Most articles about user research methods for website design are written for teams with a dedicated researcher, a recruitment budget and three months to spare. That is not how a 6-person company works. If you run a small business, your research budget is usually whatever is left after hosting, ads and the designer’s invoice.
The good news: the methods that produce the biggest design decisions are also the cheapest ones. You do not need a lab, an incentive budget of thousands, or a statistically significant sample. You need five users, a support inbox, and a willingness to watch people struggle with the thing you built.
Below are seven research methods a small business can genuinely run, often inside a single working week. For each one you get: how long it takes, what you need, what it costs, and most importantly which design decision it should inform. That last part is where most research dies. Data that does not change a design is just expensive reassurance.
Quick comparison: the 7 methods at a glance
| Method | Time | Typical cost | Best for deciding |
|---|---|---|---|
| 1. Five-user usability test | 1 to 2 days | 0 to 250 EUR | Page layout, labels, form and checkout friction |
| 2. Support ticket mining | 3 to 4 hours | Free | FAQ content, product page copy, self-service flows |
| 3. Exit survey (micro-survey) | 1 hour setup, 2 weeks collecting | 0 to 30 EUR/month | Missing information, objections, pricing clarity |
| 4. Open card sorting | Half a day setup, 1 week collecting | 0 to 100 EUR | Navigation structure and menu grouping |
| 5. Tree test / first-click test | 2 to 3 hours | 0 to 100 EUR | Menu labels, validating a new IA before build |
| 6. Analytics and session replay review | 3 hours per month | Free tier usually enough | Where to focus, which page to fix first |
| 7. Customer interviews (5 to 7) | 1 week | 0 to 350 EUR | Messaging, page order, what belongs above the fold |
First, a rule that saves you money
Before you choose a method, decide what the research is allowed to change. Write one sentence:
“After this study, we will decide whether to ______.”
If you cannot complete that sentence, do not run the study. This single habit is the difference between research that produces a redesign and research that produces a 40-slide deck nobody opens.
Research methods split into two broad families, and you need both over time:
- Attitudinal (what people say): interviews, surveys, card sorting
- Behavioural (what people do): usability tests, analytics, session replays, tree tests
Small businesses tend to over-invest in the first family because it is easier to run. Behaviour is where the money is.
1. The five-user usability test
The single highest-return research method for website design. You give five people a realistic task, watch them attempt it on your site, and keep your mouth shut. The classic Nielsen Norman finding still holds: five participants surface roughly 80% of the usability problems in a given flow. Testing fifteen people mostly means watching the same three problems happen ten more times.
How long it takes
One to two days total. Thirty minutes per session, five sessions, plus two hours to write the script and two hours to synthesise.
What you need
- Five participants who resemble your actual buyers (customers, prospects, people from a local business group, or a recruitment panel)
- A video call tool with screen sharing and recording
- A script with 3 to 5 tasks, written as scenarios, not instructions
- A note-taking template or a second person taking notes
How to run it
- Pick one flow, not the whole site. Example: “Find a service that fits my problem and request a quote.”
- Write tasks as situations: “You manage a small warehouse and need to know if this product works with your existing system. Find out.” Never say “click the Compatibility tab”.
- Ask participants to think aloud. Remind them every time they go quiet.
- When they ask you a question, answer with another question: “What would you expect to happen?”
- Record everything. Note the exact moment of hesitation, not just the failure.
What design decision it informs
Concrete, page-level changes: button labels, form field order, which content block goes above the fold, whether your pricing table is readable, whether the checkout needs an account step. If three of five people hesitate at the same point, that is not an opinion, that is a defect.
Common mistake
Testing your own site with your own team. Internal testers know where everything is. Their results are worthless. Also avoid testing with friends who want to be nice. Brief people explicitly: “We are testing the site, not you. Confusion is useful data.”
2. Customer support ticket mining
You already paid for this research. It is sitting in your inbox, your live chat transcripts and your WhatsApp Business history. Support tickets are a record of every point where your website failed to answer a question on its own.
How long it takes
Three to four hours for a first pass over 90 days of tickets.
What you need
- Export of the last 3 to 6 months of support conversations
- A spreadsheet with four columns: question asked, where the answer lives today, how many times it came up, and the page that should have answered it
How to run it
- Pull every pre-sale question. Ignore post-purchase account issues for this exercise.
- Tag each one with a theme: delivery, compatibility, pricing, returns, sizing, timelines, legal.
- Count the themes. Sort descending.
- For your top five themes, open the page where that question should have been answered. Usually the answer is absent, buried in a PDF, or written in your internal vocabulary rather than the customer’s.
- Steal the customer’s exact wording. If 20 people wrote “how long until it arrives”, your heading should not say “Logistics information”.
