How to ask customers what they actually want

To find out what customers actually want, stop asking what they want. Ask what they did the last time the problem showed up, and what that cost them. People guess badly about their future selves and describe their past selves accurately.

"What features would you like to see?" produces a wish list. The wish list is real in the sense that somebody typed it, and useless in the sense that nobody would change anything to get it.

Why "what do you want" gets a polite answer

Asking someone what they want hands them your job and expects a useful answer back.

You are asking a person to imagine something that does not exist, predict how they would behave inside it, and report the prediction as a fact. Nobody can do that. They try anyway, because you asked politely and they are a decent person.

Two forces bend the answer. People want to be helpful, so they manufacture an opinion instead of admitting they have none. And people describe the version of themselves with free time and good habits, who is not the version sitting at the desk on a bad afternoon.

So you get "more integrations would be nice". Nice is not a reason. Nobody ever rearranged their week for nice.

Ask about the last time, not the next time

The most useful question you can send is about the last time the problem actually happened.

Past behavior is a fact the person can check by opening their own inbox. Future intent is a mood, and moods do not survive contact with a calendar.

Put the two side by side. "Would you use a shared template library?" gets a shrug shaped like a yes. "When did you last need to write nearly the same message twice, and what did you do?" gets a story: they searched sent items, found something close, pasted it in, rewrote half of it anyway, and were late to the thing they were supposed to be doing.

That story contains what no survey gives you. The workaround.

Everybody with a real problem already has a bad solution for it. They built it themselves, they are slightly embarrassed by it, and they will describe it in detail if you ask about the mechanics rather than the feelings. Find the workaround and you have found the problem with a price already attached.

The questions that get a story instead of an opinion

Good questions point at one specific past event and ask what it cost.

• When did this last come up, and what were you doing right before it?
• What did you do instead, and roughly how long did that take?
• Who else got dragged in, and what did you tell them?
• What have you already tried and given up on?
• If nothing about this ever changed, what breaks first?

None of them mention your product. That is deliberate. The moment your product walks into the question, the answer turns into a review of you, and people grade their friends generously.

Treat a feature request as evidence of a problem

A feature request is evidence that a problem exists, and it is not a specification for solving it.

Someone asking for a button has already done the design work in their head, using the tools they happen to know. Their design deserves to be read. It is not the brief.

The move is one question backwards: what were you trying to finish when you wanted that? Sometimes the answer is that the thing already exists and they could not find it, which is a completely different bug. Sometimes it uncovers a problem far bigger than the button, and building the button would have hidden that from you for good.

I think public voting boards are close to useless. They rank the wishes of people who enjoy filling in voting boards, and that group is never a fair sample of the people using the thing quietly.

How to write the ask so people reply

The message that gets answered is short, asks about one thing that already happened, and gives the reader permission to ignore it.

• Say in one line who you are and why you picked them.
• Ask one question. A second one doubles the effort and halves the replies.
• Point the question at something they did, never at something they would do.
• Offer an easy exit, like "no need to reply if this never comes up for you".
• Send it while the memory is warm, close to the moment they hit the thing.

Everything that makes a cold message work applies here, because to the reader that is exactly what this is. Make the reason you picked this person visible in the first line. If nothing comes back, one follow-up is the limit, and it should be shorter than the first message.

The bit I keep relearning on Unvent: the messages that get answered are the ones where I name the exact moment I am asking about. A broad question about the product gets a polite line back. A question about the last time someone rewrote the same reply four times gets three paragraphs.

Reading the answers without flattering yourself

Read for what people did, and discount nearly everything they said about the future.

Cut the praise out of the note before you file it. Not because it is insincere. Because it takes up the space where a fact should be.

Look for repeats instead. A single customer's workaround is an anecdote. Three of them inventing the same workaround separately is a specification, written for you by people who did not know they were writing it.

Weigh what you find by cost: time lost, work redone, the apology afterwards. Problems with a real cost get fixed by whoever solves them first. A mild annoyance gets tolerated forever, no matter how many people agree it is annoying.

Then test the answer with an offer. "Yes please" means almost nothing. "Can I have it this week" means the pain is real, and somebody asking when they can start is the closest thing to proof you will get before you build.

Ask the people who left

The people who stopped using your product are the cheapest research available, and the ones nobody wants to email.

They owe you nothing. That is exactly why their answers are clean. A happy customer softens the truth to protect your feelings, while someone who has already gone has no reason to bother.

Write to them with no recovery pitch anywhere in the message. One question: what were you doing on the day you stopped? The answers tend to be smaller and more boring than the story you told yourself about why they went, which is the useful part, because boring problems are the ones you can actually fix.

Keep the people who left on a list, and read that list before you decide anything large. Not to win them back. To remember that the thing had a shape they could not use, and that you picked that shape on purpose, on some Tuesday, for reasons you have almost certainly forgotten.