The Expanding Universal Drug Table
CONTEXTWhen numbers don't tell the whole story
SmithRx MSS agents were handling 16,000 cases monthly with claim rejections comprising ~80% of the workload. The concerning metric: 30-minute average resolution time per case, with pharmacies waiting on hold and frustrated members calling back. The organizational consensus was clear—agents were inefficient because they had to juggle too many tools: Laker, Salesforce, and the incomplete Voldemort system. ZEBRA was initiated to solve this "fragmentation problem" and consolidate workflows. But something about that 30-minute average didn't quite add up.
CHALLENGEDiagnostic complexity masquerading as tool complexity
Everyone agreed the problem was "too many tools"—millions in licensing fees for Laker, Salesforce, and the incomplete Voldemort system that weren't purpose-built for claims. The solution seemed obvious: build one unified interface to eliminate the chaos. But what if the tools weren't the real problem? What if agents were spending most of their time not switching between systems, but trying to understand what each cryptic rejection code meant before they could even begin to act?
DISCOVERYThe breakthrough: diagnosis was the disease
Through agent interviews, behavioral analysis, and Looker deep-dives, I uncovered the real bottleneck: agents spent 20-30 minutes per case just figuring out what type of rejection they were dealing with. Simple, automatable cases took as long as complex prior authorizations—not because of inherent complexity, but because of diagnostic overhead. Just 20 rejection codes accounted for 70% of claim failures, yet agents had to manually "root cause" each one through three systems before they could even begin problem-solving. The tools weren't the problem—the diagnostic tax was.
Reframing the problem changed everything
My discovery completely shifted the approach. Leading the entire strategic framework after an unclear PRD, I designed Vapor around the real problem: semantic intelligence, not tool consolidation. The breakthrough was recognizing that we didn't need one tool—we needed one tool that could think. Four intelligence principles emerged to transform database outputs into agent expertise.
-
If someone can search Google, they should be able to search for their meds.
This principle replaced technical form logic with real-world language. We anchored search behavior around what members actually know — brand names, partial recall, common typos — and built the system to meet them there. This informed the creation of Find My Meds, a smart, single-field drug search powered by autocomplete and fuzzy matching.
The approach leaned on Jakob’s Law (design for familiarity), Hick’s Law (simplify decision points), and Fitt’s Law (make the action obvious)
-
Don’t just tell me what my plan is — show me where I stand.
Most benefit portals regurgitate plan PDFs. But members want to know where they are in their coverage journey — especially with deductibles as the key gating factor. Drawing on learnings from Anagram and Rally Health, I introduced a live “Plan” view that tracked member spend against deductible thresholds on a 24-hour data delay.
This wasn’t just transparency — it was relevance. Track My Spend became a clear example of designing for meaningful insight over static information.
These principles — Familiar Intuition, Recognition First, and Context Over Static — weren’t features. They were foundations. And they shaped everything that followed.
-
If it’s important, make it obvious — not memorable.
Members frequently struggled to locate their Member ID, even though it was crucial during pharmacy visits. We moved away from burying key information and instead surfaced it contextually — clearly, accessibly, and consistently. This principle drove the redesign of Show Me My ID, and aligned with the core usability heuristic: recognition over recall.
STRATEGY
PRODUCT DESIGNThe app that does only what you need it to
We rebuilt the new Member App as a responsive, mobile-optimized web app with our members in mind. Every interaction, every layout, every moment was designed to do exactly what a member needed, precisely when they needed it.
An intelligent drug search replaced the brittle form. Autocomplete handled brand and generic confusion, as well as typos, returning relevant results quickly. Local pricing appeared alongside directions to nearby pharmacies. Members could now check prices, find their ID (and save a virtual card to their device), track their plan spending, and access benefit documents — all within a unified, mobile-first experience.
Batch operations that adapt to how you think
I designed a modular system that could work forwards or backwards—matching how different ops specialists think. One user might start with 'assign Zoloft to these programs' while another thinks 'update all programs by adding Zoloft.' Through extensive workflow mapping and prototype testing, I created interchangeable 'operators' and 'affectors' that made both approaches equally valid. What took 1-2 hours in spreadsheets now completed in minutes—with zero learning curve because the system matched each user's mental model.
It all begins with an idea. Maybe you want to launch a business.
It all begins with an idea. Maybe you want to launch a business.
Speaking human, not database
Every data point is translated into ops language: structured and identifiable NDC and GPI codes are prominently ordered, with key identifiers being clear and meaningful. Technical fields remained accessible but secondary. Users finally trusted what they saw because it aligned with their expectations.
Macro of PAS view
Macro of Banjo or Connect View
Macro view on header/navigation
Configurations that anticipate
Smart drug lists and pre-filtered views eliminated the "where do I start" paralysis. Common workflows became one-click operations. The system remembered patterns and suggested next actions based on task context.
All / Connect / Banjo views stacked
NDC prompt field edit split screen
Safeguards that prevent disasters
Every high-risk action triggered contextual validation: price changes over 10% required confirmation, decimal anomalies flagged instantly, and bulk updates showed clear previews. The $203.70 pricing error that almost happened? Now impossible.
Macro price alert confirmation
OUTCOMEUser 'success' testing illuminated the excitement
During user acceptance testing, the excitement became undeniable. Clinical Review crashed our validation sessions just to test their workflows—like Banjo List exports—which transitioned flawlessly because I'd done my homework. Account Managers discovered they could pull AWP pricing and compare it against group rates for custom solutions. Teams scheduled for rollout months down the line wanted in immediately. What began as structured testing became organic advocacy.
Hours → Minutes
Operations drug management tasks
Zero training required
Users were up and running on day one
105% adoption
Exceeded target users
REFLECTIONUnshakeable foundation, transformative at scale
The UAT sessions were so rock-solid that for the first time during my SmithRx tenure, the product launched months ahead of schedule. The early delivery gave me runway to predict, design, and pre-validate the next three significant features: CSV import capabilities, a universal product detail template that could scale across the entire SmithRx ecosystem with dynamic permissions, and programmatic search that would enable true departmental scale. All designed, validated with stakeholders, and planned for rollout by day one launch.