Turn customer interviews into a decision, not a slide deck

Customer interviews fail when nobody defines the decision first. Use research on sample size, importance and past behaviour to turn transcripts into a decision memo.

The research agency presents 48 slides from 20 customer interviews. Everyone agrees it was fascinating. Three months later, nothing in the plan has changed.

This is the most common way customer interviews fail. Not because the interviews were poor, but because nobody decided in advance what the interviews were for.

Begin with the decision, not the discussion guide

Before recruiting a single customer, write down the decision the interviews will inform. Make it specific enough to have a deadline and an owner.

“Understand our customers better” is not a decision. “Decide whether to lead next year’s campaign with speed or with reliability” is. So is “decide which of two segments gets the sales team’s attention in the first half.”

Then write down what you would need to hear to choose each option. If no plausible interview evidence would change your mind, you do not need interviews. You need a decision.

How many interviews are enough?

The classic evidence comes from Abbie Griffin and John Hauser’s 1993 study, “The Voice of the Customer”, published in Marketing Science.

In their data, interviewing 20 customers identified more than 90% of the needs that emerged from interviewing 30. Based on their results, they hypothesised that 20 to 30 interviews would capture 90% to 95% of customer needs, and that one-on-one interviews might be more cost-effective than focus groups.

Treat this as a planning range, not a law. Their study covered a single product category, and the right number depends on how varied your customers are. If you sell to three distinct segments, you may need that range for each one.

Do not count mentions as importance

The same study contains a finding many teams still ignore. Griffin and Hauser tested whether the needs customers mentioned most often were the most important ones. They concluded that frequency of mention “does not appear to be a good surrogate for importance.”

This matters because the usual output of an interview programme is a frequency table: “14 of 20 customers mentioned price.” That table feels like evidence. It may simply reflect what is easy to talk about.

Importance needs a separate test. That could be a short follow-up survey, a forced trade-off exercise, or behavioural data such as what customers actually bought when both options were available.

Ask about events, not opinions

Interview quality depends on what you ask. Opinions about hypothetical futures are cheap. Accounts of real past decisions are much more useful.

Clayton Christensen and colleagues made this case in their September 2016 Harvard Business Review article, “Know Your Customers’ ‘Jobs to Be Done'”. Their argument is that customers “hire” products to make progress in a specific circumstance, and that correlation-based customer data often misses why a purchase happened.

Their example is a condominium developer that struggled to convert visits into sales. An adviser asked recent buyers to draw a timeline of how they came to buy. A surprising theme emerged: what to do with the family dining table. The obstacle was not a missing feature but the anxiety of leaving a home full of meaning.

The lesson for interview design: ask customers to reconstruct a recent decision step by step.

What triggered the search? What else did they consider? What nearly stopped them? Who else was involved?

Turn transcripts into a decision memo

The output of an interview programme should be a two-page memo, not a deck. A useful structure has five parts.

1. The decision and the options

Restate the decision and the options that were on the table before the research.

2. What we heard, with the strength of evidence

For each relevant finding, say how many interviews it came from, whether it came from past behaviour or stated opinion, and whether it differed by segment.

3. What surprised us

Surprises are often the most valuable output, and the easiest to lose in a summary. Name them explicitly, even if they do not fit the brief.

4. The recommendation

Choose an option. Explain which evidence drove the choice and what would have changed it.

5. What we still do not know

List the open questions and the cheapest way to answer each one, such as a survey, a pricing test or a sales experiment.

Share the raw material

Griffin and Hauser also recommended that multiple analysts or team members read the transcripts, because different readers notice different needs.

Practically, that means giving product, sales and service leaders access to a handful of complete transcripts or recordings, with customer permission. A short quote heard in the customer’s own words often moves a sceptical colleague more than a summary chart.

It also guards against a quieter risk: the interviewer hearing what they hoped to hear.

Common traps

A few patterns turn good interviews into wasted effort.

Recruiting only happy customers produces flattering findings. Include recent losses, churned customers and people who chose a competitor.

Letting the salesperson who owns the account run the interview introduces bias. Customers are polite to people they will see again.

Asking leading questions, such as “Would faster delivery help?”, invites agreement. Ask what happened instead.

Waiting until all interviews are complete before discussing them delays learning. Hold a short debrief after every few interviews and adjust the guide.

The test of a good programme

The test is simple. Within a month of the final interview, has a named decision been made differently, or made with more confidence, because of what customers said?

If the answer is yes, the interviews were research. If the answer is no, they were an expensive way to feel close to customers.