Context
SubHub is the internal tool NYT customer care agents use to view and manage subscriber accounts. It was structured to mirror the customer-facing Account experience: separate areas for subscriptions, billing, history, and more.
The New York Times · Customer Care · Subscriber Hub
Information Architecture Refresh for SubHub
Overview
Goal
Streamline navigation, improve discoverability, and build a more intuitive structure where information is logically grouped and easily searchable, so agents work more efficiently and resolve inquiries faster.
Problem
That nested hierarchy made it difficult for agents to find what they needed quickly, contributing to longer handle times and a frustrating experience for both agents and customers.
Opportunity
Re-architect navigation, organization, search, and labeling in one coherent push, validated with agents and tree testing, so SubHub scales as workflows and product surfaces grow.
02
The process
Four phases, from open-ended discovery, to validated concepts, to a shipped re-architecture.
Discovery
User research; mapping agent painpoints.
Define
User personas; generating hypotheses.
Develop
Creating design concepts; validating concepts with agents.
Deliver
Aligning with cross-functional partners; VQA with engineering.
03
The problem with SubHub's hierarchy
SubHub was designed to mirror the customer-facing Account experience, a nested hierarchy where subscriptions, account overview, billing, and history each lived on separate pages.
That structure works well for customers, who visit infrequently and one page at a time. It breaks down for power users like agents, who need to access multiple data points quickly. As a result, agents had to navigate across multiple sections, making their workflow fragmented and time-consuming.
04
Understanding the problem
User Personas
Understanding agent navigation patterns.
To better understand the problem, I reviewed existing research on agent behaviors related to navigation and search, focusing on their mental models and expectations. I analyzed observation studies that revealed how agents move between platforms, gather customer information, and locate necessary data. Using these insights, I created user personas to map out the primary pain points affecting our key users.
The primary pain points.
Three issues surfaced across nearly every research artifact.
Data fragmentation
Customer data was scattered across different pages or systems, making it difficult for agents to obtain a comprehensive view of a customer's history and interactions.
Fig. 03·Profile page · existing taxonomy
1
Eight tabs, and growing
Every new feature was filed under Profile rather than placed where the workflow expected it, turning the section into a catch-all for unrelated capabilities.
Critical information scattered
Customer details that belonged together were split across three to four subpages, forcing agents to memorize where each fact lived.
No anchor view
Agents lost orientation on every navigation, there was no single view that held the whole customer at once, so context had to be rebuilt manually.
Inefficient search functionality
Agents reported that search rarely yielded accurate or relevant results, slowing the process of locating customer accounts.
Labeling inconsistencies
The language used in customer-facing platforms differed from the terminology in agent platforms, creating inconsistencies that lead to confusion and inefficiency.
Customer view
"Bonus Subscriptions", what the customer sees in Account.
Agent view in SubHub
"Shared Subscriptions", same feature, different name. A constant source of confusion in calls.
I'm always five tabs deep before I can answer a basic question. The platform fights me on every call.Customer Care Agent, Observation study, Discovery phase
05
Coming up with hypotheses
Based on insights from research, I developed a set of hypotheses focused on the four key IA dimensions, and predicted how improvements in each could improve agent workflows.
01
Navigation
A flatter hierarchy will give agents quick, direct access.
Agents will benefit from a flatter navigational hierarchy because it provides quick and direct access to information and features without weeding through multiple pages to find what they need.
02
Organization
A centralized model keeps customer context in view.
Adopting a centralized organization style will ensure easy access to customer data, subscription details, past interactions, and any other pertinent information necessary for processing requests.
03
Search
Multi-attribute search will replace search-as-navigation.
A mix between an integrated and customer-focused search will enable agents to search across a few important resources, while allowing them to narrow in on key customer attributes to generate the most useful results.
04
Labeling
Familiar terminology will reduce comprehension load.
Using terminology familiar to agents can translate to more comprehension, and increase agents' ability to assist customers more efficiently.
Evaluating hypotheses with cross-functional partners.
I facilitated a workshop with internal stakeholders, product manager, operations manager, design manager, and engineering lead, to pressure-test these hypotheses and assess their feasibility. We explored two questions: would these ideas work in practice? And could we redesign the platform to support agent workflows while aligning with business goals?
Overall, the team was receptive, but the biggest challenge was engineering feasibility. Suggestions for a more robust search and a centralized organization model would require fundamental changes to how information is structured and presented across multiple pages.
06
Creating design concepts
Building on insights from research and stakeholder feedback, I developed concepts that addressed each of the four IA dimensions.
01
Navigation
To simplify wayfinding and reduce cognitive load, I restructured SubHub by consolidating six existing pages into three core sections.
- Account Overview, a central hub for general customer context.