What design decision it informs
Content hierarchy and page copy. The top three recurring questions almost always belong on the product or service page itself, above the fold or in the first content block, not in a separate FAQ page. This method also tells you which fields to add to product data and whether you need a comparison table.
3. Exit surveys and micro-surveys
An exit survey is one question that fires when someone is about to leave a key page without converting. It is the cheapest way to learn why people who were interested enough to arrive were not interested enough to act.
How long it takes
One hour to set up. Leave it running for two weeks or until you have 100 responses.
What you need
- A survey widget (most have a free or low-cost tier) or a simple custom modal
- One question, maximum two
- Exit-intent or scroll-depth trigger on pricing, product, or checkout pages
Questions that actually work
- On a pricing page: “What is stopping you from getting started today?”
- On a product page: “What information were you looking for that you could not find?”
- After purchase: “What nearly stopped you from buying?” (this one is gold)
- On a blog post: “Did this answer your question? If not, what were you looking for?”
Use open text fields. Multiple choice only confirms what you already assumed.
What design decision it informs
What is missing. Exit surveys rarely tell you to move a button, they tell you that your shipping costs are invisible until checkout, that people cannot tell which plan fits them, or that there is no proof you are a real company. Those are content and trust decisions: testimonials placement, guarantee badges, a cost calculator, a comparison module.
Warning
Do not trigger an exit survey on every page for every visitor. One page, one question, one trigger. Then turn it off. ixdf.org makes the same point with more data.
4. Open card sorting for navigation
If your menu was designed around your org chart or your internal product names, card sorting will show you exactly how wrong it is. Participants group your content the way it makes sense to them, and you discover that the category you call “Solutions” means nothing to anyone outside the building.
How long it takes
Half a day to prepare, one week to collect responses from 15 to 20 people, two hours to analyse.
What you need
- 30 to 50 cards, each one a page or content topic, written in plain language
- An online card sort tool, or physical index cards if you are doing it in person
- 15 to 20 participants (more than usability testing, because you are looking for patterns in grouping)
Open vs closed
| Type | What participants do | Use it when |
|---|---|---|
| Open | Create their own groups and name them | You are building a new navigation from scratch |
| Closed | Sort cards into categories you provide | You want to validate categories you already chose |
| Hybrid | Use your categories or create new ones | You have a draft structure but suspect gaps |
What design decision it informs
Your top-level menu, your sub-menu grouping, and your category names. Pay special attention to the labels participants invent. Those are the words your customers use, and they belong in your navigation and in your page titles, which has the convenient side effect of improving search visibility.
5. Tree testing and first-click testing
Card sorting tells you how to build the structure. Tree testing tells you whether the structure works, before you spend money building it. You show participants a text-only version of your menu and ask them to find something. No visuals, no search bar, no rescue.
How long it takes
Two to three hours including setup, collection and analysis. Participants complete it in under five minutes.
What you need
- Your proposed site structure written as a nested text outline
- 5 to 8 realistic find-it tasks
- 20 to 30 participants (it is quick and asynchronous, so volume is cheap)
- A tree testing tool, or a simple multiple-step form if you are improvising
What to measure
- Success rate: did they land on the right page? Target above 80%
- Directness: did they go straight there, or wander back and forth? Backtracking signals an ambiguous label
- First click: research consistently shows that users who get the first click right succeed far more often than those who do not
What design decision it informs
Whether to ship the new navigation or rename three categories first. It is the cheapest possible insurance against a redesign that tanks your conversions. Run it before the designer starts, not after.
6. Analytics and session replay review
Analytics will never tell you why. It tells you where to look, which is exactly what a small team needs when it can only afford to fix one thing this quarter.
How long it takes
Three hours per month, booked as a recurring calendar slot. The recurrence matters more than the depth.
What you need
- A working analytics setup with conversion events defined (not just page views)
- A session replay tool on a free tier
- A heatmap on your three most important pages
What to look at, in order
- Landing pages with high traffic and low engagement. These are your biggest quick wins.
- Drop-off between steps in any multi-step flow: form, booking, checkout.
- Internal search queries. The most under-used research source on any website. People are literally typing what they cannot find.
- Mobile vs desktop split for your top conversion pages. If mobile is 70% of traffic and 20% of conversions, you have a mobile layout problem, not a copy problem.
- Ten session replays of users who reached the key page and did not convert. Watch them at normal speed.
What design decision it informs
Prioritisation. Analytics answers “which page do we fix first”, then you run a five-user test on that page to answer “what exactly is broken”. The two methods are a pair. Using either one alone is how teams end up redesigning pages nobody visits.
7. Short customer interviews
Five to seven conversations of 20 to 30 minutes with recent customers and recent lost deals. Not a focus group. One person at a time, because groups produce consensus, and consensus is the enemy of insight.
How long it takes
One week, including scheduling. Budget 90 minutes of your own time per interview including notes.
