The Complete Guide to UX Research for Ecommerce

Most ecommerce stores make design decisions based on gut feel, industry trends, or what the competition is doing. Then they wonder why conversion doesn’t improve.

UX research methodology replaces guessing with evidence. It tells you what your actual customers are struggling with, what’s confusing them, and what’s stopping them from buying. The stores that invest in research before changing their design consistently outperform the ones that change things and hope.

The average ecommerce conversion rate is 1.4 to 2.0 percent. Stores that run systematic UX research and implement the findings typically reach 3 to 5 percent. That’s not from running more ads or changing button colors. It’s from understanding where real friction exists and eliminating it.

This guide covers every UX research method that matters for ecommerce: user interviews, usability testing, session recordings, heatmaps, surveys, card sorting, and tree testing. For each method, I explain what it reveals, how to run it for ecommerce, what questions to ask, how to recruit participants, and how to connect the findings to conversion decisions.

Why UX Research Is a Revenue Decision

Before covering methods, the business case needs to be clear. UX research is not an optional nice-to-have for ecommerce stores above a certain scale. It’s the most cost-effective way to identify revenue problems.

Consider the math: if your store does €100,000 monthly revenue at 1.5 percent conversion, moving to 2.5 percent without increasing traffic adds €67,000 per month. A round of user interviews (10 sessions at €50 per participant incentive = €500) that identifies and leads to fixing the three biggest conversion barriers is one of the highest-ROI investments you’ll make.

The alternative is testing randomly. A/B testing without a research foundation produces 5 to 8 percent win rates on most ecommerce stores. Research-informed A/B testing produces 30 to 40 percent win rates because you’re testing real hypotheses grounded in real user behavior.

Research also prevents expensive mistakes. Building a new feature that users don’t want, redesigning navigation in a way that loses existing users, or investing in mobile app development when 60 percent of your best customers shop on desktop. All of these are avoidable with basic research.

The UX research process for ecommerce draws on two types of data. Quantitative UX research methods (analytics, heatmaps, surveys with scale responses) tell you what is happening and where. Qualitative UX research methods (user interviews, usability testing, session recordings) tell you why. Both are necessary. Quantitative data without qualitative context produces changes that look good in metrics but don’t reflect real user needs. Qualitative data without quantitative validation produces anecdotes that may not represent your broader customer base. The methods in this guide cover both ends.

Method 1: User Interviews

User interviews are the foundation of qualitative UX research. You talk to actual customers or potential customers for 30 to 60 minutes about their shopping behavior, decision-making process, and experience with your store.

What user interviews reveal: The “why” behind behavior. Why did they choose your product over a competitor? What almost stopped them from buying? What questions did they have that your product page didn’t answer? What do they do when they’re unsure about a purchase? These insights cannot come from analytics data.

How to Run User Interviews for Ecommerce

Who to interview: Recent buyers are the most valuable interviewees for ecommerce. They went through your complete purchase journey and can tell you what was easy, what was confusing, and what almost stopped them. Also interview recent abandoners if you can reach them (via abandoned cart email follow-up). Their perspective on why they didn’t buy is often more instructive than why buyers did.

For 10 interviews, aim for 6 to 7 recent buyers and 3 to 4 people who match your customer profile but haven’t bought from you. The non-buyers give you perspective on consideration and hesitation.

Recruiting participants: Recent buyers are the easiest to recruit. Send an email to customers who purchased in the last 30 days. Subject: “15 minutes to help us improve? €20 gift card for you.” Expect 5 to 10 percent response rate. For 10 participants, email 150 recent buyers.

For non-buyers, use participant recruitment platforms like Dscout or Prolific. Define your screening criteria: age range, shopping frequency, product category interest, and any product-specific criteria. Budget €40 to €60 per participant including platform fees.

Where to run them: Video calls with screen sharing. Use Zoom or Google Meet. Record with permission. The ability to watch the recording later is essential, because you’ll miss things in the moment.

