How to Run a Usability Test on Your Live Website

Learn how to run a usability test on your live website to gather the evidence you need before committing to a full redesign.

How to Run a Usability Test on Your Live Website

Most redesign proposals die in committee. Not because the site is performing well, but because nobody can prove exactly where it is failing. A usability test on a live website changes that. It replaces “the site feels outdated” with timestamped recordings of real people unable to complete their most common tasks. For a product owner seeking internal buy-in, that is not research for research’s sake. It is evidence that withstands scrutiny from finance, engineering, and the C-suite.

This article walks through how to run a usability test on a live website before committing to a full redesign: participant selection, task design, remote testing tools, and the synthesis method that turns session footage into a defensible business case.

Get practitioner-led usability testing that builds your redesign business case.

Talk to a UX research lead

Why test the live site, not a prototype?

Prototypes are clean. They load instantly, contain no legacy content, and route users along carefully curated happy paths. A live website has none of those luxuries. It has slow third-party scripts, five years of accumulated navigation items, and content written by people who left the company in 2021.

Testing against the live site captures the friction users actually experience. It also anchors your findings to metrics your stakeholders already track: bounce rate on the pricing page, drop-off in a multi-step form, support tickets about a feature nobody can find. When your usability findings map directly to existing analytics pain points, the redesign argument becomes much harder to dismiss.

If you are earlier in the process and still deciding between a targeted UX audit and a broader engagement, usability testing on the live site can help you scope the right level of intervention.

Selecting participants who mirror your revenue

The single most common mistake in usability testing is recruiting the wrong people. Five articulate users who do not match your buyer profile will produce eloquent feedback that leads you in exactly the wrong direction.

Start with your actual customer segments, not demographics. A B2B platform selling to procurement teams needs procurement professionals, not “adults aged 30 to 55 who work in an office.” Pull your screener criteria from CRM data, sales call notes, or the firmographic filters your marketing team already uses.

Aim for five to eight participants per distinct segment. Nielsen Norman Group research has consistently shown that five users uncover roughly 85% of usability problems within a single user type. If you serve two clearly different audiences (say, individual contributors and their managers), run five per group. Do not blend them into a single pool and hope the patterns emerge.

Recruit from your real customer base whenever possible. Incentive rates of $75 to $150 for a 30-minute session typically yield strong participation. Panel recruits from UserTesting or Respondent work when you need speed, but verify their screener answers with a brief qualifying call.

One more thing: include at least one participant who uses assistive technology if your site serves a broad audience. Accessibility failures surface fast in usability tests, and they carry weight in executive conversations about risk. Understanding WCAG conformance requirements beforehand will help you interpret what you observe.

Designing tasks that reveal structural problems

A usability test task is not a question. It is a scenario that gives the participant a reason to act. “Find the pricing page” is a weak task. “You have been evaluating two vendors and need to compare their enterprise pricing before a meeting tomorrow morning” is a scenario that produces realistic navigation behavior.

Write four to six tasks per session. More than that and fatigue distorts the later results. Fewer and you will not have enough signal to triangulate.

Each task should target one of these layers:

Prioritize tasks that correspond to pages with known analytics problems. If your analytics show a 68% drop-off on step two of a quote request form, write a task that sends participants through that exact flow. The recording will show you why the drop-off happens, which the analytics never can.

For teams working on complex B2B sites, information architecture for buying committees is worth reviewing before you finalize task scenarios. Buying committee members navigate differently than individual users, and your tasks should reflect that.

1

Navigation and findability:

Can users locate information using the site’s primary and secondary navigation? This exposes information architecture problems that analytics alone cannot explain.

2

Comprehension:

Once users land on the right page, do they understand what they are reading? Jargon-heavy product pages often test well for findability but fail here.

3

Action completion:

Can users finish a goal-oriented flow (requesting a demo, configuring a product, submitting a form) without abandoning or backtracking?

4

Recovery:

When users make a wrong turn, can they get back on track? Error recovery reveals how forgiving (or punishing) your current site architecture really is.

Remote testing tools: what actually matters

The tooling conversation consumes too much oxygen. Here is what you actually need: screen and audio recording, a way to assign tasks in sequence, and a mechanism for the participant to think aloud without feeling surveilled.

UserTesting remains the most complete platform for unmoderated remote tests. It handles recruitment, task sequencing, recording, and basic tagging. Maze and Lookback are strong alternatives if you want tighter integration with design tools or prefer moderated sessions.

For moderated remote tests (which yield richer qualitative data), a video call on Zoom or Microsoft Teams with screen sharing works. Record the session, get consent, and use a dedicated note-taker so the facilitator can focus on follow-up questions rather than scribbling.

A few practical decisions that matter more than tool choice:

  • Test on the participant’s own device and browser. Do not send them a staging link on a clean machine. You want to see real conditions: browser extensions, screen size, network speed.
  • Disable any A/B tests or personalization that would create inconsistent experiences across participants. You need a stable baseline.
  • Record the participant’s face (with consent) alongside the screen. Facial expressions during moments of confusion are powerful evidence in stakeholder presentations.