What you need
- A list of customers who bought in the last 60 days (memory fades fast)
- A discussion guide with 6 to 8 open questions
- Recording permission and a transcription tool
- A small thank-you: a voucher, a discount, or a donation in their name
Questions worth asking
- “Walk me through what happened the day you decided to look for this.”
- “What did you try before you found us?”
- “What other options did you look at, and how did you compare them?”
- “What almost made you choose someone else?”
- “What did you need to know before you felt comfortable paying?”
Avoid “would you like a feature that…”. People are generous and inaccurate about hypothetical futures. Keep every question anchored in something that already happened. growthdrivendesign.com makes the same point with more data.
What design decision it informs
Messaging hierarchy and page order. Interviews tell you what has to be resolved first, second and third on a page, and in which words. They are also where you discover that your main competitor is not another company but a spreadsheet, a manual process, or doing nothing at all. That changes your entire homepage.
A realistic 90-day research plan
You do not run all seven at once. Here is a sequence that works for a small team with limited time, and it carries you through a full redesign cycle.
| Weeks | Methods | Output |
|---|---|---|
| 1 to 2 | Support ticket mining + analytics review | A ranked list of problem pages and unanswered questions |
| 3 to 4 | Five-user usability test on the worst flow | A defect list with severity ratings |
| 4 to 6 | Customer interviews + exit survey running in parallel | Messaging priorities and objection list |
| 7 to 8 | Card sorting | Draft navigation structure |
| 9 | Tree test the draft structure | Validated or corrected navigation |
| 10 to 12 | Design and build, then re-test with five users | Confirmation before launch |
Turning findings into actual design decisions
Research only pays off if every finding is attached to a decision and an owner. Use a table like this and keep it in whatever tool your team already opens daily.
| Finding | Source | Severity | Decision |
|---|---|---|---|
| 4 of 5 users could not find delivery times | Usability test | High | Add delivery estimate block under the price |
| 31% of exit survey replies mention unclear pricing | Exit survey | High | Publish starting prices on the service pages |
| Nobody grouped cards under “Solutions” | Card sort | Medium | Rename top-level item to “What we do” |
Severity, in plain terms
- High: blocks a purchase or a contact request. Fix this sprint.
- Medium: causes hesitation or extra clicks. Fix this quarter.
- Low: cosmetic or preference-based. Fix when you are already in that file.
Five mistakes that waste a small research budget
- Asking people what they want. Observe what they do. The gap between the two is enormous.
- Recruiting the wrong participants. Five people from your target market beat fifty random panel respondents who will never buy from you.
- Running one big study instead of six small ones. Continuous small studies compound. A single annual study goes stale in three months.
- Testing after launch. Tree tests and prototype tests cost a fraction of post-launch repairs.
- Never closing the loop. If you change something because of research, tell the team which finding caused it. That is how research becomes a habit instead of a project.
Frequently asked questions
How many users do I need for a usability test?
Five per user group for qualitative testing. If you have two genuinely distinct audiences, for example buyers and administrators, run five of each. For quantitative studies such as tree tests or card sorts, aim for 15 to 30 participants because you are looking for statistical patterns rather than individual struggles.
What are the main types of user research methods?
They are usually split along two axes. Qualitative vs quantitative (why something happens vs how often it happens) and attitudinal vs behavioural (what people say vs what people do). Interviews and surveys are attitudinal. Usability tests, analytics, tree tests and session replays are behavioural. A healthy research mix pulls from all four quadrants over the course of a year.
Can I do user research without a budget?
Yes. Support ticket mining, internal search query review, analytics review and session replays cost nothing but your time. Even usability testing can be free if you recruit from your existing customer list and offer a small discount instead of a cash incentive. The real constraint is calendar time, not money.
How often should a small business run user research?
Something small every month beats something large once a year. A sensible rhythm: a three-hour analytics review monthly, one five-user test per quarter, a support ticket audit twice a year, and an exit survey running continuously on your most important conversion page.
Should I use AI tools for user research?
Use them for the grunt work: transcription, tagging open-text survey responses by theme, clustering support tickets, summarising interview transcripts. Do not use synthetic or simulated participants as a substitute for watching a real person fail to find your contact form. The value of usability testing lives in the hesitation, the sigh and the wrong click, and none of that can be generated.
What is the difference between user research and A/B testing?
A/B testing tells you which of two options performs better. User research tells you what the options should be in the first place, and why the losing one failed. Most small businesses do not have enough traffic for statistically valid A/B tests, which makes qualitative research methods far more practical and far more informative.
Where to start this week
Pick one. Open your support inbox, export the last 90 days of pre-sale questions, and tally the themes. It takes an afternoon, costs nothing, and will almost certainly hand you three page changes worth more than your next round of ad spend.
Then book five usability sessions for next week. The combination of those two methods, run honestly, gives a small business more useful direction than most expensive discovery phases ever deliver.