- Subscriptions, a dedicated page for all subscription-related details.
- Account Activity, a single source for customer interactions and historical data.
This reduction in pages aimed to make navigation more intuitive, minimizing unnecessary clicks and helping agents access key information more efficiently.
02
Organization
Grouping related information and adding lightweight filters so agents could navigate without losing customer context.
Concept 03
Multi-attribute customer search
Letting agents look up customers using multiple attributes, beyond just email and account number.
Strengths
- Lookup still works when agents only have partial identifiers
- Filters map to how agents think about accounts (status, subscription, payment)
- Reduces dead ends when email or account number is unknown
Tradeoffs
- Depends on data quality and cross-system indexing
- More fields can add load if the UI does not prioritize the essentials
- Engineering work to expose and keep criteria reliable at scale
Concept 04
Consistent labeling across SubHub
Aligning SubHub's terminology with the language agents use internally, and with what customers see, so the same word means the same thing on every screen.
Strengths
- Shared vocabulary between agent tools and customer-facing surfaces
- Labels match agents' mental models from training and legacy tools
- Less time clarifying terms while resolving customer issues
Tradeoffs
- Renaming can confuse agents during a transition period
- Requires coordination with product and editorial on customer copy
- Historical records may still use older terminology
07
Validating with users
A focus group of five agents, four objectives, four hypotheses to test.
After developing these concepts, I tested them with agents through a focus group with five participants. The goal was to gather their feedback on how the designs could improve their workflows and identify any gaps.
Objectives
Navigation findability
Assess whether the proposed navigation changes made it easier for agents to find key information.
Centralized context
Understand how a more centralized organization system impacted their ability to maintain customer context.
Search efficiency
Evaluate whether expanded search capabilities helped agents locate accounts more efficiently.
Label clarity
Gather insights on label and terminology alignment to ensure clarity and ease of use.
My learnings.
The focus group validated several hypotheses while uncovering preferences that shaped the final design.
Navigation
Partially proven. Agents do want quick access to information about customers, but they preferred SubHub's secondary sticky navigation over a flatter style, which required more scrolling and clicking. The win wasn't flatter, it was persistent.
Organization
Validated. Providing as much context as possible, when appropriate, reduces unnecessary navigation, letting agents access critical information without switching pages.
Labeling
Validated. Consistent terminology across SubHub and customer-facing interfaces was essential for clarity. Aligning labels with agents' mental models reduced confusion and improved retrieval.
Search
Validated. Agents expressed a strong need for multi-attribute search, reinforcing the assumption that expanding criteria beyond email and account number would help them locate accounts more efficiently.
The final designs.
01
Navigation
I restructured navigation by surfacing key actions, keeping customer details accessible, and consolidating redundant pages.
02
Organization
Related information grouped together, with filters that help agents navigate while keeping context.
03
Search
We launched Advanced Search, multiple attributes, greater flexibility, and laid the groundwork for an AI-powered tool that shipped three months later.
04
Labeling
Standardized terminology across Account (customer-facing) and SubHub (agent platform), same feature, same name.
Customer view
"Bonus Subscriptions", unchanged for customers.
Agent view in SubHub
Now also "Bonus Subscriptions", labels aligned, confusion gone.
Advocating for the suggested changes.
One of the biggest challenges was convincing the product team to implement the suggested changes. Research and agent feedback strongly supported a more streamlined structure, but some of the proposed updates, consolidating pages, introducing a centralized organization system, and enhancing search, required significant engineering effort.
To drive alignment, I broke the work into near-term and long-term solutions, framing the changes in smaller, more manageable chunks. This helped the team focus on low-effort, high-impact improvements that could immediately enhance the agent experience while keeping larger structural changes on the roadmap.
08
The impact
Since implementing consistent labeling, persistent customer information, and a more robust search experience, we've seen measurable improvements.
Customer experience
−37%
Customer queue times, agents could retrieve information more efficiently, leading to faster resolutions and shorter wait times.
Agent efficiency
−4%
Average handle time, agents spent less time toggling between pages and more time addressing customer needs.
Long-term
AI Copilot
A more robust search infrastructure laid the foundation for an AI-powered tool that shipped three months later, further improving how agents access customer insights.
What endures.
The work shaped a more scalable, efficient, and effective support system, but the deeper outcome was a documented IA framework the rest of the product team now uses when adding new surfaces. The structure isn't done; it's maintainable, which is the actual win.
The biggest miss was waiting until the workshop phase to bring engineering leads into the IA proposal directly. They had the strongest intuition about which renames would carry implementation cost and which were free, earlier input would have saved a round of revisions. The biggest lesson: holding the line on three top-level modes when stakeholders pushed for four, then five. Research kept three honest. Tree-test data kept it shippable.