B2B SaaS · Benefits

Benefits Destination

Enrollment season used to mean all hands on deck just to get sites out the door. We fixed that by building a system that could scale without scaling the team.

My role
Lead UX/UI designer
Type
Product design, design system
Platform
Web (B2B SaaS)
Industry
Employee benefits

Outcomes

faster site delivery
↑40%
enrollment completion rate
↓60%
client configuration time
100s
of clients, one system

Background

Where this started

Every year, enrollment season hit the same way. A pile of client requests, not enough time, and sites that were technically different but doing the same thing. Each one was built from scratch. Each one had its own quirks. And every customization added to the next one’s timeline.

The business needed a way to serve hundreds of employers without treating every single one as a unique project. I needed to figure out how to design for flexibility without losing the structure that made the product work.

The real problem wasn’t delivery speed. It was that we had no shared foundation. Everything was custom by default, even when it didn’t need to be.

Discovery

Understanding the full picture

Before jumping into solutions, I spent time mapping the actual pain. I talked to the teams configuring the sites, looked at the patterns across existing clients, and traced where the bottlenecks lived. What I found was that most variation was surface-level (branding, language, a few plan differences), but the underlying structure was nearly identical across clients.

That told me we didn’t need more flexibility. We needed better defaults, with flexibility in the right places.

Problem 1
No shared component library meant every client started from zero, even for identical UI patterns.
Problem 2
Clients couldn't self-serve basic changes. Every update required a designer in the loop.
Problem 3
Employees were dropping off mid-enrollment because the guidance was inconsistent across plan types.
Problem 4
No feedback loop. Once a site launched, we had no visibility into how it was actually performing.

Process

How I approached it

I started with the components that appeared in every single client site and worked backwards from there. Once I had a clear inventory of what was truly shared vs. what was client-specific, the system structure became obvious.

01
Audit existing sites
Catalogued patterns across 20+ client builds
02
Define the template layer
Separated structure from skin
03
Design the config tools
Made customization a client-facing feature
04
Test with real enrollment flows
Validated with actual employee journeys
Template builder. Component configuration view.

The work

What we built

The core of the solution was a templatized system where the structure was locked but the content, branding, and plan configuration were exposed as client controls. Clients could set up their own benefit websites without needing to open a support ticket for every change.

On the employee side, I redesigned the enrollment flow to actually guide people through their options rather than just presenting them. That meant surfacing the right information at the right step. Not everything upfront, and not making people hunt for it.

Client config panel

Employee enrollment flow

Final assembled templates across clients

Accessibility & inclusion

Built for every employee, not just the ones who read English

Benefit enrollment is already complicated. Doing it in a second language shouldn’t make it harder. For clients who opted in, the platform supported translated benefit content, a combination of machine translation and human review to make sure the language was accurate where it mattered most.

My job was designing the workflow that made that possible inside the platform. How content moved from English source to translated output, where human reviewers stepped in, and how the internal team managed it without a separate tool or process.

Machine translation got us 80% of the way there fast. Human review caught what automated tools miss: benefit terminology, plan names, legal language that doesn’t translate literally. The workflow had to make that handoff clean.

01
Content flagged for translation
Client opts in, content blocks are queued for processing
02
Machine translation runs
Automated first pass. Fast, covers standard content.
03
Human review queue
Reviewers check flagged content: terminology, legal language, plan names
Translation workflow. Side-by-side content editor with English source and Spanish translation.

Learnings

What I’d do differently

This project taught me a lot about designing systems that other people configure, not just systems that designers control.

1
Defaults matter more than options. The more we got the defaults right, the less clients felt the need to customize, and the better the end result looked.
2
Flexibility without guardrails is a trap. Early versions gave clients too much control. The sites they produced were inconsistent and some were genuinely hard to use. We had to pull some of that back.
3
The employee experience was almost an afterthought. Most of our conversations were about client tooling. We should have been talking about enrollment completion from day one.