6+
New clinic adoptions
Healthcare AI Product Redesign
Worked on redesign that balances clinical practicality and improved reliability can lead to new adoption.
A UX/UI redesign project for the healthcare AI app LocoStep, a gait analysis application for orthopedic clinics. The iOS app allows physical therapists to record patients’ walking videos and use AI to assess gait function. Although the service offered clear business value through insurance reimbursement eligibility, sales opportunities repeatedly stalled at the proposal stage.
Role
UX/UI design
Duration
8 weeks
Team
PdM, iOS Engineer, ML Engineer, Sales, Regulatory Affairs
Contribution
UX/UI design, User research
Project Overview
Score-based to measurement-based outputs
Through research, I identified two key issues behind this stalled sales momentum: low clinical usability in the results experience, and lack of trust due to insufficient medical evidence. In response, I shifted the concept of the results screen from a score-based model to one centered on measurement-based outputs supported by medical evidence, while also expanding the range of metrics shown. This redesign transformed the experience into one that supports more comprehensive whole-body evaluation. The shift contributed not only to business results, but also to long-term strategic direction and a stronger foundation of trust.
Key Result
30%→45%
Deal conversion likelihood
4+
University research collaborations initiated
20%→57%
Progression rate to next stage
57%→5%
Stalled deal rate
6-7件
User interviews conducted
6+
Stakeholders engaged
4→10+
Metrics expanded
5-6
Touchpoints designed
- Re-proposals to clinics that had previously passed on the product were well-received after the redesign, resulting in 6+ new adoptions within two months of release.
- Post-redesign, improvements were observed across several sales-related metrics — including deal conversion likelihood, stage progression rate, and stalled deal rate — though these reflect multiple contributing factors.
- Sales team members shared that they felt the product improvements had a direct positive effect on deal outcomes.
- The redesign direction also led to collaborative research partnerships with universities, forming a medium-to-long-term strategic foundation for building evidence and academic credibility.
Business Challenge
This product was originally co-developed with a pharmaceutical company before being transferred to its current organization. With KPIs consistently unmet, it had become clear that incremental feature improvements were no longer sufficient — a fundamental rethinking of the product concept was needed.
Working with the PdM and sales team to identify the root causes of stalled deals, I framed two hypotheses for the product team to prioritize:
Insufficient clinical usability in real-world care settings
Lack of trust due to limited medical evidence
The previous results screen centered on four score-based metrics — walking speed, left-right asymmetry, body sway, and rhythm. While simple and easy to understand, it lacked the depth of expertise, explainability, and clinical credibility needed for hospital use and adoption decisions.
Physical therapists found it difficult to explain why a particular score was generated, and the level of confidence in the scores varied among practitioners. This made it hard to support the trust needed for patient communication and internal adoption discussions.
The core business challenge was not visual clarity — it was that the results experience wasn't being used, referenced, or trusted in the field.
Why This Was Hard
The primary challenge was redesigning the results experience in a way that strengthened medical credibility, while working within a strict constraint: the product could not appear to provide diagnostic outputs.
The trade-off between simplicity and clinical depth
The existing score-based display was simple and easy to communicate to patients. However, because the AI was outputting pre-interpreted scores, physical therapists and physicians found it difficult to understand how scores were calculated or what they were being compared against — making it harder to use as a basis for clinical judgment.
At the same time, shifting to measurement-based outputs introduced its own set of challenges:
Numerical values are not intuitively interpretable without context
Physical therapists would need to apply their own clinical reasoning to interpret results
Users satisfied with the existing scores might face increased burden when explaining results to patients
More information risks making the results experience harder to navigate
Regulatory considerations
The product is classified as Class I medical device, which means it cannot provide diagnostic conclusions as a service. Any design approach needed to reinforce medical evidence and trust without guiding users toward diagnostic-style interpretations.
Design requirements
This project required meeting the following criteria:
Ensure medical accuracy and credibility
Support physical therapist interpretation — not diagnosis
Create an experience that feels trustworthy and is easy to explain to patients
Remain feasible within technical constraints
Deliver value that holds up in sales and adoption discussions
Approach
I worked with the PdM to propose a redefinition of the results concept and drive hypothesis validation. Through user interviews, segmentation, and direction-setting, we revisited the core value of the product.
Concept directions to validate:
Shift to measurement-based outputs
Research into the trade-offs and product value of moving from scores to raw measurements
Metric expansion
Evaluating which additional metrics to include, with consideration for technical feasibility
Addition of medical evidence
Identifying evidence that would resonate with physical therapists and physicians, assessed against the product's overall value proposition
Validation process:
01
Hypothesis Setting
Identifying likely causes of stalled sales; redefining the concept
02
User Understanding
Interviews with physical therapists and physicians
03
Segmentation
Classifying usage contexts and information needs
04
Direction Setting
Redesigning the information architecture
Research & Key Insights
I conducted 10+ interviews with physical therapists (primary users) and physicians (decision-makers) to validate the core concept.
Preparing for faster hypothesis validation
To accelerate the validation process, I worked with the PdM to align on hypotheses and implemented two approaches:
Standardized interview guide
Designed a reusable interview template for Sales, PdM, and Design. The goal was to ensure consistent depth of insight regardless of who conducted the interview.
AI-assisted rapid prototyping
Used generative AI to quickly produce multiple prototypes, enabling short validation cycles to test concept acceptance.
