Specific screens and internal design details are protected. This piece focuses on process and methodology. Full work available on request in a conversation.
Udemy's login and signup flows supported passwordless authentication but the UX had not been redesigned to reflect this shift. The flows had usability issues creating friction at one of the most critical points in the conversion funnel, the moment a user decides to sign up or sign back in.
I led this audit end to end over 7 weeks, defining the methodology, running the analysis, producing redesign recommendations, and presenting findings to the PM, PM Director, Growth Design team, and Growth Engineering team.
Login entry, signup entry, email verification flow, and associated error and edge case states.
I used Nielsen's 10 usability heuristics and Laws of UX as the foundation to build a structured set of evaluation questions for each screen. Every finding was mapped to a principle so recommendations were grounded in established UX thinking rather than subjective opinion.
Used Nielsen's 10 heuristics and Laws of UX to create a structured set of questions for each screen. Every issue found had to map to a principle. No findings without a reason.
Systematically evaluated login, signup, and email verification across all screen states and edge cases. Documented every usability issue with severity and heuristic category.
Analyzed login and signup patterns across 20 products in 4 industries to identify what users already expect from authentication flows and where Udemy diverged from established patterns.
Produced redesigns for the two highest-traffic screens with specific changes tied to each finding. Presented to PM, PM Director, Growth Design team, and Growth Engineering team with prioritized recommendations.
Every usability issue in the flow traced back to a single problem: the experience had been updated for passwordless authentication without being redesigned for how users actually experience and trust a passwordless system.
Third-party login options lacked clear labeling. The distinction between social login and organization SSO was not immediately clear. Users had to read carefully to understand their options rather than recognizing them at a glance.
The passwordless system was new and there was no signal telling users it was intentional. No notification, no explanation. For a security-sensitive action like login, that gap in trust communication is a real barrier to conversion.
Sharp rectangular buttons inconsistent with the direction the product's visual language was heading. Inconsistent use of background colors on secondary actions. Small decisions that compound into a flow that feels unpolished at a critical moment.
"The passwordless system was technically sound. The design had not been updated to help users trust it. Every friction point in the flow came back to that gap between what the system was doing and what users could understand about it."
I analyzed login and signup flows across 20 products in 4 industries. The patterns that emerged across every category informed the prioritization of every recommendation.
Focused on simplicity and reduced friction. Clear third-party options, minimal steps to get into the product.
Set the benchmark users compare everything to. High clarity, strong trust signals, consistent labeling across every auth option.
Multiple verification paths for flexibility. Login and signup optimized for speed. Every extra step is a potential drop-off.
SSO is prominent and expected. Users signing into work tools have higher security awareness and respond well to explicit trust signals.
I produced redesigns for the login and signup entry screens with specific changes mapped to each finding. The goal was to give the engineering and PM teams something actionable and prioritized, not a list of observations.
15 out of 16 products in the analysis used text plus icon for every third-party option. "Other log in options" creates an unnecessary cognitive step. Clear labels and icons let users recognize their options immediately.
Users encountering a passwordless flow for the first time need a signal that the absence of a password field is intentional. Explicitly communicating "no password required" addresses the trust gap that the system change had created.
The email verification step only offered a code. Adding a magic link gives users a faster path back into their account, one click instead of typing a six-digit code. More options for the same action reduces drop-off.
Rounded corners reduce cognitive load and feel more approachable. The signup screen was also missing value propositions. A reason to commit to creating an account at the exact moment users are deciding whether to.
The projected $1.2M GMV increase and $3M annualized figure were estimates based on conversion assumptions provided by the PM. The findings and redesigns were handed off as a deliverable for the team to prioritize and implement.
As an intern this was my first experience connecting UX findings directly to revenue projections. Understanding that design decisions have a business number attached to them changed how I approach every project since.
A heuristic evaluation is only as good as the argument you make for it. The research was the easy part. The harder part was walking into a room with a PM Director and making a clear, confident case for which changes mattered most and why they should be prioritized.
The audit was a deliverable handed off at the end of the internship. Being in the room as recommendations get built and prioritized would have added a layer of learning that no research phase alone provides.
Expert evaluation surfaces issues that principles predict. User testing surfaces issues that real behavior reveals. Pairing the two would have made the recommendations harder to argue with and easier to prioritize.
There was no hard deadline but I spent more time in the evaluation phase than I needed to. I have learned since that moving faster from findings to recommendations, even without external pressure, gives the work more time to actually land and be acted on.
Categorizing findings by type is a start. Assigning severity scores would have given engineering and PM a clearer triage framework after handoff rather than leaving that judgment to them.