User Research: How to Know What Your Customer Actually Wants Before You Pay for Any Design

User Research: How to Know What Your Customer Actually Wants Before You Pay for Any Design

User research is structured listening to user behaviour before you spend a design budget. You do not always need it: an expert audit can diagnose problems visible in analytics. It is essential when the questions are open: who the user is, why they churn, what they do outside your product.

The phrase "user research" shows up in proposals as a vague line item. It adds weeks to the schedule and a number to the invoice, and nobody explains what happens in those weeks. One agency treats it as mandatory for any serious work. Another jumps straight to screens. You sit between the two with no way to judge.

This guide is written from the side of the table that sells the service. That position lets us say what proposals will not: the research is not always needed. And when it is needed, it is the cheapest decision in the whole project. The goal is that you know which case is yours before you sign anything.

What user research actually is

The definition we work from: user research is structured listening to user behaviour before you spend a design budget. Every word earns its place. "Structured" means a repeatable method, not a friendly chat with two clients you already know. "Behaviour" means watching what people do, not what they say in meetings. And the timing carries all the value: a finding that arrives before design costs one session. The same finding after launch costs a rebuild.

The difference that justifies the cost is evidence versus opinion. "Customers want a simpler app" is an opinion from a meeting. "Seven of ten participants stopped at the address step and never finished checkout" is evidence from observation. The first opens an endless debate. The second closes it.

The international standard points the same way. ISO 9241-11, in its 2018 revision, defines usability as effectiveness, efficiency and satisfaction in a specified context of use. That context, which device, which place, what pressure, never appears in internal meetings however long they run. The research is how you learn the actual context of use rather than the imagined one.

When you genuinely need user research

The rule we apply to our own proposals: the research becomes essential when the questions are open. Open means nobody in the room can answer them with evidence, while everyone answers with confident, contradicting opinions.

Who is the user in the first place? If your team's answers range from "young people" to "small businesses" to "everyone," you are designing for someone nobody has met.

Why do customers churn? Analytics shows where they leave. It says nothing about why. The reason lives outside your product: in the alternative they opened after you, and in the job they could not get done.

What job does the user hire your product for? Products get hired to accomplish something. Often it is not the thing you built them for.

Are you entering a new market? A product that succeeds in one Gulf market carries assumptions that may not survive the next one. The reverse holds too.

The cost of skipping these questions has numbers attached. In its 2026 update, CB Insights reviewed 431 venture-backed companies that shut down since 2023. Of those, 43% cited poor product-market fit among the reasons. Building the wrong thing well is the most expensive research you never ran.

When the research can wait: the part proposals leave out

Speaking as the side that sells this: full research is not required on every project. A proposal that makes it mandatory for every client is selling time, not answers.

If the problem is visible in analytics, it needs a diagnosis, not a study. A cart abandoned at one specific step, a signup form everyone quits at the fourth field: a UX audit diagnoses these in a few weeks at a fraction of the cost, and it has its own guide.

If the friction is the kind an expert spots in the first review, weak contrast or a seven-step checkout, then paying for a full round to confirm the obvious is waste wrapped in methodology.

If you are pre-launch with no users, a small launch usually beats any study. An early version in front of a hundred real users answers faster and cheaper than months of hypothetical sessions about a product that does not exist yet.

The dividing line is simple. Closed questions that your data and an expert eye can answer: start with the cheaper route. Open questions about people you have never met: the research is the only road.

The methods that earn their cost

Textbooks list dozens of methods. Three of them do most of the work in commercial products, and each answers a different kind of question.

User interviews reveal what people say. A one-to-one, semi-structured conversation. Its output is motive and context: why the customer chose their current solution, and what they are really trying to get done. You also learn which alternatives they tried before you. Interviews open questions well. They predict behaviour poorly, because what people say about themselves is not what they do.

Usability testing reveals what people do. You give a user a real task and watch them attempt it without help. Nielsen's classic finding applies here. Jakob Nielsen published it in 2000: testing with about five users typically uncovers around 85% of usability problems. He attached the conditions himself: comparable user groups, and iterative rounds with fixes between them. The finding covers usability-testing rounds. It does not extend to surveys.

Analytics reveals what happens at scale. Where everyone drops off, and which paths get used at all. It answers "where" and never gets near "why."

MethodThe question it answersTypical size
InterviewsWhy does the user behave this way?5 to 8 per segment
Usability testingWhere do they fail during the task?About 5 per round
AnalyticsWhere does everyone drop off, at what scale?All current users

The say/do gap is why you combine them. A user tells the interviewer your pricing is clear. Then in the test you watch the same person open three pages hunting for shipping fees. One method alone gives you half the picture, confidently presented as the whole.