The structure: Open with rapport-building and context-setting. The shopper should feel like they’re having a conversation, not being tested. Then move through these sections:

Shopping behavior and context (10 minutes)

“Walk me through the last time you bought [product category] online.”
“How do you usually find the products you’re looking for?”
“What would make you choose one store over another for this type of product?”

Your store experience (20 minutes)

For buyers: “Tell me about when you bought from us. What made you decide to buy?”
“Was there anything confusing or uncertain during that process?”
“What almost stopped you from completing the purchase?”
For non-buyers: share your store screen and ask them to walk you through how they’d shop for [product category]. Observe without guiding. Then ask about hesitations.

Specific friction points (15 minutes) Based on your analytics data or hypotheses, probe specific areas:

“When you were looking at the product page, what questions did you have that the page didn’t answer?”
“What did you think about the shipping costs when you saw them?”
“What did the return policy tell you about buying from this store?”
“On your phone, how did the checkout feel?”

Wrap-up (5 minutes)

“What’s the one thing we could change that would make this store better for you?”
“Is there anything you wish you’d known before buying from us?”

How to synthesize findings: After 10 interviews, you’ll hear patterns. The same hesitation about returns mentioned by 6 out of 10 people is not a coincidence. It’s a conversion problem with a clear solution. Create an affinity map: write each insight on a card (physical or digital in FigJam or Miro), group similar insights, and identify themes.

Themes from user interviews become your hypotheses for design changes and usability tests. “6 out of 10 interviewees couldn’t find the return policy without looking in the footer” is a testable hypothesis: “Adding the return policy near the ATC button will increase conversion on the product page.”

Conversion impact of user interviews: Teams that run quarterly user interview programs consistently identify 3 to 5 high-impact improvements per round. At 10 percent conversion improvement per fix, 3 fixes over a year compound to a 33 percent overall improvement. For a €100,000 monthly store, that’s €33,000 in additional monthly revenue from systematic research.

Method 2: Usability Testing

Usability testing is structured observation. You give participants specific tasks and watch them try to complete those tasks on your store. You don’t help. You observe.

What usability testing reveals: Where shoppers get stuck, confused, or frustrated. The difference between where you think the friction is and where it actually is. Checkout flows that seem logical to you but break for users. Navigation structures that make sense internally but fail externally.

How to Run Usability Tests for Ecommerce

Moderated vs. unmoderated:

Moderated: you’re present during the test, observing and following up with questions. Better for complex research questions. Requires scheduling and facilitator time.
Unmoderated: participants complete tasks independently, recorded automatically. Faster and cheaper. Better for high-volume testing on specific flows.

Start with moderated tests for foundational research. Use unmoderated for validation and for testing specific changes.

Task design: Tasks must be realistic and specific without leading the participant. Bad task: “Add a product to your cart using the add-to-cart button.” Good task: “You’re looking for a birthday gift for a friend who enjoys cooking. Find something under €50 that you think they’d like, and get to the point where you’d pay for it.”

The good task mirrors real behavior. It has a goal (birthday gift, cooking, under €50) without directing how to accomplish it. The bad task describes the UI element by name, which tells the user exactly what to do and reveals nothing about how they’d naturally navigate.

Standard usability test tasks for ecommerce:

Find a specific product by browsing (not searching)
Find a product using search
Evaluate a product page and tell you whether they’d buy
Add a product to cart and complete the checkout
Find the return policy and tell you what it means for their purchase decision
Find a customer review that would help them decide

5 to 6 tasks per session. Each session runs 45 to 60 minutes.

The rule of five: 5 participants in a usability test reveal 85 percent of usability problems in an interface. You don’t need 30 participants to find the major issues. Run 5 sessions, synthesize, fix the major problems, run 5 more sessions to validate. This loop is faster and cheaper than a single large test.

Running the session: Start with this briefing: “We’re testing the website today, not you. There are no wrong answers. Please think aloud as you go. Tell me what you’re looking at, what you’re thinking, what you’re trying to do. If something is confusing, say so. If you’d give up in real life, tell me.”

