The legacy doctor homepage was where members made up their minds. But it wasn't where they could act on that decision.
I saw a legacy product limitation as the opportunity to improve the doctor search experience. Kaiser Permanente's doctor homepages on My Doctor Online draw close to a million pageviews a month from members choosing a primary care doctor or specialist. But the legacy page only profiled the doctor, so members who found the right one had to switch to a separate product, Choose Your Doctor, to make it official.
Choosing a doctor is one of the most personal decisions in a member's care, and it deserves one continuous experience: learn about your doctor, feel confident, and choose them, all in the same place.
From a profile to a decision product
Finding a doctor and choosing one used to be two separate journeys. The redesign made them one. Working with business and clinician leads, I rebuilt the physician homepage around the decision members came to make. After launch, clicks rose 28% year over year on 4.6% more sessions, and engagement grew from 0.83 to 1.02 clicks per session. The standalone Choose Your Doctor flow was retired, and its final steps now live on the physician homepage.
Ratings and reviews members can trust
I led design and research for a new ratings and reviews system. Business wanted to implement ratings without reviews for the MVP. After interviews, usability testing and several rounds of prototyping, I recommended delaying the feature until the backend could support reviews alongside ratings. Transparency with members mattered more than shipping features that users didn't fully trust.
A clear overview, at a glance
Name, title, specialty and empanelment status are easy to scan thanks to a deliberate type hierarchy. The doctor's photo anchors it all, so I created photography standards to make thousands of clinician portraits look like one set. Every photo now follows this style guide for image ratio, cropping, lighting, and subject placement.
The right next step, for every doctor
A clear call to action was one of the redesign's top business priorities, and the new CTA adapts to the doctor type. On primary care physician pages, including OB/GYNs and pediatricians, the empanelment message shows at a glance whether the doctor is accepting new patients.
Faster access to content on demand
Physician pages are content-heavy, so I recommended a progressive disclosure strategy using the design system's content toggles. Availability, ratings and the call to action are always visible. Supporting sections expand when members need them, and each section's default state is configurable. A jump-link navigation gets members anywhere on the page in one tap.
One visual language, top to bottom
I led visual design for the entire page, applying current design system components and style guide standards to every section. The result is one consistent visual language across thousands of clinician profiles, and a page that's easier for development teams to build and maintain.
The legacy product: deciding on one page, choosing on another
Research revealed two main friction points in the legacy journey. The first came while members assessed fit, when what mattered most was a first impression they could trust and what other patients had to say. The second came after members decided, when the page sent them to a separate product to act.
I researched, designed, prototyped, and tested every part of the new homepage to resolve both. Two design stories unfolded: rebuilding the first impression and bringing the choice home to the doctor's page, and figuring out what makes a ratings system worth trusting.
Designing the new doctor overview section - earning trust at a glance, and how to act on it
What does a member need to see in the first few seconds to trust a doctor and act?
The redesign launched in the Mid-Atlantic States region first, then came to Northern California a year later. The overview section had three jobs: introduce the doctor, make panel status unmistakable, and offer the right next step from the same page.
How we tested
Erica Raleigh, Senior UX Researcher, led two rounds of unmoderated mobile testing. I designed the variants and prototypes and synthesized recommendations.
Round 1 (May 2024): 13 participants, KP members and non-members.
Prototypes (A/B layout testing): specialist and primary careRound 2 (August 2024): 11 participants, KP members and non-members.
Prototype (End to end specialist flow): specialist referral flow
Initial explorations for testing
I designed three directions. One tested a large, warm portrait. One moved the name and specialty to the top. One kept the entire overview, call to action included, above the fold.
The photo wasn't about size, It was about visual tone and consistency
Photo size and name placement didn't change how participants chose a doctor, and every participant found the compact layout easy to scan. The photo didn't need to be the sole focus. It needed to anchor the overview and set a warm, professional tone. With thousands of portraits in the system, the real challenge was consistency, so I developed a photography style guide to make every portrait look like part of one set.
Say it once, where members are looking
The panel message tells members whether they can choose a doctor. Research showed it was getting lost before members ever reached a doctor's page.
On the search results page, every card opened its panel line with "Accepting patients," and the detail that changed, like "by referral only," came last. We noticed research participants stopped reading after the first few words whenever text repeated, so the detail that mattered most was the one they skipped.
