Enterprise Data Platform (400+ Screens)

Context

Jellyfish is a global digital media company serving enterprise clients through data-driven marketing services. The company operates a suite of interconnected data products -- J+IQ for competitive analysis, J+Bidding for ad spend optimization, Share of Model for AI brand analytics, and others -- all of which depend on a shared data infrastructure for their underlying data collection, processing, and delivery.

That infrastructure is J+Lake. It is the platform that manages how data flows into the entire Jellyfish product ecosystem: collecting data from 22 advertising and analytics connectors (Google Ads, Search Ads 360, Display & Video 360, Campaign Manager 360, and 18 others), processing 20,000+ data collection jobs daily, and managing the credentials, templates, and monitoring systems that keep the data flowing for 500+ clients across 32,000+ accounts.

This is not a user-facing analytics tool. It is the foundational layer that every other product depends on. When J+Lake breaks, everything downstream breaks -- reports fail, dashboards go dark, client deliverables miss deadlines. Designing for this kind of infrastructure means designing for reliability, clarity, and scale simultaneously.

The Challenge

J+Lake's complexity came from three intersecting dimensions.

Technical fragmentation: 22 data connectors, each with different data models, authentication methods, and operational patterns. Google Ads handles credentials differently than SA360. DV360 uses different scheduling cadences than CM360. Templates that collect data run at frequencies ranging from every 15 minutes to weekly, each with different error states, retry logic, and monitoring needs. The platform had to present a unified experience across all of this heterogeneity.

Three distinct user roles with different mental models: CSMs (Customer Success Managers) think in terms of clients and accounts -- they need to validate credentials, link advertising accounts, and manage multi-connector setups for their book of business. Data Engineers think in terms of templates and pipelines -- they build, modify, and monitor data collection templates, troubleshoot failures, and manage scheduling. Paid Media Specialists think in terms of campaigns and performance -- they navigate client data to optimize advertising spend. A single platform had to serve all three without forcing any role into an unnatural workflow.

Scale that shows in the UI: This was not a prototype-level challenge. Client lists run to hundreds of entries. Template tables show 100+ linked clients per template. Monitoring dashboards need to surface meaningful signal from thousands of daily jobs. Pagination, filtering, search, and status visualization at this scale require careful information architecture -- every decision about what to show, what to hide, and how to organize affects whether a CSM can find a problem before it cascades into a client escalation.

And underlying all of this: users were uncertain about the tool's purpose. Feedback indicated that people did not understand how J+Lake would integrate with their daily responsibilities. I was designing for a product whose value proposition was not immediately obvious to its own users.

My Role

Title: UX Designer

Actual scope: Lead product designer for the entire J+Lake platform -- end-to-end ownership of discovery, research infrastructure, design, and usability testing across all user roles

Team: Cross-functional collaboration with product owners, tech leads, and developers; I was the sole design authority

Strategic influence: Developed a Tech/UX/Business review prioritization framework that directly influenced the product roadmap

The scope of this work -- 400+ screens across 22 connectors for 3 user roles, with research infrastructure built from scratch -- is Staff-level design under a UX Designer title. I owned not just the interface design but the research methodology, the prioritization framework, and the product strategy influence that connected user insights to development decisions.

Research & Discovery

Structured Discovery Questionnaire: I designed a comprehensive discovery questionnaire covering six theme areas: data collection monitoring priorities, automated health check workflows, template management ideal processes, historical data usage patterns, template discovery and organization, and template quantity and customization needs. The questionnaire went through iterative refinement -- I created both draft and final versions, testing question phrasing to ensure I captured actionable insights rather than generic feedback.

This was not a single survey. It was a research instrument designed to systematically map the workflows and mental models of three distinct user roles across a complex infrastructure product.

