Case study - Internal tooling · Entitlements unification

One place to configure every customer

A unified internal platform replacing a scatter of legacy tools for setting up contracts, offerings, and feature availability. I designed the end-to-end Client Configuration experience - a searchable customer directory, a six-step setup wizard, and a customer profile with full audit history - validated it with a working coded prototype, and handed it to development after all five usability test participants preferred it over the current tooling.

My role
Lead product designer - research, IA, UI, handoff, development support
Team
PM, engineering
Timeline
Q1 & Q2 2026
Tools
Figma, Claude Cowork & Code, UserTesting
Client Configuration - the customer directory

Overview

Context · Problem · Constraints

01

Context

A configuration model with real depth, and nowhere to put it.

Each customer buys a contract tier that determines which healthcare offerings their members receive - expert medical opinion, bariatric surgery, cancer care, trusted-partner programs. Every offering can be tuned further: bariatric surgery might be voluntary use for one customer and a mandatory consult for another, and eligibility can be limited to specific member groups.

Before this project, that configuration lived across multiple internal tools, personal notes, and slide decks acting as unofficial sources of truth. There was no reliable answer to "what does this customer actually have?" and no way to tell who changed a configuration, or when.

"Searching through a mountain of configurations."

The intended users - implementation, content, and sales teams - are non-developers, configuring all of this without engineering help.

02

The problem

Internal teams had no unified, reliable way to configure and audit a customer's entitlements - and no way to review the whole picture at once.

Setup meant stitching together contracts, toggles, eligibility rules, and partner settings across disconnected tools.

Who is affectedImplementation, content and sales teams as primary users; engineering and support, who inherit the misconfigurations; and ultimately members, who see wrong or missing benefits.
Cost of inactionMisconfigurations translate directly into members denied care they're entitled to, support escalations, slower client onboarding, and contractual risk.

03

Constraints

Domain complexityOfferings differ structurally - simple on/off toggles, multi-value parameters, group-level eligibility, launch dates, and mutual-exclusivity rules between offerings.
TechnicalEvery concept on screen needed a well-defined counterpart in the entitlements service's domain model.
OrganizationalTeams used conflicting vocabulary for the same concepts - the work had to unify language, not just UI.

Discovery

Research · Synthesis

04

Research

Interviews with implementation managers, plus an audit of the existing tools and documentation.

Three insights that shaped the direction

01Configuration is a sequence, not a form - who the org is, what they bought, how offerings behave, which partners, app settings, review.
02Language was a first-order problem. Overloaded terms made a shared glossary a prerequisite for a shared UI.
03Change over time matters as much as initial setup. Contracts get amended and offerings deactivated - the tool needed history and scheduled changes, not just a create flow.

05

Synthesis

Two core problems, in order of weight.

Fragmentation

No single place where a customer's configuration lives. Setup and lookup were spread across multiple tools and informal records - so there was no authoritative answer to "what does this customer have?", no audit trail, and configuration knowledge locked in individual heads.

Design response: one platform covering the full lifecycle - setup, overview, change history, deactivation.

Developer dependency

The people who own the configuration couldn't perform it. Implementation, content and sales teams knew what each customer should get, but changes routed through engineering. Every request became a ticket, every ticket a delay, and engineering time went to data entry instead of building.

Design response: a guided, self-serve flow safe enough for non-developers - stepped structure, per-step validation, incompatibility guards, and a mandatory review before anything publishes.

Supporting workstream - aligning the domain language

Not the core problem, but a prerequisite for a self-serve tool: if users configure entitlements without an engineer translating, the words on screen must mean the same thing to sales, implementation and the backing service. We restructured the draft glossary into a clean ladder, which then shaped the wizard's information architecture - contracts and attributes each got their own step because they are distinct domain concepts.

ContractOfferingAttributeEntitlement

Guiding principle

Make the invisible contract visible - every entitlement decision should be inspectable, reviewable and traceable, by the people who own it.

Design

Process · Iteration · Validation

06

Process & iteration

Weekly crit and review with engineering throughout development.

Language as a collaboration tool

The glossary was built jointly with PM and engineering, with "replaces…" annotations to ease the transition. It kept UI and service responsibilities aligned as both moved.

Contract configuration depth

From a single contract to multiple contracts, group targeting (all members versus selected groups), per-group launch dates, and opt-out rules.

Contracts - selection

Rule complexity

Dedicated treatments for eligibility rules, incompatible offerings, and inline errors - the places where a wrong click has a real member consequence.