Then be quiet. Don’t answer questions (“Where would you look for that?”), don’t guide (“Try scrolling down”), don’t react to what you see on screen. Your job is to observe and occasionally probe with “What are you thinking?” or “What did you expect to happen there?”

Take notes on: where participants hesitate, where they click incorrectly, what they read vs. skip, where they express confusion or frustration, and where they give up.

Unmoderated testing tools: Maze integrates directly with Figma prototypes and live URLs. Participants follow tasks and are recorded. Maze auto-generates a report showing task completion rates, click paths, and time on task. Useful for testing specific flows (checkout, product page) with 20 to 50 participants quickly.

Connecting usability test findings to conversion: If 3 out of 5 participants can’t find the size guide, that’s a conversion problem visible in your analytics as add-to-cart abandonment on size-variable products. If every participant asks “what’s the delivery time?” while on the product page, that’s a conversion problem visible as high exit rates at the delivery question in checkout.

Method 3: Session Recordings

Session recordings are video recordings of real users on your actual store. Every mouse movement, scroll, click, and keystroke is captured. You watch these recordings to see real behavior at scale.

What session recordings reveal: How real shoppers navigate your store, where they rage-click (clicking the same element repeatedly in frustration), where they drop off, and what they look at before leaving. Unlike usability tests, session recordings show you natural behavior without a facilitator or task structure.

How to Use Session Recordings for Ecommerce

Tools: Microsoft Clarity (free), Hotjar, Fullstory. Clarity is the starting point. It’s free, has native Shopify integration, captures checkout sessions with proper GDPR handling, and includes rage-click and dead-click detection automatically.

What to watch: Don’t watch randomly. Filter sessions by segment and behavior:

Filter 1: Mobile sessions on your product page for your top-selling product. This is where you have the most traffic and the highest potential impact. Watch 20 sessions. Look for: Where does the scroll depth stop? Do they reach the ATC button? Do they tap the variant selector? Do they look at reviews?

Filter 2: Sessions with rage clicks. Clarity and Hotjar both identify sessions with rage-click events. Sort by highest rage-click count and watch the top 10. Each rage-click session shows you a broken interaction in real use.

Filter 3: Sessions that include cart adds but don’t complete checkout. These shoppers had purchase intent and something stopped them. Watch their behavior in checkout. Where specifically did they stop?

Filter 4: Sessions longer than 5 minutes that didn’t convert. Long sessions without conversion often mean the shopper wanted to buy but couldn’t find key information. Watch what they were searching for.

What to document: After each session watch, note:

The exact point where behavior changed (started scrolling back up, stopped interacting)
Any visible confusion signals (circular clicking, repeated scrolling of the same section)
What they clicked before leaving
What they didn’t click or didn’t scroll to

Volume: Watch 30 to 50 sessions per segment when investigating a specific issue. 10 sessions reveal obvious problems. 30 to 50 sessions confirm whether they’re widespread or isolated.

Converting findings to action: If 40 percent of mobile sessions stop scrolling at the same point on your product page, something there is either satisfying their need or blocking them. Identify what’s at that scroll depth. If it’s where your ATC button appears, great. If it’s where a confusing UI element appears, that’s your friction point.

Method 4: Heatmaps

Heatmaps aggregate behavior across thousands of sessions into a visual overlay showing where users click, scroll, and focus attention. They answer questions about aggregate behavior that session recordings (which show individual behavior) cannot.

What heatmaps reveal: Which page elements get attention and which are ignored. Whether users scroll to your key content. Whether they click elements that aren’t links. How scroll depth varies across device types.

Types of Heatmaps for Ecommerce

Click maps: Show every click on a page as a dot or heat area. On product pages, check: Are users clicking the product images (do they want to zoom/expand)? Are they clicking elements that look clickable but aren’t? Are they clicking the variant selector before the ATC button?