Workshops with Personas and Journey Mapping: I facilitated workshops that progressed from understanding users to understanding their workflows to designing solutions. I created persona cards for at least 4 user types, mapped user journeys for each role, developed user stories, and designed user flows. The workshop artifacts show a structured progression: discovery and goals first, then J+ stories (user scenarios), then journey mapping, then flow design. Each step informed the next.

Role-Specific Usability Testing: I created dedicated test scripts for each user role with defined objectives, step-by-step instructions, and expected outcomes across 5 distinct test scenarios. For CSMs, I tested client validation workflows -- verifying credentials across CM360, DV360, Google Ads, and SA360, linking and unlinking accounts, managing account transfers, and identifying error states. For Data Engineers, I tested template management and monitoring workflows. For Paid Media Specialists, I tested navigation and data access patterns.

I conducted 20-minute remote usability testing and in-person interviews with 5 users across all three roles (2 CSMs, 2 Data Engineers, 1 Paid Media Specialist) and produced a structured findings report organized by category (UI/UX, Template Management, Data Management, System Features, Security) with quantified metrics, user quotes, and prioritized recommendations.

Design Process

Designing for Three Mental Models: The core design challenge was creating a unified platform that respected each role's mental model without fragmenting into three separate products. My approach was to design shared infrastructure (navigation, visual language, interaction patterns) that maintained consistency with the broader J+ product suite, while creating role-specific workflows within that shared framework.

For CSMs, I designed client validation flows that mirror their account-management mental model. The Validate Client interface presents multi-connector account management for CM360, DV360, Google Ads, and SA360 in a unified view -- allowing CSMs to see all of a client's connected accounts, their status, and any credential issues in one place. Account actions (transfer, unlink, add) are contextual to each connector, because different connectors have different requirements, but the interaction pattern is consistent.

For Data Engineers, I designed template management systems with the depth they need -- scheduling controls (from every 15 minutes to weekly), modification tracking, run history, linked client visibility (showing templates serving 1 to 101+ clients), and status indicators. The template list shows 15+ global templates with filtering by status, product, modification date, and frequency.

For operational oversight, I designed monitoring dashboards that surface real-time pipeline status: jobs count by status with daily budget/spend/ideal spend charts, a detailed job table with status indicators (active, paused, error, ended), connector-specific views, and notification systems across Slack and email. The monitoring view answers the question every operations person asks first: "Is anything broken right now?"

The Notification System: Data pipeline monitoring demands proactive alerting. I designed a multi-channel notification system spanning Slack (per-channel alerts and dedicated Slack App concepts), email notifications, and a centralized notification settings interface. Users needed to be alerted immediately when a pipeline failed -- but not overwhelmed by noise from routine operations.

Influencing the Product Roadmap: Beyond interface design, I developed a Tech/UX/Business review prioritization framework that evaluated each feature across four dimensions: Uncertainty (Green/Yellow/Red), Effort, UX Value, and Business Value. I applied this framework to 10+ features including platform account listing, access testing, scheduled collection monitoring, data catalog API, manual mapping, and admin UI. This tool translated UX research insights into a format that product and engineering could use for roadmap decisions, bridging the gap between what users needed and what got built.

Solution

The delivered J+Lake platform spans 400+ screens unified under a coherent design language:

Monitoring Dashboard: Real-time pipeline monitoring with jobs count by status visualization (active, paused, error, ended), daily budget and spend charts, detailed job tables with inline status indicators, and connector-specific filtering. Designed to answer "is anything broken?" within seconds of opening.

Template Management: Comprehensive template creation, modification, and monitoring system. Templates display linked client count (1 to 101+), scheduling frequency (15 minutes to weekly), last run timestamp, modification history, and status. Supports creation, duplication, editing, and archiving workflows.

Client Validation: Multi-connector credential management enabling CSMs to verify, link, unlink, and transfer advertising accounts across Google Ads, SA360, DV360, and CM360. Validates credentials per connector, shows account status, and surfaces error states requiring attention.

Notification System: Multi-channel alerting across Slack (per-channel and dedicated app) and email, with centralized settings for controlling notification preferences, thresholds, and routing.

