shopping_cart E-Commerce Platform water_drop Industrial / B2B

Xylem Solver

Unifying 4 E-Commerce Sites into One Platform

Leading the UX strategy for a 3-year enterprise redesign — from fragmented legacy systems to a unified, research-driven platform targeting $4B+ revenue. Complete design system, 90+ applications support, and end-to-end UX framework delivery.

$4B+
Revenue target for 2026 with unified e-commerce platform — reducing task completion from 6-7 clicks to ≤4 clicks, increasing search success from 65% to 85%, and achieving 100% Design System adoption.
Solver UX Framework — 6-phase methodology for enterprise UX

Client

Xylem Inc. — Water Solutions

My Role

Lead Product Designer (UX Team)

Platform

Enterprise E-Commerce (Angular)

Tech Stack

Angular 17/18, AG Grid, Tacton Configurator, Azure, Material Design 3

Duration

3 years (2023–2026)

Brands

Bell & Gossett, Flygt, Goulds, Lowara + others

Partners

Accenture (Development & UX Audit)

Four Sites, Zero Consistency, Millions in Lost Efficiency

Xylem's digital ecosystem had grown organically over years — four separate e-commerce sites serving different brands (Bell & Gossett, Flygt, Goulds, Lowara), each with its own navigation patterns, component libraries, and user flows. The UX team supported 90+ applications, but Solver was the strategic priority: a unified platform that would consolidate these fragmented experiences into one cohesive system.

The existing Solver platform mirrored internal product structures rather than user mental models. Engineers, distributors, and technicians struggled with cognitive overload — too many clicks, redundant filters, hidden documents, and no guided workflows. Accenture's UX audit confirmed what VOC sessions revealed: users adapted despite the system, not because of it.

touch_app6-7 clicks per task

Users navigated through deep hierarchies just to reach product results — multiple entry points with no clear orientation or breadcrumbs.

support_agent40% support dependency

Users contacted Application Engineers for help because the interface provided no guidance, feedback, or self-service troubleshooting.

content_copy74% template duplication

Frontend codebase had massive redundancy — similar components built differently across brands, driving maintenance costs and inconsistency.

trending_down32% adoption rate

Only a third of target users actively used Solver — the rest relied on offline spreadsheets, phone calls, or competitor tools.

adjustMy Scope & Ownership

I was the sole Lead Product Designer on Solver — no UX process existed before my engagement. I owned the end-to-end UX strategy from discovery research through production handoff: the full information architecture, all 5 core user flows (Search, Compare, Configure, Quote, Projects), the Design System Foundation for 90+ applications, all error states and edge cases, account management flows, and sprint-by-sprint developer collaboration with Accenture. I also introduced the 6-phase UX framework that became the standard process for Xylem's digital team.

Owned end-to-end

IA, all user flows, component design, error states, account & permission model, dev handoff specs, and Confluence documentation

Decided independently

Design system architecture, navigation model, search strategy, Tacton integration approach, RFP/quote generation model, and governance structure

Aligned with stakeholders

Sprint scope and technical feasibility (Accenture), business priorities and KPIs (Product Management), Angular/AG Grid constraints (Dev leads)

Usage Guidelines — Icon system with sizes (Small 20, Default 24, Large 40, XL 48)