Scroll maps: Show how far down the page users scroll. If 70 percent of users scroll past your product description but only 30 percent reach your reviews section, most of your social proof is invisible. Move it up.

Move maps (desktop only): Show cursor movement patterns. Users tend to read text by moving their cursor along it. Move maps reveal which copy blocks users engage with and which they skip.

Attention time maps (Fullstory, Contentsquare): More sophisticated than cursor movement. Show where users actually spend visual attention based on cursor dwell time. Useful for understanding which product image or description element drives engagement.

How to use heatmaps for specific ecommerce decisions:

Product page optimization: Run a scroll map. Find the scroll depth where 50 percent of users drop off. Everything above that line is seen by most visitors. Everything below is seen by half or fewer. Move your highest-converting elements above the 50 percent scroll line. On mobile, this often means your reviews and trust signals need to be much higher on the page than you currently have them.

Category page optimization: Run a click map. Which products get clicked? The pattern should align with your merchandising logic (bestsellers first, promotional products highlighted). If it doesn’t, your visual hierarchy or sort order isn’t working.

Homepage optimization: Click maps on your homepage reveal which navigation items and hero CTAs actually drive traffic. You may discover that 80 percent of homepage clicks go to 20 percent of the available links. That tells you what your actual homepage needs to do versus what you’ve designed it to do.

Method 5: Surveys

Surveys collect structured feedback from a large number of users simultaneously. Unlike interviews (which are deep but slow) and heatmaps (which show behavior but not reasons), surveys can capture large-scale quantitative data combined with qualitative open-ended responses.

What surveys reveal: Reasons behind behavior that analytics can’t show. Purchase motivations, barriers to buying, satisfaction with specific features, comparison with competitors.

Three Surveys Every Ecommerce Store Should Run

Survey 1: Post-purchase survey (the “why you bought” survey)

Send via Klaviyo 24 to 48 hours after purchase. Keep it to 3 to 4 questions maximum.

Questions:

“What almost stopped you from buying?” (open-ended)
“How did you find us?” (dropdown with options including “other”)
“How confident were you about your purchase when you clicked Buy?” (1-5 scale)

The most valuable question is the first one. The answers reveal your highest-friction points from the perspective of people who made it through. Common responses: “Wasn’t sure about sizing,” “Wasn’t sure about shipping time,” “Wasn’t sure about returns,” “Price seemed high but I bought anyway.” Each of these is an actionable signal.

Survey 2: Exit-intent survey (the “why you’re leaving” survey)

Show this survey when a user shows exit intent on a product page or in checkout. Keep to 1 question.

On product page: “You’re about to leave. What would have made you more likely to buy today?” (multiple choice with “other” + text field). Options: “Not sure it fits / suits me,” “Price too high,” “Wanted to compare elsewhere,” “Needed more information,” “Just browsing,” “Technical problem.”

In checkout: “You’re about to leave your cart. What stopped you?” (multiple choice). Options: “Shipping cost,” “Didn’t want to create an account,” “Wanted to pay a different way,” “Not ready to buy yet,” “Security concern,” “Other.”

Exit surveys on checkout pages have 5 to 10 percent response rates. At 1,000 checkout exits per month, that’s 50 to 100 data points per month with no additional traffic spend.

Survey 3: NPS survey (the “would you recommend us” survey)

Send 14 days after delivery to give customers time to experience the product. The standard NPS question: “How likely are you to recommend [store] to a friend or colleague?” (0-10 scale).

Follow up promoters (9-10) with: “What do you love most about shopping with us?” Follow up detractors (0-6) with: “What would we need to do to earn a higher score?”

The detractor follow-up responses are your most valuable research data. These are customers with a grievance who are willing to articulate it. Every pattern in their responses is a conversion and retention problem.

Survey Tools for Ecommerce

Typeform: Best UX of any survey tool. Completion rates are significantly higher than Google Forms or SurveyMonkey because each question appears one at a time. Critical for exit-intent surveys where you have limited attention. The Shopify integration handles post-purchase trigger automatically.

