How to Run a Customer Interview
Berlin, 1904. Hans, the horse belonging to the maths teacher Wilhelm von Osten, answered sums by tapping his hoof on the ground. Addition, subtraction, fractions, calendar questions. The shows were free, the crowds were large, and the horse's fame had reached as far as The New York Times.
In September 1904 a thirteen-member panel chaired by the psychologist Carl Stumpf examined the horse. It included a vet, a circus manager, a cavalry officer, teachers and the director of the Berlin Zoo. The verdict: there was no trickery involved.
Stumpf's student Oskar Pfungst would not let it go. He changed the questioner, put blinkers on the horse, and sometimes prepared questions to which the questioner did not know the answer. The picture became clear: when Hans could see the questioner, he got 50 out of 56 questions right; when he could not, only 2 out of 35. If the questioner did not know the answer, neither did the horse.
According to the book Pfungst published in 1907, the secret was simple. As the hoof approached the right number, the questioner's face and posture tensed without their realising it, and relaxed at the correct tap. Hans was not doing sums; he was reading the expectations of the person in front of him. Nobody had tried to fool him. The questioners themselves had put the answers they wanted to hear into the horse's mouth.
The Hans in your interview
The same trap is set when you talk to customers to test a startup idea. You carry your excitement, your hope and your love for your idea into your voice and your questions. The person in front of you agrees with you because they are polite. And you write down "customers want it".
Jeanette Mellinger, head of user research at BetterUp, who previously led the same work at Uber Eats, suggests a three-step way out of this trap for founders in an interview with First Round Review: plan, listen without bias, find the pattern.
How to run a customer interview
Start by writing down the decision you need to make
According to Mellinger, the first question is not "Who should I talk to?" but "What is my next decision?" Imagine yourself in the review meeting a few weeks from now. What do you need to know at that table? Pick a single assumption and test it with a small round of research of five to eight interviews.
Narrow down who you talk to
"Everyone" is not a target audience. As well as characteristics like age, city and experience, look at how people think and behave. A small group that feels the problem most acutely will teach you more than a large but indifferent crowd.
Ask open, short, unvarnished questions
Questions that end in yes or no, like "Would you use this feature?", are like Hans's hoof; you decide the answer. Instead, say "Tell me about the last time you ...". Do not put adjectives like "great" or "tedious" in the question; that adjective is your opinion. Ask one short question at a time. If it is not understood, rephrase it in two or three different ways.
Save your product for the end
Mellinger highlights three biases. The first is confirmation bias: listening for what you want to hear rather than the uncomfortable truth. The second is acquiescence bias: the other person trying to please you. The remedy is to mention your product as late as possible. The third is retrospective distortion: people describe both what they have experienced and what they will do differently from how it really is. Instead of hypotheticals like "Would you use it?", ask about a concrete memory: what happened last time, what did you do? According to Mellinger, the strongest indicator of future behaviour is past behaviour.
Start wide, zoom in, then zoom out again
Open the interview with the person's world: their work, their day, their priorities. Then move into the area where the problem happens and ask for detail. Finally, open up again. One of the questions Mellinger describes goes like this: if you had a magic wand, how would you solve this problem?
Take your notes while they are fresh
After each interview, set aside fifteen minutes and write down three things separately: the words you heard and what you observed, the interpretation you draw from them, and the actions to take. In a weekly one-hour meeting, put that week's interviews side by side and ask "why". When the round is over, set aside half a day; write each insight on a piece of paper, stick it on the wall and group similar ones together to see the pattern. Holding on to the data also stops you jumping to a solution too early.
Interview: seeing each other
In the 1510s the English word "interview" described a face-to-face meeting. Its root is the French "entrevue", from "s'entrevoir": to see each other, to visit briefly. Inside it are "entre", between, and the Latin "videre", to see. Its journalistic meaning today only appeared in American English in 1869.
The Turkish word for interview, "mülakat", comes by a similar route. The Arabic "mulāqāt" means an encounter or meeting, and the letters l-q-y at its root describe coming across one another. The first definition given by the Turkish Language Association (TDK) still preserves this: a mutual conversation held by meeting somewhere.
In both languages the essence of the word is not interrogation but mutuality. That was the lesson of Pfungst's experiment too: as much as you can see Hans, he can see you. A good customer interview is one in which you show the other person your curiosity, not your face.
A final word
According to Diogenes Laertius in his "Lives of the Eminent Philosophers", Zeno of Citium said this to a young man who was talking idly:
"The reason we have two ears and only one mouth is that we may listen more and talk less."
In your last customer interview, who talked more, you or them? And was the "yes" you heard really theirs, or was it read off your face?
This article is based on Jeanette Mellinger's guide to customer discovery in First Round Review. If you want to know how to build your first version after the interviews, have a look at What Is an MVP and How Do You Plan One?