On the doctor's homepage, I treated the panel message as headline content. It sits directly under the doctor's name and specialty, in bold, high-contrast type inside the overview band. The status and its condition read as two clear parts: "Accepting new patients; available by referral only." In Round 2, every participant noticed the referral message and understood that specialists required a referral.
One page, the right next step
For primary care, the call to action worked from the first round. Every Round 1 participant knew how to choose a primary care doctor from the homepage. "Select me as your doctor is a very clear statement," one said.
Specialists were a different story. Members can't choose a specialist directly. They need a referral first, and the path to one ran through a "How to make an appointment" page that participants found overwhelming. Some skimmed it and missed that a referral was required at all. Others came away with questions the page never answered:
"Do I have to pay and make an appointment with a PCP first?"
"What is a scheduling ticket?"
"If the system knew I already had a referral on file, would I still see this screen?"
Research rated it a high-priority fix. I simplified the referral message on the specialist overview so members saw "by referral only" before they clicked, and I redesigned the "How to make an appointment" page around the questions members actually asked.
"It is broken down not only how to get a referral but what to do after that referral," one said. "So, I like that a lot."
End to end specialist flow: Specialist referral UXR prototype
The product payoff
In Northern California, choosing a doctor now happens on the doctor's homepage. The separate Choose Your Doctor flow was retired. On the new homepage product, clicks rose 28% year over year from January through April, with clicks per session up from 0.83 to 1.02.
Round 1 also showed that most participants wanted ratings and reviews before choosing. That raised a different question…
Ratings and reviews: the case for waiting
The pressure
The Mid-Atlantic business team wanted ratings for Phase 1, and their reasoning made sense. Star ratings make a doctor look established, and members shopping for a doctor are used to seeing them everywhere else.
There was one constraint. Written reviews depended on data and API licensing from a third-party vendor, and those couldn't be secured until well after launch. The business and design team wanted the same thing: a feature members would trust. What we needed to show was how much of that trust depends on reviews, and we did it through design iterations, user testing, and recommendations.
The question
My first exploration was the full industry standard. Stars told members how a doctor rated. Reviews told them why.
Without reviews, the version the business asked to launch kept the per-visit star ratings with only a date beside them. That raised the central question: what would members make of a star rating with nothing behind it?
In a December 2024 design review, I laid out the concerns. A star rating for a single visit, with no review and no patient name, doesn't explain itself, and two kinds of ratings on one page create cognitive overload.
My recommended version cut the per-visit ratings and added one experiment: a ratings distribution summary. Without reviews, would more detail build trust or raise more questions? I suspected the latter.
The evidence
In January 2025, research ran unmoderated mobile testing of the ratings prototype with 11 participants, including two KP members from Northern California. Participants liked the homepage overall and gave it 4.0 out of 5 for helpfulness. The ratings section was where it fell short.
All but one participant said written reviews were missing. Ratings on their own didn't help them decide.
Several participants mistook the rating categories for written reviews. They expected reviews and went looking for them.
The star breakdown made things worse. When participants expanded a category, they started recalculating the averages themselves and wanted to know why someone had given a low score.
"I would like to see ratings in areas that are more pertinent to your actual visit," one said. "Did she explain things well? Did she show compassion?"
Recency mattered. Participants trusted recent ratings more than older ones.
The argument
I presented the findings to stakeholders around one question. A 4.0 average seems fine, but if it were a 3.2, wouldn't you want to know why? For a choice as personal as a doctor, a chart without reasons gives members half the information they're looking for.
The recommendation followed from that. Add written reviews, because members' expectations for ratings come from Amazon, OpenTable, and Yelp, and a healthcare site should match the world members already know. Until then, simplify what we show: remove the star breakdown, relabel the categories, and make recency visible.
The decision
We recommended pausing ratings until reviews could launch with them. The Mid-Atlantic stakeholders chose a middle path: launch ratings now and add reviews later. But the ratings that shipped were shaped by the research. Per-visit ratings and the star breakdown were removed. The categories were renamed "Patient experience ratings" and spelled out in full. A short note explains that ratings appear only for clinicians with at least 20 completed surveys, and response date range shows how recent the ratings are.
It wasn't the outcome we advocated for, but members got a more transparent version than the one first requested.
The win
A year later, the research went to work again. When Northern California prepared for Phase 2, the same evidence persuaded business and clinician leads to hold ratings until reviews could ship with them. Northern California launched without ratings. Designs for a full ratings and reviews experience were completed and approved for a future release, and that approach became the standard for other regions.
Sometimes the most valuable thing a designer can ship is a well-supported "not yet."