Strategic Choices That Shaped the Platform

Decision What I Chose Why Tradeoff
Design system foundation Material Design 3 with Xylem Blue brand tokens, not a custom system from scratch MD3 provided enterprise-grade, accessibility-compliant components out of the box — enabling design system governance across 90+ applications without building foundational patterns twice Teams unfamiliar with MD3's token architecture required training; adoption took multiple onboarding sessions before self-service was possible
Navigation architecture Flattened IA from 5+ levels to 3 levels maximum User Friction Reduction was the top priority — 6-7 clicks per task was the primary abandonment driver. Removing depth directly reduced drop-off in the search-to-selection flow Required restructuring all existing content hierarchies, URL patterns, and redirects — significant development effort with SEO implications
Search model Predictive Global Omnibar as the primary entry point across the entire platform Conversion Rate Optimization (CRO): VOC showed 35% of users abandoned when search failed. A unified entry point eliminates dead-end navigation paths and consolidates the highest-traffic interaction Complex backend integration across multiple product databases; relevance ranking required significant investment from Accenture's development team
Configurator integration Tacton embedded inline within the product detail flow, not launched as a separate context Launching the configurator externally broke the user's selection lifecycle — context-switching caused session abandonment. Inline embedding keeps the full Configure → Quote journey within one continuous session Deep Tacton API integration added frontend complexity; required close Accenture collaboration and thorough error state documentation for constraint violations
Quote & payment integration model In-app RFP/RFQ generation with pricing editor and PDF export — not a cart checkout or redirect to an external ERP B2B industrial procurement runs on Request for Proposal (RFP) and Request for Quote (RFQ) cycles, not shopping carts. Keeping pricing negotiation, discount management, and quote generation inside Solver eliminates the highest-friction CRO drop-off: the gap between product selection and commercial commitment. Keeping the user in one session directly improves conversion from "product found" to "quote submitted" No real-time ERP pricing sync in MVP — list prices with manual discount overrides by sales reps. Full SAP/ERP integration deferred to Phase 2. Pricing accuracy depends on distributor manual input rather than automated data feeds until that integration ships

6-Phase Methodology: From Discovery to Deployment

The Xylem UX team established a unified process to ensure every digital initiative follows measurable, user-centric practices. For Solver, this framework was applied systematically across every phase — combining human-centered design with engineering feasibility and design-system governance.

Solver UX Framework — 6 phases from Understand to Deploy

auto_awesome AI-assisted: The prompt for this framework visualization was structured with Claude to map UX phases, research methods, and design system governance into a coherent artifact. The visual was generated using Figma AI.

1

Understand & Align

Build shared understanding of business context, users, risks and product vision. Stakeholder interviews, VOC review, heuristic evaluation, competitor scan.

2

Research & Discover

Learn how users work today to identify pain points, needs and critical journeys. User interviews, contextual inquiry, persona refinement, journey mapping.

3

Define & Prioritize

Turn insights into clear problem statements, UX focus areas and MVP scope. Affinity mapping, requirements mapping, complexity scoring, roadmap planning.

4

Ideate & Plan

Explore solution patterns, IA and navigation concepts, define vertical slices. Co-creation workshops, concept sketching, user stories, acceptance criteria.

5

Design & Prototype

Translate concepts into detailed, Solver DS-aligned designs and interactive prototypes. High-fidelity layouts, interaction design, component documentation.

6

Test, Iterate & Deploy

Validate solutions with users, refine designs based on evidence. Usability testing, A/B testing, accessibility validation, dev implementation support.

Understanding Users at Enterprise Scale

Research wasn't theoretical — it was systematic data collection across multiple sources: 19 VOC sessions, Accenture's UX audit, heuristic analysis, and competitive benchmarking against 6 industry leaders.

The goal was to validate assumptions, quantify pain points, and create a shared evidence base that every design decision could reference.

VOC Sessions (Voice of Customer)

19 sessions with B&G / Flygt users + 4 Accenture interviews

Key themes: Navigation & wayfinding issues, complexity & cognitive overload, broken defaults and repeated work, search gaps for duty-point and replacement.

Heuristic Evaluation

Full Solver platform audit against usability principles

Identified inconsistent component naming, outdated typography, missing responsive states, and no feedback confirmation patterns.

Contextual Inquiry

3 Distributors, 2 Engineers, 1 Field Technician observed in-context

Desktop users lose progress when re-filtering. Field users can't scroll tables on mobile. Users export screenshots to compare products offline.

Competitive Analysis

Grundfos, Pentair, Armstrong, KSB, Taco, Systemwize

Competitors win on discipline and focus — fewer modes, consistent patterns, clearer guidance. Solver has capabilities but exposes them with friction.

auto_awesome AI-assisted: Claude was used to structure the research plan and synthesize VOC data into prioritized themes. Interview transcripts and documentation were reviewed with Gemini to surface consistent patterns across 19 sessions and identify signal from noise at scale.

Four distinct personas emerged from research — each with different experience levels, devices, goals, and pain points. Design decisions were validated against all four to ensure no user segment was left behind.

