Everything downstream rests on this. The framing, the prioritisation, the roadmap argument you will be having in three months — none of it can be better than the material you collect now.
Five methods cover most of what you need. Interviews and surveys tell you what people say. Contextual inquiry, heatmaps and session recordings show you what they do. The gap between the two is where most product failures live, so run at least one of each kind before you believe anything.
Pick the method from the question, not from the calendar. If you cannot state in one sentence what you are trying to find out, you are not ready to run anything — you are just booking meetings.
Interviews
One person at a time, and your job is to follow the interesting answer rather than finish the script.
That is the whole advantage. A survey freezes your assumptions into the question set; an interview lets you abandon them mid-sentence. When someone mentions they gave up on the old tool last spring, your prepared next question is worthless and "what did you do instead?" is worth the session.
Ask about the past, not the future. "Walk me through the last time you did this" produces facts — dates, tools, people they had to chase. "Would you use something that..." produces speculation dressed up as a commitment.
Run interviews early, while you are still working out what is true. Use surveys later to find out how many people it is true for.
- What it's like to interview as a product designer — article
- How To Conduct User Interviews Like A Pro (UX Design) — video
Surveys
Surveys count things. That is what they are for: you have a hypothesis from ten conversations and you want to know whether it holds across ten thousand accounts.
Which is exactly why a survey is the wrong first move. Writing good closed questions requires already knowing the answer space. Guess at the options and you get clean numbers about categories you invented.
The failure modes are predictable. Leading questions — "how useful would you find..." — harvest agreement. Free-text boxes generate data nobody ever codes. And the people who reply are the people who reply to surveys: engaged, opinionated, still here. The ones who left have to be chased by email, so treat any in-product number as a survey of your survivors.
A number you cannot explain is not a finding. It is a reason to book more interviews.
- How to use product development surveys for customer satisfaction — article
- How to Run Surveys at Every Stage of the Design Cycle — article
- Designing a Survey — article
Contextual Inquiry
Go to where the work actually happens and watch someone do it. Ask as you go, but let them lead — you are an apprentice, not an interrogator.
What you are hunting for is the workaround. The spreadsheet that shadows your product. The laminated sheet taped to the monitor. The one login three people share because the permissions model does not match the team. Nobody reports these in an interview, because they stopped noticing them years ago. It is just how the job gets done.
Context means everything around the screen too: interruptions, a supervisor leaning in, gloves that make tapping unreliable, a shift that ends at six.
Budget half a day per site. The useful hour starts once they forget you are there.
- Contextual Inquiry: Inspire Design by Observing and Interviewing Users in Their Context — article
- Contextual inquiry: A comprehensive guide — article
- Contextual Enquiry — video
Heatmaps
A heatmap aggregates clicks, cursor movement or scroll depth across thousands of sessions and paints them onto the page. Cheap, passive, always on.
Use it for questions shaped like "is anyone getting this far?" Scroll maps show where the page effectively ends, and that line sits further up than anyone on the team assumes. Click maps show people clicking things that are not links — a reliable sign that something looks interactive and is not, or that they expected an action to live there.
What it cannot tell you is why. A cold region might be ignored, or might be so clear that nobody needs to touch it. Cursor position is a weak stand-in for attention, and on a touch screen there is no cursor at all — mobile gives you taps and scroll depth, nothing more.
A heatmap tells you where to aim the next question. It is not the answer.
- Using heatmaps to improve your website's UX: 5 ways to get started — article
- How to Use Website Heatmap Data to Improve Your Site — video
Session Recordings
A replay of one real session: the scrolling, the pause, the three attempts at the same button, the tab closing.
They are the antidote to self-report. Someone will tell you the checkout was fine; the recording shows them stalled on a postcode field. Rage clicks and back-and-forth navigation are the patterns worth hunting.
The discipline is sampling. Do not watch recordings in a row hoping for enlightenment — that is television. Filter to a segment with a question attached (dropped at step three, first week on the new plan) and watch ten of those.
Two obligations. Mask personal and payment fields at capture time, not afterwards. And say so in the privacy notice: in the UK and EU, replay tools need consent before the recorder loads, because they write to the visitor's device. This is a real person being observed, and it deserves the care you would give a lab session.