MOBILE EXPERIENCE · UX RESEARCH
Tested understanding of personalization and onboarding experience of an upcoming health & wellness app for a large media company.
ROLE
UX Research Intern · Summer 2026
TIMELINE
2 weeks
METHODS
Structured Interviews · Affinity Mapping · Synthesis
01 / PROBLEM
The onboarding experience plays a huge role whether a user will continue using an app. For a pre-launch wellness app, that promise had to do a lot of work: explain how to use the app's unfamiliar concept, justify why personalization would be useful, and give people a reason to come back tomorrow.
With such a unique app format, I had a very important question to answer: Does the new onboarding experience help users understand the app's value, believe the experience is tailored to them, and feel ready to begin using it?
The onboarding feature will be integral in communicating the app's value to new users. Thus, it was my job to understand how the user actually feels and interprets the experience when moving through it.
02 / APPROACH
As my internship began to wrap up, I had two weeks to complete this task. Understanding the time limitations, I had to adapt as a full report would not be done in time.
I chose unmoderated testing to field ten sessions using a think-aloud, task-based protocol. Since I was testing the feel of the onboarding experience as a whole, no extra probing from a moderator was needed. Unmoderated testing also allowed me to jump straight into the video analysis and synthesis process.
I scoped the deliverable to the timeline to ensure quality results. Rather than being rushed to build a full report I couldn't finish well, I proposed to Product and Design to deliver a prioritized set of findings, opportunities, and recommendations. This allowed me to spend time digging into the findings and delivering thoughtful recommendations.
03 / PROCESS
01 - Align
Met with Product and Design to clarify goals, intended outcomes, and what decisions the findings needed to support.
02 - Plan
Drafted a research plan covering background, objectives, methodology, and session structure; documented it in Confluence so the team could follow the reasoning.
03 - Build
Wrote structured interview questions and built the study in UserTesting.
04 - Fail, diagnose, rebuild
The first round came back unusable because the participants had misread my instructions. I diagnosed it as a prompt-framing problem rather than a participant problem, rewrote the prompts with far more explicit setup, and re-fielded.
05 - Field
Ran ten sessions successfully on the revised protocol.
06 - Synthesize
Reviewed every recording against high-level categories, isolated the recurring friction points, and built themes through affinity mapping.
07 - Pressure-Test
Used our team’s UXR AI playbook to run a second pass over the transcripts. AI was able to help surface supporting quotes, generate an independent theme set, and check my own conclusions against it. Where the two passes disagreed, I went back to the recordings.
04 / ARTIFACTS
See my finished report below!
05 / IMPACT
People were enthusiastic. They just couldn't tell you what they'd be doing.
Four findings drove that:
Enthusiasm was high, comprehension was partial. Participants liked what they saw and wanted to get their hands on it through self exploration. They largely wanted to find out what the app actually does.
Onboarding asked too little to promise so much. The personalization questions were simple enough that people couldn't work out how the app was generating recommendations from them. This led to a bit of skepticism in the accuracy in personalization.
Specific features remained unclear even after onboarding, like the logging feature and the app's visual status indicator. Once onboarding ended, users still experienced confusion over what those features were useful for.
Named talent and the learn-and-practice format read as genuine differentiators. This was the part of the promise people understood immediately and valued against competing products.
Onboarding was successfully selling the product, but it still needed work in explaining it.
The findings gave Design and Product a concrete basis for revisiting onboarding structure with retention and expectation-setting as the framing rather than aesthetics or length.
The convergence mattered too. This study and a separate usability study on the same product arrived at the same conclusion from opposite directions: the risk wasn't that people couldn't use the product, it was that they couldn't articulate why it was worth using. Two independent methods pointing at one problem made the case considerably harder to set aside.
06 / LEARNINGS
1) In unmoderated research, the prompt is the moderator. My first round failed because I wrote instructions in a way that left room for ambiguity and interpretation. By focusing on concrete visual indictors, users will have a less confusing end-of-task signal to look for.
2) Communicating conflict is not a bad thing. Since I only had two weeks to finish a full study, I worried over not getting everything done in time. Communicating this issue to my colleagues and coming to a compromise helped build trust among the team and ensure the deliverable's quality.
3) I am the researcher, and AI is my tool. Using AI to run through the raw data and extract themes, quotes, and patterns helped supplement my own conclusions. However, my judgement and understanding of the product was something AI could not replace.
What I'd do next: test revised onboarding flows against comprehension measures rather than sentiment, and explore whether more personalization questions improve trust in the recommendations that follow.