If your organization already uses Google Analytics or a session replay tool like Hotjar, pull behavioral data for the same pages you are testing. Triangulating replay data with live test observations makes your findings far more credible.

Running the session without contaminating the data

The facilitator’s job is to stay out of the way. Do not explain the site. Do not rescue participants when they struggle. The discomfort of watching someone fail is the entire point.

Open each session with a brief warm-up: ask what the participant does at work, how they typically evaluate tools or vendors, what device they usually browse on. This establishes rapport and gives you context for interpreting their behavior later.

Then present tasks one at a time. After each task, ask three questions: “How confident are you that you found the right answer?” “Was anything confusing?” and “What would you do next if this were a real situation?” These post-task prompts surface problems participants adapted to without mentioning.

Resist the urge to add questions mid-session. Every unplanned probe introduces interviewer bias. Write your follow-up questions in advance and stick to the script.

Synthesis: turning recordings into a decision document

This is where most usability efforts fall apart. Teams record eight sessions, feel informed, and then present vague summaries like “users found navigation confusing.” That is not evidence. That is an opinion wearing a lab coat.

Use this process:

The strongest internal business cases pair usability test findings with analytics data and competitive benchmarks. Three sources of evidence are harder to argue against than one.

If your team is planning to follow this usability test with a broader research effort, structuring it as a UX research sprint can compress the timeline and keep momentum before stakeholder attention drifts.

1

Tag observations, not interpretations:

Watch each recording and log discrete observations with timestamps. "Participant 3 clicked ‘Solutions’ expecting to find pricing (2:14)" is an observation. "The navigation is broken" is an interpretation. Stay at the observation level first.

2

Cluster by page or flow, not by participant:

Group observations around the site areas they affect. You will quickly see which pages generate the most friction and which problems are isolated versus systemic.

3

Assign severity:

Use a simple three-tier scale. Critical: the user could not complete the task. Serious: the user completed it but with significant delay or confusion. Minor: the user noticed the issue but recovered quickly. Count how many participants hit each problem at each severity level.

4

Map to business metrics:

Connect each cluster to a measurable outcome. "5 of 6 participants failed to find the enterprise pricing page" maps to "the pricing page receives 40% less traffic than competitor benchmarks suggest it should." This is the translation layer that makes findings legible to finance and leadership.

5

Recommend scope, not solutions:

Your deliverable should specify which areas of the site need intervention and how severe the problems are. It should not prescribe design solutions. That comes later, during the actual redesign engagement.

Presenting findings to get the green light

Executives do not watch 45-minute session recordings. They watch 90-second highlight reels. Edit your most compelling moments into a short compilation: the VP of Marketing’s target persona unable to find the product page, the enterprise buyer abandoning the contact form after two failed attempts.

Pair each clip with the corresponding analytics data point and the estimated business impact. “This page has a 72% bounce rate. In our test, 4 of 5 participants could not identify what the product does within 10 seconds of landing.” That framing connects user pain to revenue risk, which is the language budget holders speak.

Present three options: do nothing (and accept the quantified cost), fix the critical issues only, or commit to a phased redesign. Giving stakeholders a choice, rather than a single recommendation, increases the chance they choose action.

FAQs

How many participants do I need for a usability test on a live website?

Five to eight participants per distinct user segment is the practical range. Five users typically uncover about 85% of usability problems within a single user type. If you serve multiple segments with different goals, recruit five per segment rather than blending them into one group.

Can I run usability tests internally without hiring an agency?

Yes, if you have someone trained in facilitation who can avoid leading participants or rescuing them during tasks. The risk of running tests internally is confirmation bias: teams unconsciously design tasks and interpret results to validate what they already believe. An external facilitator brings neutrality that strengthens the credibility of findings.

What is the difference between a usability test and a UX audit?

A UX audit is an expert review where a specialist evaluates the site against heuristics, accessibility standards, and best practices. A usability test puts real users in front of the live site and observes their behavior. The two are complementary: audits catch issues experts recognize from pattern knowledge, while usability tests reveal problems that only surface through actual use.

How long does a usability test take from start to finish?

A focused usability test on a live website can be completed in two to three weeks. Week one covers participant recruitment and task design. Week two is for running sessions. The first half of week three is synthesis and report creation. Timelines stretch when recruitment is difficult or when stakeholder review cycles are slow.

What remote testing tools are best for usability testing a live website?

UserTesting is the most complete platform for unmoderated remote tests, handling recruitment, task delivery, and recording. For moderated sessions, Zoom or Microsoft Teams with screen sharing works well. Maze and Lookback are strong alternatives depending on your workflow. The tool matters less than consistent task design and disciplined facilitation.

Turn usability findings into a funded redesign

You now know how to structure a usability test that produces defensible evidence. Our UX research leads run these engagements end to end, delivering the severity-scored findings and business case your stakeholders need to approve the project.

Book a usability test consultation