How a session actually runs: no theatre

A good session is boring to watch. A user holds their own phone. The task is real: "order the product you usually buy." The researcher stays quiet and takes notes. No slideshow, no tour of the interface, no rescue at the first stumble. The stumble is the data.

On a phone, specifically. The CST's Saudi Internet Report 2024 settles why: internet penetration at 99%, mobile as the browsing device in 99.4% of usage, and 48 GB of data per person per month, three times the global average. A session that observes behaviour on a desktop monitor studies behaviour that barely exists in this market.

We ask the participant to think aloud while working. Then the researcher shuts up. Silence is the hardest skill in this craft, because every intervention contaminates the sample. And a leading question produces a comfortable lie. Ask "wasn't that step easy?" and most people politely say yes. You pay for that politeness later, in conversion rates. The right question suggests no answer: "what did you expect to happen here?"

Research in Saudi Arabia and the Gulf

Recruiting is the hardest step here, and the one that decides whether the results are true. A participant who does not resemble your customers gives you clean data about the wrong audience. Sessions run in the participant's own dialect, not stiff formal Arabic, and on their own device with their own connection. Change either one and you change the behaviour itself.

A bilingual product needs testing in both languages, never in English with the findings "generalised" across. The user reading your interface in Arabic meets different text lengths, a different direction, and different numeric fields than the English reader. Problems do not transfer between the two versions.

And no global report substitutes for one local observation. Watch a user in Riyadh or Jeddah attempt your checkout. She stalls at the address field because her new district is missing from the list. That teaches what no dashboard shows. Running ux research in this region is not a translation of foreign playbooks. It is the same evidence discipline applied to a genuinely different context of use.

What you receive, and what happens next

The core deliverable is a findings report ranked by evidence. Every finding carries how many participants hit it, its severity, and the design decision it touches. A report ranked by the team's enthusiasm gets argued down in the first meeting. A report ranked by evidence ends the argument before it starts.

Personas, we deliver only when the data actually produces them. Manufacturing four characters with names and stock photos because "that is what UX projects look like" is theatre. It decorates the report and changes no decision. We have seen entire reports of that kind buried in company archives.

After the report comes the translation into decisions. Findings redraw the user journeys, then flow item by item into UI design, which has its own guide. When you compare proposals from any UX design agency, ask each one to name the decisions the research will feed. "A comprehensive report" is not an answer. A serious user research project ends in a list of evidence-backed decisions, not in an elegant PDF opened once.

Related reading: design system, ui/ux design.

A classic UX reference: Nielsen Norman Group — UX Research.

Similar News

View All News
Hiring a UX Design Agency: What You Actually Get, and How to Measure the Return

Hiring a UX Design Agency: What You Actually Get, and How to Measure the Return

resources
What Is a Design System? And Why It Saves You Years, Not Months

What Is a Design System? And Why It Saves You Years, Not Months

resources
UX Design in Saudi Arabia: The Complete Guide

UX Design in Saudi Arabia: The Complete Guide

resources

Common Questions

About five per round, and the number has a source. Nielsen's classic finding, published in 2000, is that testing with around five users typically uncovers about 85% of usability problems. The conditions matter: participants from a comparable group, and iterative rounds with fixes in between. The finding applies to usability-testing rounds only. Surveys need far larger numbers before they say anything reliable.

For a mid-sized product, two to four weeks: framing the questions, recruiting participants, running the sessions, then analysis and the report. Recruiting is the step everyone underestimates, especially when you need specific segments in specific cities. Anything much faster usually means the participants were whoever was easiest to find, and the findings inherit that shortcut.

It depends on the type of question, and an honest answer refuses to pick one for all cases. "Why" questions belong to interviews: motives, context, the job behind the purchase. "How many" questions belong to surveys and analytics: proportions at scale. The common mistake is a survey that asks "why," collects ready-made rationalisations, then treats their percentages as behavioural facts.

Yes, for some methods, and it beats doing nothing. Reading your analytics and running first interviews are workable in-house. The pitfalls are known: leading questions, recruiting people who resemble the team, and reading results as confirmation of existing beliefs. The founder hears whatever supports the roadmap. If you go in-house, let the least invested person moderate, and record the sessions for a colder review later.

Rarely on purpose, but the say/do gap is real. Participants want to be helpful, so they praise what they struggled with and predict behaviour they will never perform. That is why serious sessions weigh what a participant did over what they said. A recording that shows three minutes of hunting for a button outranks a closing statement that everything was clear.

When a new open question appears, not on a fixed calendar. A shifting audience, a major feature, a new market, an unexplained churn pattern: each one reopens questions that evidence should answer, not meetings. Small regular rounds beat a giant project every two years, because each round feeds decisions while they are still cheap to change. Analytics fills the space between rounds.