Insights from interviews
Through 6–7 interviews and prototype testing with physical therapists, I identified key gaps in the existing results experience and clarified the direction for improvement.
01
The measurement-based concept itself holds value
Physical therapists responded positively to measurement-based results. Displaying raw numerical values without AI-interpreted scores — giving practitioners more room for professional judgment — was well-received in principle.
02
Raw values alone are not enough — context and reference points matter
Without clear benchmarks or an indication of what constitutes a normal or concerning value, the cognitive load of interpretation increases for both therapists and patients.
03
The required depth of information varies by context
The information needs differ between a quick patient-facing explanation during a limited rehab session and a detailed clinical evaluation by the therapist.
04
Real-world assessment is multi-dimensional
Physical therapists don't evaluate gait through a single score — they integrate observations across multiple dimensions to form a holistic picture of the patient's condition.
User segmentation
Interviews with physical therapists revealed distinct segments in how practitioners expected to use LocoStep and what value they were looking for. This informed the decision that a shift to measurement-based outputs alone would not be sufficient.
The redesign also needed to expand the range of metrics for comprehensive assessment and allow users to adjust the depth of information based on their context of use.
Design Decisions
Core design challenges
The central design challenge was balancing three competing needs:
・Supporting physical therapists' clinical reasoning through detailed numerical data
・Keeping results easy to explain to patients within limited rehabilitation time
・Maintaining a level of trust and credibility without crossing into diagnostic territory
For metric expansion, I worked with the PdM and ML engineers to review medical evidence and define the direction of the results experience.
4 score-based metrics → 10+ measurement-based metrics
Added metrics tied to more specialized clinical evaluation — including head and pelvic sway and trunk lateral tilt angle — expanding from 4 to 10+ metrics for a more comprehensive whole-body assessment.
Tab structure: Summary View ⇄ Detailed View
To support different use contexts, the results screen was split into two tabs:A Summary View focused on 3 high-evidence metrics, designed for efficient patient communication; and a Detailed View for therapists to perform in-depth, whole-body clinical evaluation.
Interpretation support labels
Added comparative labels — including average value comparisons and reference ranges — to help physical therapists draw clinical insights without being guided toward diagnostic conclusions. This supports practitioners in understanding the clinical significance of measurements and communicating them clearly to patients.
Previous interface
↓
Redesigned interface
Design Iteration
Balancing clinical depth, trust, and ease of explanation
Throughout this process, the focus was not on visual polish, but on identifying the right information, in the right order, to help users understand, explain, and make decisions. Each direction was refined through user interviews.
A案
B案
C案
View showing results alongside video playback
View showing all metrics at once
View modeled after an electronic medical record format
Feedback: Too many functions; difficult to use in clinical settings
Feedback: Poor scannability; made it harder to explain results to patients
Feedback: High cognitive load for less-experienced physical therapists when interpreting results
Expanding the number of metrics raised clinical depth, but also increased information density — which in turn raised the burden of interpretation. The solution was to add supporting context (e.g., average value comparisons) and introduce a tab structure to allow users to adjust the level of detail based on their situation.
Cross-Functional Collaboration
To align user needs, technical constraints, medical validity, and implementation feasibility, I worked closely with multiple stakeholders — facilitating perspective-sharing and helping to define and drive the product direction.
PdM
Structured the root causes of stalled deals; aligned on improvement hypotheses and priorities
ML Engineer
Discussed the validity and credibility of metrics and interpretation-support labels
Software Engineer
Coordinated on SwiftUI implementation specs ahead of design finalization
Sales
Clarified what information was most useful to communicate during adoption proposals
Regulatory Affairs
Reviewed expression risks and explainability requirements
Board members
Facilitated alignment on the improvement direction and criteria for next-phase decisions
Outcomes
The redesign contributed to improvements in both near-term adoption and sales momentum, as well as longer-term credibility-building.
Short-Term Business Impact
▸
6+ new clinic adoptions within two months of release
▸
Stage progression rate improved from 20% to 57%
▸
Stalled deal rate decreased from 57% to 5%
▸
Deal conversion likelihood improved from 30% to 45%
Long-Term & Strategic Impact
▸
Initiated collaborative research with multiple university institutions
▸
Shifted the product concept toward a comprehensive evaluation experience
▸
Established a foundation for academic publications
User Trust
▸
Enabled presentation of detailed, data-rich results in a form that practitioners found understandable and credible
▸
Received feedback through sales channels that the improvements made clinics more willing to consider adoption
▸
Addressed the gap in clinical confidence and practical usability that the previous simplified score-based design could not fill
▸
The tab structure allowed therapists to switch between patient-facing summaries and in-depth clinical views — increasing specialist depth without significantly raising the burden of explanation
Reflection
The key insight from this project was that adding more information does not solve a trust problem.Early in the process, I was thinking in terms of adding more metrics or improving accuracy.
The real breakthrough, however, came when I reframed the question.
Not "what data should we show?" but "in what order does a physician actually make decisions?"
The underlying data didn't change. What changed was the structure around it.
Take away
By distinguishing between "what data exists" and "what decisions that data needs to support" — and designing the information architecture around the latter — I learned that it's possible to meaningfully improve usability and address the problem at its root.
Next step
There is ongoing need to deepen user understanding across different patient conditions and clinical specialties, in order to refine how the product is used and what metrics matter most in each context.I also want to further improve the overall results experience by better integrating the video playback feature with the results view.
