did: 40 to 200 people, 600+ customers, 1 or 2 marketers, no CMO. Somebody has told you to talk to the customers before you touch the website.

The common failure starts there. Somebody books 15 to 20 calls, the transcripts pile up, and 6 weeks later nobody can say what changed.

Key takeaways

• Name the decision first. A research question with no decision attached produces a report nobody uses.
• Start with the evidence you already have, so customer interviews test a hypothesis instead of building one from scratch.
• Agree what counts as evidence before you start reading, or you will find what you already believed.
• Stop when 3 new sources in a row tell you nothing new.

Define the business decision before you collect customer evidence

Every useful research plan begins with a sentence in this shape: We need to decide X by Y, and we are missing Z.

Compare two briefs. The weak one says we want to understand our customers better. It has no end point, so it never ends. The strong one says we need to decide by month end whether the website speaks to operations managers or to finance, and we do not know which of them drives the purchase. That one tells you who to speak to, what to ask, and when you are finished.

Write the decision at the top of the document before anything else goes in it. If nobody can name a decision that the outcome would change, the research is a reading exercise and the budget is better spent elsewhere.

Six sources of customer evidence to review before booking interviews

Interviews are the source everyone thinks of first, but they take more time to organise and run. Before booking any, start with the evidence you already have.

1. Recorded sales calls. Free, already in your tools, and full of buyers describing the problem in their own words before anyone coached them. Best for language and objections.

2. Support tickets and onboarding notes. Best for what goes wrong after the purchase, which is where renewal and expansion messaging comes from.

3. Lost-deal notes. Best for the alternatives you actually lose to, including internal builds and doing nothing.

4. Search and site data. Best for the words people use before they know a category exists. Your own site search box is the most underused source in the building.

5. Review sites and community threads. Best for how your market talks when nobody is selling to them, including about competitors.

6. Customer conversations. Best for understanding the reasoning behind a decision, which the sources above cannot fully explain. Because they take more time to arrange and conduct, use them for the questions that existing evidence cannot answer.

The first five sources are usually already available inside the business. Reviewing them first gives you a clearer hypothesis to test when you speak to customers.

Use the evidence you already have to form a hypothesis before interviewing customers

The order matters more than the effort. Spend the first few days on what you already hold: Read 20 sales calls, a month of support tickets, and every lost-deal note from the last quarter. Write down the recurring phrases exactly as people said them.

That reading produces a hypothesis, and the hypothesis is what makes the interviews worth booking. You arrive with a specific thing to test rather than asking people to explain your market from scratch, which turns a vague 30 minutes into a sharp one. It also means you notice when a customer contradicts the pattern, because you know what the pattern is.

The interview guide itself is a separate craft, covered in customer discovery interviews.

Decide what counts as a finding before you review the evidence

Research in a company your size fails more often through wishful synthesis than through bad fieldwork.

First, set the bar before you look. Decide in advance how many independent mentions make something a finding. Three customers saying the same thing unprompted is a pattern. One articulate customer saying something you already believed is a quote you liked.

Second, count what does not appear. If nobody mentions the feature your team is proudest of, that silence is data and belongs in the write-up. Absences never reach a highlights reel unless you look for them.

Third, separate what people did from what people said. A buyer explaining why they chose you is reconstructing the past, and the CRM record of what they actually looked at is a useful check on that story.

Stop when 3 new sources in a row tell you nothing new

That point arrives sooner than people expect, and it beats a target number of interviews, because a well-designed spread of 12 conversations beats 30 from the same customer type.

The other stopping condition is the decision itself. If you can now answer the question you wrote at the top of the document with evidence behind it, the research is finished, whatever the calendar says.

Then write the answer down and keep the raw material. The conclusion goes into the portfolio company positioning play, and the verbatim phrases go against the customer record, which GTM modernization after an acquisition covers getting into shape.

What the method produced in a 60-person Nordic software company

Take a newly acquired Nordic software company we work with: Around 60 people, 600+ customers, two marketers, no CMO. The decision was concrete: Who should the website speak to, now that the customers being won were larger than the ones the copy described.

The team started with evidence the company already had. Recorded sales calls showed buyers using operational language the website never used, support tickets showed the same words appearing again after purchase, and search data showed people arriving on symptom terms rather than product terms.

Only then came the conversations, aimed at one question: Which person in the buying group actually decides. The answer was someone nobody in sales had met. The round took a fortnight, half of it reading material the company already owned.

FAQ

Finish the research when the evidence is strong enough to make the decision

Name the decision, pick the sources that can settle it, read what you already own before you book anything, set the evidence bar in advance, and stop when the answers repeat. Done that way a round takes about 2 weeks and produces something the company acts on. No random acts of marketing.

If the answer changes who the company should be selling to, the fix sits under the value-creation programme for portfolio companies, and smaller pieces of research sit in the add-ons that stack onto an active module.

When you are ready to run a round for a company you have just bought, book a strategy session. You will leave with a bird's-eye plan, whether or not we work together.

Book a strategy session →