Customer Interviews That End in a Decision

Most brands run customer interviews and come away with a slide of quotes. The quotes are real. The decision does not move. Six weeks later the same question is back on the table and somebody suggests talking to customers again.

The customers are almost never the problem. The design of the exercise is. A programme built to produce evidence looks nothing like a programme built to produce colour.

Write down the decision before you write the questions

Start with the decision on the table, not the topic. Not “understand our customer”. Something like: do we launch the 250g pack at a lower price point, or hold the 500g and fix the listing. Then write the two or three things you would need to believe to pick each option.

Those beliefs become your interview guide. Everything else is optional. If a question does not test a belief that could flip the decision, cut it. This single step removes about half of the questions teams usually plan to ask, and it is the reason the output arrives as a recommendation rather than a highlight reel.

How many conversations before the pattern repeats

There is no universal number, and anyone who quotes one is selling a template. What matters is the mechanism. Run in small waves. After each wave, log the themes you heard and mark which are new. When a full wave adds nothing you had not already recorded, the segment has saturated and more calls are just cost.

Two things move the number. Question width: a narrow question inside one homogeneous segment saturates quickly, while a broad question about category behaviour takes far longer. Segment count: saturation is per segment, not per project. If you want to compare a metro repeat buyer against a tier two first time buyer, you owe each of them their own wave. Teams routinely forget this, split fifteen calls across five segments and end up with three anecdotes per group.

Recruit people who bought, not people who agree

Recruit from your own order data. Pull real cohorts: recent buyers, one time buyers who never returned, subscribers who lapsed, people who raised a return, and people who reached checkout and abandoned. Each of those is a different conversation.

Screen on behaviour, never on attitude. “Have you bought a product in this category in the last sixty days” is a screener. “Are you interested in healthier snacking” is an invitation to say yes. Incentive driven recruitment attracts people who are good at qualifying for studies, and they will find the answer you appear to want. Pay for the time, pay everyone the same regardless of what they say, and make that explicit at the start of the call.

In India, scheduling is its own hurdle. WhatsApp confirmation with a fixed slot outperforms email invitations by a wide margin. Expect to over recruit, because no show rates on unpaid or low incentive calls are high.

Ask about the past, never about the future

“Would you buy this at 499” is not data. It is politeness. People are poor forecasters of their own behaviour and excellent reporters of what they actually did.

So anchor every question to a specific past episode. Walk me through the last time you bought this category. Where were you. What else was on the shelf or the screen. What did you nearly buy instead. What made you stop. How much did you pay, and do you remember whether there was an offer. Ask for artefacts: the order history, the pack sitting in the kitchen, the screenshot of the cart. Memory improves sharply when the customer is looking at the evidence.

The questions that reliably mislead

A short list, all of which sound reasonable in a planning meeting.

Hypothetical pricing. People say yes to prices they will not pay, especially to a polite interviewer.

Feature ranking without a tradeoff. Everything is important when nothing has to be given up. Force a choice between two real options instead.

“What would make you buy more.” This invites the customer to invent a promotion.

“Why did you choose us.” You will get a tidy story built after the fact. Ask what they did, in order, and reconstruct the why yourself.

Any question that names your brand first. It reframes the whole conversation around you, when the customer’s actual frame is the category, the occasion and the household budget.

The failure that ruins most interview programmes

The single most common mistake is asking customers to design the product. Do you want a resealable pack. Should we launch a mini size. Would a subscription help.

Customers are genuine experts in their own problem and complete amateurs at your solution. They do not know your cost structure, your fill rate, your shelf constraints or what your co-packer can run. So they answer with the nearest familiar object, which is usually a slightly better version of something that already exists.

Your job is to extract the constraint, not the feature request. When someone asks for a resealable pack, the finding is not “build resealable”. The finding is that the product is consumed across multiple sittings in a humid kitchen and the current pack fails between uses. That constraint might be solved by a different pack, a different grammage, a different closure, or a different SKU altogether. The customer named one option. You own the option set.

Turning transcripts into a decision

Do not summarise interviews one by one. That produces a document per person and no pattern.

Code across the set instead. For each transcript, tag four things: the problem in the customer’s own words, the context it occurs in, the workaround they already use, and the moment they switched brand or gave up. Workarounds are the most valuable of the four, because a workaround is a customer paying a cost to solve something you have not solved.

Then write one page per decision. State the option you are choosing, the two or three findings that support it, the strongest counter evidence you heard, and what would have to be true for you to reverse it. Add a named owner and a date. If the page cannot be written, the interviews did not test the decision and you should fix the guide rather than run more calls.

Quotes belong in a footnote. The decision is the deliverable.