When users churn, support tickets keep coming in, or a feature gets ignored, the cause is rarely the feature itself, it's usually the experience around it.

Structured evaluation of your product against Nielsen's 10 usability heuristics applied by a practitioner who has reviewed 50+ production interfaces, not by a junior analyst with a checklist. Each heuristic produces specific findings, not general impressions: "the password error message says 'invalid credentials' rather than identifying whether the email or password is wrong" is a documented heuristic violation with severity Major and a specific fix, not "error messages could be more helpful." Severity classifications: Cosmetic (fix if time permits, no usability impact); Minor (causes occasional confusion, workaround exists); Major (causes repeated failure or significant friction, no obvious workaround); Critical (prevents task completion, causes user exit). The evaluation covers every key flow in the product, not just the primary conversion path, but the flows that drive retention, support volume, and feature adoption. Findings de-duplicated across heuristics: a form with no inline validation fails multiple heuristics, but it appears once in the report as one specific fix.

Review of your existing analytics data to identify where user behaviour diverges from design intent, and to identify the gaps in your current tracking that leave key decisions unmeasured. Google Analytics 4, Mixpanel, Amplitude, or PostHog: funnel step completion rates, page exit rates, feature event frequency, and session duration by entry path. Funnel anomalies that trigger investigation: a step with 40% drop-off that engineering assumes is normal but has never been benchmarked against industry comparables; a feature with 20% weekly usage among users who activate it and 0% usage among users who don't, a discoverability problem or a value problem? Time-on-page vs. scroll depth correlation: a page with high time-on-page and low scroll depth suggests confusion at the top of the page, not engagement with the content. Cohort analysis where available: retention by onboarding path reveals whether users who complete a specific onboarding step retain at higher rates. Missing tracking identified and prioritised: the analytics setup almost always has gaps that, once filled, would answer the questions the product team is currently guessing at.

Systematic analysis of session recordings from Hotjar, FullStory, or LogRocket, not random browsing of sessions but targeted review focused on the flows where analytics show problems. Rage click analysis: elements that receive repeated frustrated clicks are either broken (the element doesn't respond) or misleading (users expect it to be interactive but it isn't). Hesitation patterns: a user who pauses for 8 seconds before a form field is confused about what to enter, the field label, placeholder, or help text is insufficient. Unexpected navigation: users who arrive at a feature page and navigate immediately to the documentation are encountering a learning curve that should be addressed by the feature's onboarding, not by documentation. Click map analysis for critical decision points: where do users click when presented with multiple options? The click distribution often contradicts what the team believes the primary path is. Scroll depth on landing and conversion pages: below-the-fold content that contains your primary value proposition is invisible to users who leave before scrolling. Session filters for specific segments: recordings from users who abandoned at a specific funnel step, from paid traffic that didn't convert, from mobile users specifically.

Moderated or unmoderated usability test sessions with 5-8 representative users, sufficient to identify approximately 85% of significant usability problems per Nielsen's research, without the diminishing returns of larger panels. Task scenario design: tasks framed as realistic goals ("find a plan that fits a team of 15 and start a trial") rather than instructions ("click on Pricing"), so we observe whether users can navigate to the outcome rather than whether they can follow directions. Think-aloud protocol in moderated sessions: users narrate their reasoning as they work through tasks, surfacing the mental model mismatches that explain why they try the wrong path. Task success and failure metrics: pass/fail per task, time-on-task, number of incorrect navigation actions before success. Unmoderated sessions (UserTesting, Maze) for faster turnaround when the test questions are well-defined. Remote moderated sessions for access to users in specific geographies or industries. Separate sessions for distinct user types where the workflow differs significantly, admin users and end users experiencing different problems require different test scripts. Timestamped video evidence for each finding, so the report can include a clip link rather than a text description of user confusion.

WCAG 2.1 AA compliance evaluation, the standard required for legal compliance in the UK (Equality Act), US (ADA for public-facing digital products), EU (European Accessibility Act 2025), and Australian government procurement. Automated audit using axe-core (catches approximately 30-40% of WCAG violations that are machine-detectable) plus manual review for the issues automated tools miss. Common findings in production products: colour contrast failures where marketing brand colours don't meet the 4.5:1 ratio for normal text and 3:1 for large text; missing focus indicators on interactive elements styled away by outline: none CSS; custom dropdown menus and modal dialogs missing ARIA role, state, and property attributes; images with missing or generic alt text ("image.jpg" is not a useful description); form inputs without programmatically associated labels that screen readers can announce; keyboard navigation traps in modals and flyouts that lock keyboard users inside a component. Manual screen reader testing with NVDA on Windows and VoiceOver on macOS covers the interaction patterns that automated tools cannot assess. Each finding includes the specific WCAG success criterion, the affected element with a selector, and the code-level fix.

Consolidated findings from all audit methods, de-duplicated and prioritised into a document that a product team can take directly into sprint planning rather than commissioning a new discovery project to translate the findings into stories. Report structure: executive summary (top 5 issues with expected business impact, suitable for stakeholder communication); detailed findings list sorted by priority tier; per-finding format: finding title, evidence (screenshot, session recording timestamp, or analytics data), severity rating (Critical/Major/Minor/Enhancement), WCAG criterion where applicable, root cause analysis, and specific design recommendation with sufficient direction for a designer to act on immediately. Priority matrix plotting findings on impact vs. implementation effort: high-impact, low-effort Quick Wins at the top; high-impact, high-effort Strategic items for sprint planning; low-impact items documented but explicitly deprioritised. Separate section for findings that require analytics instrumentation changes rather than design changes, these surface the measurement gaps that need to be closed first. The report is delivered as a Notion document, Confluence page, or PDF depending on your team's workflow, with all evidence assets linked and accessible.