Klaviyo forms: Native to your email platform. Best for post-purchase surveys delivered via email. Responses feed directly into customer profiles for segmentation.

Hotjar Feedback: The tab that appears on the side of pages where users can leave feedback at any time. Not a structured survey. More of an open channel. Useful for catching issues you haven’t thought to ask about.

Method 6: Card Sorting

Card sorting is a research method where participants organize a list of items (topics, products, categories) into groups that make sense to them. It reveals how your users’ mental model of your product catalog compares to how you’ve actually structured it.

What card sorting reveals: Whether your category structure makes sense to shoppers. What they call things versus what you call them. Where your navigation decisions diverge from user expectations.

When to run card sorting for ecommerce: Before redesigning navigation, restructuring your product catalog, or adding new category pages. Also valuable when analytics show high drop-off in navigation or when session recordings show users searching for things they should be able to browse to.

Running Card Sorting for an Ecommerce Store

Open card sort (for discovery): Give participants 20 to 40 cards, each with a product name or product attribute. Ask them to sort the cards into groups that make sense to them, then name each group. This shows you how they naturally categorize your products.

Example: if you sell outdoor equipment and your categories are “Camping,” “Hiking,” “Climbing,” and “Water Sports,” but 80 percent of participants group camping and hiking items together and call it “Outdoor Adventures,” your two separate categories may be causing unnecessary navigation friction.

Closed card sort (for validation): Give participants the same product cards plus your existing category names. Ask them to sort the products into the categories. This shows you how well your current structure matches their mental model. High placement accuracy (70 percent or higher) means your categories are clear. Low accuracy means they’re confusing.

Tools: Maze has a card sort feature. Optimal Workshop is the dedicated tool for card and tree testing. Both generate visual dendrograms showing which items were most commonly grouped together.

Participant count: 15 to 20 participants for card sorting. Fewer produces unreliable patterns. More is rarely necessary for typical ecommerce category research.

Conversion impact: Navigation that matches user mental models reduces the search effort required to find products. The direct conversion impact is measurable in reduced bounce rates on category pages and increased product page views per session. Stores that fix category structure based on card sort research typically see 10 to 20 percent improvements in navigation engagement metrics.

Method 7: Tree Testing

Tree testing is the complement to card sorting. Where card sorting tells you how users want to organize things, tree testing tells you whether they can find things in your current structure.

What tree testing reveals: Navigation success rates for specific finding tasks. Where users get lost in your information architecture. Which navigation labels are ambiguous or misleading.

How tree testing works: Participants see only your navigation structure (text only, no visual design). You give them a finding task: “Where would you go to find X?” They navigate through your tree until they select a location.

The output is: task success rate, direct success rate (found it first try), time to task completion, and a tree map showing where participants went when they got lost.

When to use tree testing for ecommerce:

Validating navigation structure before a redesign
After card sorting identifies a potential restructure
When analytics show high “back” button usage in navigation (sign of wrong-path navigation)
When users in session recordings visibly struggle to find specific types of products

Sample tasks for an ecommerce tree test:

“You need a gift for someone who enjoys cooking. Where would you look?”
“You bought something last week and want to check on your order. Where would you go?”
“You want to know if this store accepts returns. Where would you find that policy?”
“You’re looking for running shoes in size 43. Where would you start browsing?”

Run these tasks with 20 to 30 participants. A task success rate below 70 percent means your navigation structure fails for that task. Tasks with success rates below 50 percent are critical failures with direct revenue impact.

Method 8: Customer Journey Mapping

Customer journey mapping is the process of documenting every touchpoint a customer has with your store, from initial awareness through post-purchase. As a UX research method, it synthesizes data from multiple sources (interviews, analytics, session recordings) into a single visual artifact that shows the complete shopping experience from the customer’s perspective.