Unvalidated Client Management: Dedicated workflows for identifying and resolving clients with missing, expired, or invalid credentials -- preventing data collection failures before they impact downstream products.

The design maintains visual consistency with the broader J+ product suite (navigation patterns, card-based layouts, icon language, typography) while accommodating the infrastructure-specific demands of data pipeline management. Users transitioning between J+Lake and other J+ products encounter familiar patterns.

Results & Impact

Usability Outcomes

• 8.9/10 average usability score across all three user roles
• 100% task completion rate across all tested tasks and scenarios
• Users praised the unified experience: "It is surprising how different sources have different ways of being connected, and here everything is streamlined and easy to set-up. It is easy to manage so that even non tech people or experts can manage them. Really useful for CM or CSM."
• The design was described as having a "smooth and simple UI" that users appreciated for "resembling other products in the J+ suite"

Scale Served

• Platform supports 500+ clients and 32,000+ accounts
• Processes 20,000+ daily jobs across 22 data connectors and 8 products
• Templates serve from 1 to 101+ linked clients each, running at frequencies from every 15 minutes to weekly

Design Scope

• 400+ screens designed across monitoring, template management, client validation, notification systems, and unvalidated client workflows
• 3 user roles served through a unified platform with role-specific workflows
• 22 connectors unified under a coherent interaction model

Strategic Impact

• Tech/UX/Business review framework directly influenced the product roadmap, translating UX insights into prioritized engineering work
• Research identified that users needed proactive communication about J+Lake's role and benefits, leading to adoption-support recommendations alongside interface improvements
• J+Lake's coherent design enabled the products built on top of it (J+IQ, SOM, J+Bidding) to maintain data reliability for their own users

Reflections

J+Lake taught me something fundamental about the relationship between infrastructure design and user-facing product design. When I designed J+IQ or Share of Model, I was designing products people interact with directly -- tools they open, explore, and make decisions from. J+Lake is different. Most of its value is invisible. When it works well, nobody notices. Its success is measured not in delight but in the absence of failure.

Designing for that kind of product requires a different mindset. Every workflow I designed had to account for error states, edge cases, and scale implications that would never surface in a prototype walkthrough but would absolutely surface at 20,000 daily jobs. The monitoring dashboard needed to surface the one failing pipeline among thousands of healthy ones. The client validation flow needed to handle the case where a credential expires mid-collection across multiple connectors simultaneously.

The biggest lesson was about designing across mental models. CSMs, Data Engineers, and Paid Media Specialists do not just have different tasks -- they think about the same system in fundamentally different ways. The information architecture decisions I made (how to structure navigation, what to show on each view, how to organize filtering) had to serve all three without defaulting to the vocabulary or priorities of any single role. The 100% task completion rate across all roles tells me those decisions worked.

Core patterns this project demonstrates

• Territory Mapping: Systematic discovery across 6 theme areas with role-specific research instruments before designing -- understanding the 22-connector ecosystem, 3 user roles, and their distinct workflows before proposing any solution
• Infrastructure Building: Research questionnaires, role-specific test scripts, structured findings frameworks, and a Tech/UX/Business prioritization framework -- all built from scratch as reusable research methodology infrastructure
• Silo Bridging: Designing a unified platform that bridges three organizational functions (CSMs, Data Engineers, Paid Media Specialists) with fundamentally different mental models and workflows
• Operating Above Title: 400+ screens of enterprise data platform design, research practice building, and product roadmap influence as a UX Designer -- Staff-level scope under a designer title

J+Lake is also the foundation that every other Jellyfish product depends on. I did not just design individual products at Jellyfish -- I designed the interconnected system. J+Lake feeds J+IQ, SOM, J+Bidding, and the rest of the suite. Understanding that ecosystem -- how data flows from connector to template to pipeline to product -- is what made it possible to design coherent experiences across the entire J+ platform.

Key Artifacts