Solver Personas — Power Rep, Consulting Engineer, Inside Sales, Field Technician
VOC Key Themes and Representative Quotes
Prioritized UX Issues List with severity ratings
Solver vs Competitors — heuristic scoring across 7 flow areas

info Key insight: Systemwize praised for simplicity, Armstrong leads on mobile responsiveness. Opportunity: Blend guidance + mobility + structure under Solver Design System.

Journey J1 — Duty-Point Sizing & Selection with pain points per step

Baseline vs. Target — Measuring the Transformation

Every design decision was validated against measurable KPIs. These baseline-to-target metrics defined what success looked like for the Solver transformation.

groupAdoption Rate

  • Baseline: 32%
  • Target: ≥55%
  • Validation: Analytics

searchSearch Success

  • Baseline: 65%
  • Target: ≥85%
  • Validation: Usability Test

touch_appAvg Clicks/Task

  • Baseline: 6-7
  • Target: ≤4
  • Validation: Task Logs

paletteDS Usage

  • Baseline: 40%
  • Target: 100%
  • Validation: Audit

sentiment_satisfiedSatisfaction

  • Baseline: 2.5/5
  • Target: ≥4/5
  • Validation: VOC

Building the Xylem Design System from Scratch

I created the Foundation file for the entire Xylem organization — a token system based on Material Design 3, configured with Xylem Blue (#0066B3) as the primary brand color, Roboto typography, and an 8px spacing base. This foundation serves 100+ applications across the company.

auto_awesome AI-assisted: Claude was used throughout design system work — from structuring token documentation and naming conventions to drafting component specifications and writing usage guidelines for the shared library.

palette
Color Tokens
  • Xylem Blue (50–900)
  • Gray scale (50–900)
  • Semantic: Success, Warning, Error
  • Surface & Background
chevron_right
text_fields
Typography
  • Roboto (all weights)
  • Headings 1-6
  • Navigation & Menu
  • Labels, Buttons, Helper
chevron_right
space_bar
Layout
  • 8px base unit
  • Margins & Padding
  • Spacers system
  • Grid: 12/8/4 columns
chevron_right
deployed_code
Production
  • Angular 17/18
  • CSS Variables
  • Component library
  • 90+ apps support
Solver Design System — Content colors and Background colors with hex values
XDS Typography — Roboto typeface with all text styles for desktop
Layout Basics — Margins, Spacers, Padding with Local Variables
Usage Guidelines — Icon system with sizes (Small 20, Default 24, Large 40, XL 48)

Production-Ready Components for Enterprise Scale

Beyond the foundation, I designed critical components specific to Solver's needs — complex data tables for AG Grid, multi-brand filter systems, global search with predictive results, and responsive navigation patterns that work across desktop and mobile.

Solver Template — Full page layout with header, navigation, grid, and footer
Global Search Omnibar — 4-step variants for Series, Catalog, Replacement, Application, Product, Brand
Header Components — Global Navigation Menu, Market selector, Language selector with spec answers
Form Element Value — Properties matrix with sizes and states
Component Card — Default, Product, Product Family, Line variants with real brand examples
Component Table — Multiple table variants and Table Results configurations
Menu Item and Filters — Flygt, Bell & Gossett brand filters, Table Filters, Basic Filter with technical parameters

Agile Delivery with Stakeholder Alignment

Design wasn't done in isolation. Every flow was developed through co-creation workshops with UX, Product Managers, Dev Leads, and Accenture consultants. User stories were written with acceptance criteria, and prototypes were validated before any code was written.

I used Figma Make and Claude Console for rapid prototyping — generating variations faster than traditional wireframing, then refining with stakeholder feedback in real-time sessions.

Workshop annotations with sticky notes showing design decisions for Projects flow and Quote modal
Projects Details flows 3.7-3.10 with documented User Stories and multiple UI states
Header Secondary Menu Flows — Applications and Products navigation journey with connected screens

From Search Query to Quote — The Full Selection Lifecycle

The core user journey in Solver isn't a single action — it's a multi-step selection lifecycle that mirrors a B2B procurement flow: find the right product, configure it to spec, validate the selection, and generate a deliverable. Every step was designed to reduce friction and prevent dead ends.