What customer journey mapping reveals: Where customer expectations and actual experience diverge. The emotional high and low points across the shopping journey. Handoff gaps between stages, such as where a shopper transitions from product browsing to checkout and something breaks. Which touchpoints drive confidence and which create friction. And critically: problems that individual methods won’t surface because they only capture one stage in isolation.

How to Build an Ecommerce Customer Journey Map

Define the scope. A customer journey map covers one persona completing one task. For ecommerce, the most valuable map is your core buyer persona completing a first purchase. Don’t try to map every possible path. Map the one that matters most.

Identify the stages. Standard ecommerce journey stages: Awareness (how they discovered you), Consideration (browsing products, reading reviews), Decision (product page, variant selection, add to cart), Purchase (checkout, payment), Post-purchase (delivery, product experience, returns). Each stage is a column in your map.

Fill each stage with research data. For each stage, document:

What the customer is doing (the actions)
What they’re thinking (their questions and goals)
What they’re feeling (their emotional state: confident, uncertain, frustrated)
The touchpoints involved (your website, email, social media, packaging)
Identified pain points (from interviews, recordings, surveys)

Source data from methods you’ve already run. Customer journey mapping is a synthesis tool. The data comes from your user interviews (“What almost stopped you from buying?” maps to the Decision stage), session recordings (where they hesitate on the product page), exit surveys (why they left checkout), and post-purchase surveys (what confused them after ordering). The map is the frame that organizes findings from multiple methods into a coherent view.

Tools: Miro and FigJam are the most-used tools for collaborative journey mapping. Both have pre-built templates. A spreadsheet works equally well for the first version. Don’t let tooling slow you down.

Conversion impact: The value of a journey map is in identifying the stage with the most friction. If your map shows that 6 out of 8 interviewees expressed uncertainty at the Decision stage but felt confident by checkout, your product pages are the problem. If the post-purchase stage shows high confusion and support volume, your order confirmation and delivery communication need work. The map shows you where to look. The other methods tell you exactly what to fix.

Run a customer journey map exercise once per year as part of your research calendar, and update it after any major store change. It keeps your team aligned on where customers actually struggle versus where you assume they struggle.

How to Recruit Research Participants for Ecommerce Studies

Recruitment is the step most teams skip or do badly. Who you recruit determines the validity of your research. The wrong participants give you irrelevant findings.

Recruiting from your customer base:

Your email list is your best participant source. Send a recruitment email to recent buyers (within 60 days for recency, within 12 months for broader experience). Offer a meaningful incentive: €20 to €30 gift card for 30 minutes, €50 for 60 minutes.

Be specific in your screening. For product research on kitchen equipment: “Do you cook at home at least 3 times per week?” and “Have you bought a kitchen product online in the last 6 months?” These two questions screen for the right profile quickly.

Expected response rates: 5 to 10 percent from a warm customer list. For 10 participants, email 150 to 200 customers.

Recruiting from external panels:

When you need users who haven’t bought from you (for unbiased product or category research), use:

Prolific (prolific.com): Academic-grade research participants. UK and EU focused. Well-screened. €8 to €15 per participant per 30 minutes.
Dscout: Specialized in UX research recruitment. Better for complex screener criteria. €40 to €60 per participant.
Respondent.io: Focuses on professional and consumer B2C research. €50 to €100 per participant but very high quality.

For unmoderated studies (Maze, usability testing), use Maze’s participant panel. It’s integrated, fast, and covers EU markets.

Screener questions for ecommerce research:

Standard screening criteria:

Do you shop online at least once per month? (Must answer yes)
Have you bought [your product category] online in the last 6 months? (Must answer yes)
Which device do you primarily use for online shopping? (Collect data, don’t filter unless you need device-specific research)
How old are you? (Collect range, filter to your core demographic if relevant)

For mobile research, add: “Do you shop on your smartphone at least sometimes?” and test with actual mobile users, not desktop users who “also use mobile.”

Synthesizing Research Into Actionable Recommendations

Raw research findings are not useful. Synthesized findings with prioritized recommendations are. The synthesis step is where most research programs fail: teams collect data, produce a report, and file it without action.

