Talk to the people who use what you invent

Your time is too valuable to spend it creating things people don't want. As counterintuitive as it may seem, talking to people can save you months or years of frustrating, fruitless effort. You might be afraid that talking to customers (or potential customers) will be a waste of time. That they will only tell you things you already know. Or, worse, that it will be awkward and you'll accidentally offend people.

That's okay. But you also might be looking for new opportunities. You might be wondering why people cancel. You might be wondering how to get more people to buy. You might be trying to figure out which features to add next. And your data may not be giving you clear answers.

— Deploy Empathy, Michele Hansen

The best thing about this passage is that it does not argue from virtue. It does not say you should talk to people because it is the right thing, or because the customer is at the centre, or through any of the formulas that have been repeated so long they no longer mean much. It argues from cost: talking is cheap and building is expensive, and therefore not talking is the worst investment decision you can make in digital product.

And yet almost nobody does it regularly. It is worth understanding why, because the reasons are more interesting than the recommendation.

The three objections and what they hide

Hansen lists the fears with uncommon honesty: that it will waste time, that they will only tell you what you already know, that it will be awkward and you will offend someone. All three are real and all three point to a different problem than the one they name.

That they will only tell you what you already know does indeed happen very often. But it is almost always a symptom of the question rather than of the customer. If you ask about solutions — what would you like it to have, what is missing, what would you add — you will get a slightly reordered version of your own roadmap, because the person you are talking to is doing your job with less information than you have. If you ask about the situation — what happened last time, what you were trying to do, how you ended up solving it — material appears that sits in no dashboard.

That it will be awkward is a correct intuition that has been misread. The conversation is uncomfortable, yes, but not for the customer: for you. Asking in depth means exposing yourself to an answer that dismantles a decision you already made, often in front of someone on your team. That is the real cost, and better scripts do not remove it.

That it will waste time is, of the three, the one most often supported by false arithmetic. The visible cost of the conversation — scheduling, travel, transcription — gets compared against an invisible one: the months spent building on an untested conjecture. Since the second does not appear on any spreadsheet until the project is over, the comparison always favours not talking.

What conversation does that data cannot

There is a line in Hansen worth underlining: and your data may not be giving you clear answers. She does not say the data is wrong. She says it does not answer.

Behavioural data tells you what happened and how many times. It does not tell you what someone was trying to achieve, which alternative they considered, what made them give up, or the criterion by which they judged the result. All product analytics is an exercise in inferring intentions from traces, and traces admit several compatible stories. Conversation is not a rival source to that data: it is what lets you choose which of those stories is worth looking for in it.

Hence the best tradition in this craft insists on context rather than opinion. Holtzblatt and Beyer's contextual design was built on the idea of going to see the work where the work happens, because nobody can accurately describe an activity they have mastered: expert practice becomes invisible to the person performing it. Christensen pushed in the same direction from economics: we do not buy products, we hire progress, and progress is only intelligible by looking at the whole circumstance. Von Hippel took the argument to its conclusion by showing that much of the relevant innovation had already been done by the advanced user, alone and with worse tools, and that the company only needed to go and look.

All those traditions say the same thing in different vocabularies: the information you are missing is not an opinion, it is a situation. And situations are not collected with surveys.

The sustainable version

The most common mistake, once the argument is accepted, is to turn this into a project. Research gets commissioned, twenty interviews get run, a fifty-page report gets produced, it is presented to a committee, and then fourteen months pass without anyone talking to a customer again. The report ages faster than it can be read.

What works is the opposite: small quantity, high frequency. One or two conversations a week, run by the same people who will decide, with no intermediaries aggregating and cleaning the information before it arrives. Teresa Torres calls this a habit, and the word is well chosen: what makes it work is not the rigour of any single interview but the continuity. An isolated conversation produces an anecdote; thirty conversations spread over six months produce the ability to recognise a pattern when it shows up.

And the temptation to delegate it is worth resisting. When the person talking to the customer is not the person deciding, the loop still exists but gets longer: someone listens, someone summarises, someone prioritises the summary, and at each step the anomalous disappears — which is exactly where the new information was.

Asking well is a skill, not an attitude. It can be learned, it costs little to practise, and it remains the cheapest instrument you have for not spending months building what nobody asked you for.

When did you last talk to someone who uses what you build, without a report in between?