Org details - population segmentation and eligibility rules

From setup tool to lifecycle tool

Later iterations added the customer profile: history, scheduled changes, and structured deactivation flows. The tool stopped being about day one and started being about year three.

The prototype as a communication artifact

Alongside the static designs I built a working prototype in Claude Code - real step navigation, toggle states, rule builder, validation. That made unmoderated remote testing possible and surfaced flow-level problems static frames can't. It also gave engineering an unambiguous behavioural reference: reviews happened against working interactions, not hypothetical descriptions.

07

Try the prototype

The coded prototype used in the usability study - six-step wizard, live validation, rule builder, and the Customer Overview.

Client Configuration - prototypeOpen in a new tab ↗︎

08

Validation

Unmoderated study, 5 internal participants - the tool's real future users.

The scenario: set up a fictional customer, "SkyRoute Airlines", end to end - a Wayfinding Plus contract, expert medical opinion limited to specific employee groups and categories, one offering excluded, trusted partners with an age-based eligibility rule - then review the Customer Overview and audit log.

What validated

"I'm super excited to use this. It seems way more user friendly than current tools."

5 / 5participants preferred it over the current tooling
Overviewthe strongest moment of the test - everyone immediately saw a single source of truth replacing lookups, notes and Slack threads
Historylanded unanimously - a direct fix for the "who changed this?" problem

The step-by-step structure, contract selection, include and exclude toggles, and the Attributes step all tested well.

What it surfaced
The one real problemOrg-level eligibility rules confused all five participants - defining rules before configuring offerings matched nobody's mental model. Crucially, most came around once they experienced the reuse logic, so the fix is onboarding, inline explanation and clearer subgroup sourcing labels, not restructuring the flow.
A second language layer"Subgroups", "eligibility groups" and "eligibility rules" were still being used interchangeably.
Interaction fixesA missed Apply button, a calendar-only date picker, an ignored progress bar (now per-step headlines), and an In-App Settings step needing context and correct defaults.
A domain catchMutually exclusive second-opinion products - something no review would have found, now a system-level conflict validation.

The study produced 6 flow recommendations and 11 design fixes - auto-save with draft mode, manual date entry, hover tooltips on rule chips, "select all" for partner programs, open-ended age ranges, customer archiving, an exportable overview - all folded into the designs before handoff.

Solution

Three pillars

09

Customer directory

A searchable list of every customer with their contracts, offerings and launch status - active, pre-launch, terminated, expiring - filterable by each. One glance answers "what does this customer have?"

10

Configure New Client

Six steps mapping one-to-one onto the domain model.

01Org detailsIdentity, IDs, child organizations, subgroups and eligibility rules
02ContractsTier selection, group targeting, launch dates, opt-out rules
03AttributesOfferings grouped by domain, each with its parameters and exceptions
04Trusted partnersPartner programs and segment setup
05In-app settingsMember-facing app configuration
06ReviewA full validated summary before publishing

Craft details: autosave with a visible timestamp, and validation both on continue and at the final step.

Attributes - configuration
Trusted partners - segments configuration
Review - validated summary before publishing

11

Lifecycle management

The customer profile covers the full life of a configuration: change history, scheduled and planned future changes, individual end dates, notes, and structured deactivation - whole customer or per offering.

Client Profile - audit log

Impact

Outcomes · Reflection

Outcomes & impact

Validated before a line of production codeUnanimous positive reception from target users, with Overview and History as unambiguous wins. "A fantastic start for new customer configuration - with just a few small adjustments we should be in a great place."
De-risked the launchTesting showed the novel eligibility-rules concept needs onboarding, not redesign - avoiding a costly late restructure of the core flow.
Handed off completeA full component library, specified edge cases, and 17 test-driven refinements incorporated before development began.
One vocabularyThe unified glossary was adopted across design, engineering and services.

12

Reflection

Naming is designThe highest-leverage early work wasn't a screen - it was killing ambiguous terms. Once the vocabulary settled, the information architecture was largely designed. The subgroups confusion in testing proved that work isn't one-and-done.
A coded prototype changes the conversationIt moved reviews from "I think this would feel…" to observed behaviour - and enabled unmoderated testing that static frames and basic click-throughs can't support.
Users forgive a new mental model if the payoff is visibleParticipants who initially rejected org-level rules came around after experiencing the reuse logic. A static walkthrough would likely have produced a false "redesign this" signal.
Test with people who live in the domainThe mutual-exclusivity rule came from a participant's product knowledge - no heuristic review would have caught it.