The synthesis process that drives decisions:

Step 1: Collect all observations. After each research session, write down every observation immediately while memory is fresh. One observation per note (physical sticky note or digital equivalent in FigJam). Include: what you saw or heard, where it happened, and why it matters.

Step 2: Group by theme. Spread all observations across a surface. Group related observations together. “Couldn’t find return policy,” “Concerned about what happens if it doesn’t fit,” and “Needed more confidence before buying” cluster into a theme: purchase confidence and risk perception.

Step 3: Prioritize themes by frequency and impact. Frequency: how many participants showed this pattern? Impact: what’s the likely conversion effect? A friction point experienced by 8 out of 10 participants in checkout is higher priority than a confusion point seen in 2 out of 10 product page sessions.

Step 4: Generate specific, testable recommendations. Each theme becomes a recommendation:

Bad recommendation: “Improve product page clarity.” Good recommendation: “Add the return policy (or a 1-sentence summary) within 2 rows of the Add to Cart button on all product pages. Current: policy is in the footer. Problem: 7 out of 10 interview participants either couldn’t find it or didn’t know it existed before buying.”

The good recommendation has: the specific change, where to make it, and the evidence supporting it. Anyone reading it knows what to do.

Step 5: Estimate impact and prioritize.

For each recommendation, estimate:

Conversion impact (high/medium/low)
Implementation effort (high/medium/low)
Confidence level (high if 8/10 participants showed the pattern, low if 2/10)

Prioritize high-impact, low-effort, high-confidence changes first. This is your implementation roadmap.

Building a Research Program: How Often and What to Run

Research is not a one-time project. The stores that maintain research programs consistently outperform those that run research once and consider the problem solved.

A practical ecommerce research calendar:

Quarterly: User interviews (10 sessions) Focus on a specific question each quarter: What stops people from buying? How do customers find us? What do they compare us against? Each round costs €500 to €1,000 in incentives and produces 3 to 5 high-impact findings.

Monthly: Session recording review (2 to 3 hours) 30 to 50 sessions reviewed per month, focused on your highest-traffic pages. Document patterns. Add to your known-issues list.

After every major change: Usability testing (5 participants) New checkout flow, new product page layout, new navigation structure. Test before you launch. 5 sessions take 2 to 3 days to schedule and run. They save you from shipping something that creates worse problems than it solves.

Ongoing: Post-purchase survey Set it up once. Let it run permanently. Review results monthly. The data compounds over time and shows you when a new problem emerges (sudden increase in “wasn’t sure about returns” after you changed your policy language, for example).

Annually: Tree testing and card sorting When you’re planning navigation changes or catalog restructuring. Run before the redesign, not after.

The Research Mistake That Kills Conversion Gains

The most common research failure is running research without a clear question. Teams collect hours of session recordings without knowing what they’re looking for. They run user interviews without a specific hypothesis. They survey without knowing what decisions the data will inform.

Every research activity should start with: “What decision will this research help us make?”

“We want to understand why our mobile conversion is 0.8 percent while desktop is 2.4 percent” is a clear research question. It leads to: session recordings filtered by mobile, usability tests run on mobile devices, interviews with mobile-first shoppers.

“We want to improve UX” is not a research question. It leads to broad research that produces broad findings that nobody acts on.

Start with the conversion metric you want to move. Identify the stage in the funnel where you’re losing people (analytics). Form a hypothesis about why. Design research to test that hypothesis. Synthesize findings. Implement the change. Measure the result.

This loop, repeated consistently, is what moves ecommerce stores from 1.5 percent to 3 percent to 5 percent conversion over time.

The tools are accessible. The methods are learnable. The incentive costs are modest. What separates stores that do this consistently from those that don’t is treating research as infrastructure rather than as a one-off project.

Start with 10 user interviews and 30 session recording reviews this month. That’s enough to identify your 3 biggest conversion problems. Fix those 3 problems. Then do it again.