B2B Procurement Context

Unlike B2C e-commerce, Solver's end goal isn't a cart checkout — it's an RFP (Request for Proposal) or RFQ (Request for Quote) submitted to a distributor or regional sales team. The flow involves multiple roles: an Application Engineer or Consulting Engineer selects and configures, a sales rep reviews pricing and applies discounts, a manager approves the quote before submission. Every friction reduction decision — flattened IA, predictive search, inline configuration, in-app pricing editor — was validated against this multi-stakeholder, multi-session procurement reality.

search
Entry
Global Search
User types into the Omnibar — predictive results appear immediately
chevron_right
filter_list
Step 1
Filter & Compare
Apply duty-point filters, compare products side-by-side in AG Grid
chevron_right
tune
Step 2
Configure
Run Tacton configurator inline — validate parameters against model constraints
chevron_right
folder_open
Step 3
Add to Project
Save configured product to a project for batch quoting or future reference
chevron_right
request_quote
Step 4
Generate Quote
Review line items, pricing, and product specs before finalizing
chevron_right
task_alt
Success
Export & Share
Quote downloaded as PDF, shared with client or submitted to procurement
Entry Homepage — Global Search
Homepage with global search and predictive results
Step 1 Advanced Search — Filter & Results
Advanced search with duty-point filters and AG Grid results table
Step 1b Cards View — Application Browse
Application-based cards view with sub-application drill-down
Step 1c Express Select — Brand Entry Point
Express Select — brand-specific model selection with curves and table results
Step 2 Compare Products — Side-by-Side
Compare Products — side-by-side curves, hydraulic data and efficiency comparison
Step 3 Product Configurator — Tacton Inline
Product Configurator — inline Tacton integration with pump curve and motor parameters
Step 3b Product Details — Specs & Features
Product Details — system application data, fluid properties and product description
Step 3c Documents Repository
Documentation tab — filterable document repository with type tags and download actions
Step 3d Accessories — Compatible Parts
Accessories tab — compatible accessories with compatibility check and inline add-to-project
Step 4 Add to Projects — Create Project
Add to Projects modal — create new project with name, customer and application context
Step 4b My Projects — Projects List
My Projects list — sortable table with status badges, share status and actions
Step 5 Project Details — Overview
Project Details — product line items with pricing summary and customer context
Step 6 — Success Pricing & RFP — Generate Quote
Pricing and RFP tab — line item discounts, margin summary and Request for Proposal generation

Designing for Failure at Enterprise Scale

In an enterprise selection tool, error states aren't edge cases — they're workflow blockers. A failed search or configurator conflict can send an engineer back to a phone call with Application Engineering, exactly what Solver was built to eliminate. Every failure mode was documented and designed with three rules: explain what happened, show the closest valid path forward, and never lose user progress.

search_off Error State 01
No Search Results — Empty State

When a query returns no products, the empty state shows query suggestions, related categories, and an "Ask Application Engineering" escalation path. Users are never left on a dead-end page — they always have a next action.

CRO impact: VOC data showed search abandonment was the #1 drop-off point. A useful empty state recovers users who would otherwise leave the platform.
warning Error State 02
Configurator Constraint Violation

When selected duty-point parameters produce no valid model in Tacton, the configurator shows an inline constraint message identifying the conflicting values and suggests the closest valid alternatives — keeping the user in the configuration flow rather than resetting to search.

Design decision: Never force a full reset — show the delta between what was requested and what's achievable so engineers can make an informed substitution.
receipt_long Error State 03
Quote Generation Failure

When the quote service is unavailable, the system auto-saves the project state and shows a service status message with a retry option. Users don't lose their configured selections — the quote generates automatically when the service recovers.

Design decision: Auto-save at every project action — losing a complex configuration after a service error would drive users permanently back to offline tools.
manage_accounts Error State 04 — Account & Permission Management
Access Control, Brand Subscriptions & Self-Service Account Model

When a user tries to access a brand, feature, or pricing tier outside their account permissions, Solver shows a contextual access message — not a generic 403 — specifying the exact role required and providing a direct self-service path to request it. No IT ticket, no dead end.

User Role Model

Three role tiers govern access: Application Engineer (full configuration + pricing), Distributor (browse + quote, restricted net pricing), and Read-Only (product research, no quote generation). Roles are set at account creation and upgradeable via self-service request — no admin required for standard upgrades.

Brand Subscriptions

Each user account has a subscribed brand set (e.g., Bell & Gossett only, or Flygt + Lowara). Accessing an unsubscribed brand shows an upgrade prompt with a business justification field. Subscription scope controls which product families, pricing tiers, and technical documents are visible — critical for multi-brand distributors managing separate contractual agreements per brand.

Self-Service Profile

Users manage brand subscriptions, unit defaults (metric/imperial), frequency settings, notification preferences, and language from a single profile. Role upgrade requests and access history are visible inline. Designed to reduce the 40% support dependency by giving users direct ownership of their access — the same friction reduction principle applied to the product flows.

Design rationale: In B2B industrial, access control is a real mid-workflow blocker — engineers get locked out of brands mid-project, or distributors discover they need pricing access only after 20 minutes of configuration. Every permission boundary in Solver was designed with a self-service escape hatch to prevent support calls, preserve workflow state, and keep users in the platform.
token
Design System Consistency

Same components, tokens, and patterns across all 90+ apps — users feel at home the moment they switch tools

verified
Inline Validation

Configuration forms validate in real-time — engineers know immediately if a parameter is out of range, before submitting

step
Progress States

Multi-step flows show where the user is and what's saved — no surprise resets or unexplained loading states

save
Auto-Save Projects

Every project action is persisted — users can close and return without losing configured selections or quote drafts

Deliverables & Impact

science
UX Research Framework 6-phase methodology with VOC integration, persona cards, journey maps, and competitive benchmarks
token
Xylem DS Foundation Complete token system (colors, typography, spacing, elevation) serving 90+ applications company-wide
widgets
Solver Component Library 40+ components including AG Grid tables, multi-brand filters, global search, navigation, and forms
account_tree
Complete Flow Prototypes Search, Compare, Configure, Quote, and Projects flows with full state coverage and responsive breakpoints
menu_book
Handoff Documentation Confluence wiki with component specs, user stories, acceptance criteria, and developer guidelines
handshake
Accenture Partnership Continuous collaboration with development partner — workshops, reviews, sprint planning integration
Working with proto-personas when you can't access end users. Without direct access to external customers, we built proto-personas from VOC data, support tickets, and internal Application Engineers who knew user pain points intimately. These proxies were validated iteratively as research expanded — starting broad, then refining with each data point.
Negotiating with development teams requires shared evidence. Design proposals that seemed obvious from a UX perspective often conflicted with technical constraints or existing code patterns. The key was building a shared evidence base — research findings, competitive benchmarks, and KPI targets — that both sides could reference when making tradeoffs.
Enterprise scale requires governance, not just components. With 90+ applications, creating a design system wasn't enough — we needed adoption mechanisms, documentation standards, and clear ownership. The Foundation file became a shared contract between UX, development, and brand teams.
AI-assisted prototyping accelerates exploration without replacing judgment. Using Claude Console and Figma Make let me explore more variations faster — but the actual design decisions still came from research, stakeholder feedback, and domain expertise. AI was a production multiplier, not a substitute for design thinking.

Design System Governance: Enabling 90+ Teams Without Becoming a Bottleneck

The Problem

With 90+ applications sharing the Xylem Design System, every DS update had the potential to break UI in apps I hadn't designed. Too much control = bottleneck. Too little = fragmentation and brand drift across the product portfolio.

The Decision

Two-layer governance: a locked Foundation layer (tokens, primitives, core components owned by UX) and an extensible Application layer (teams build on top but can't modify the Foundation). Changes to the Foundation go through a documented review process with a clear breaking-change policy.

The Tradeoff

Some teams found the locked Foundation layer frustrating — they wanted to modify core components for their specific use case. Each exception required a formal proposal, which slowed down feature delivery in the short term but maintained long-term consistency across the platform.

90+
Applications served by the Xylem Design System Foundation
40+
Production-ready components in the Solver Component Library
−40%
Task completion time — from 6-7 clicks to ≤4 with flattened IA
$4B
Revenue target for the unified Solver platform by 2026

Qualitative impact: The 40% support dependency on Application Engineers was directly tied to platform friction — users couldn't self-serve. By reducing friction through flattened IA, predictive search, and inline configuration guidance, Solver moves the conversion point from "call for help" to "complete task independently." The DS consistency also reduces the onboarding cost for new Xylem products joining the platform.