Case study · Information architecture
KM Modernization
From content repository to decision support: a legacy knowledge base redesigned so frontline agents reach the right answer in seconds. Planned end to end and proven in a pilot.
Before · organized by team
Team A
Update Profile
Book Travel
Report Fraud
Team B
Book Travel
Modify or Cancel Bookings
Issue Goodwill Miles
Team C
Report Fraud
Investigate & Resolve Fraud
Handle Partner Escalations
After · organized by task, filtered by access
Manage Account & Balance
Update Profile
Report Fraud
Investigate & Resolve Fraud
Redeem & Manage Rewards
Book Travel
Modify or Cancel Bookings
Resolve Program Issues
Issue Goodwill Miles
Handle Partner Escalations
01 · The problem
Agents were guessing
A high-volume loyalty contact centre relied on a knowledge base built on a legacy SharePoint 2010 platform, organized by team name.
Core frontline agents mostly worked in one section, but the domain-specific and expert teams, who handle specialized processes, exceptions and escalations, worked across several. They had to remember which team owned each process, open that team's section, and back out to try another when they guessed wrong, all mid-call while the customer waited.
Research with frontline agents surfaced six themes:
02 · The approach
Restructure, not rewrite
Turned a migration into a redesign
A move to a new platform was already being discussed. Rather than carry the old structure across as it was, I pitched using the moment to rebuild it. The content was not the problem. Its structure was.
-
Replaced team-based navigation with a permission-based model
Every agent sees the same main navigation, organized by task. What appears under each heading depends on their access level, and the levels build on each other:
- General access (core frontline teams) sees the general pages.
- Specialized access (domain-specific teams) sees the general pages plus specialized ones.
- Advanced access (expert teams handling exceptions and escalations) sees everything above plus their own processes, which no other team sees.
Agents no longer need to remember which team owns a process: everything their role covers is in one place. And because every agent sees the same main navigation, a frontline agent moving into a specialized or expert role keeps working with labels they already know. Only the pages underneath grow.
Designed a task-first IA
Research showed agents think in terms of the action they need to take, not the team that owns it, so most labels are written as verbs: Update Profile, Book Travel, Resolve Travel Exceptions. I replaced deep hierarchies with a wide, shallow structure, so major tasks are reachable in three clicks or fewer, and front-loaded every label with its key word so agents can scan a list at a glance. Both changes reduce cognitive load and help agents answer faster mid-call.
Built a metadata and content model
One topic per page, consistent templates and shared labels, plus an owner and a review date for every page.
Piloted where it was hardest
The new structure went first to Travel Support, the back-office team that handled invoicing, so every new and changed booking passed through them. If it worked under that pressure, it would work for everyone.
03 · The outcome
Proven in pilot, planned for scale
4.3 → 2
clicks to reach a topic
~50 sec
saved per lookup
~$22,500
saved a year by the 16-agent pilot team, roughly 1,100 hours
$100K+
projected yearly savings at full scale, if each lookup saved just one click
In the Travel Support pilot, navigation guesswork disappeared, and leaders asked for the same landing-page pattern in every other section. The full rollout was planned and costed: across all contact-centre teams, saving just one click per lookup would save more than $100K a year. The pilot saved more than two clicks per lookup, so that projection is conservative. A company acquisition ended the program before it could be implemented.
Keep exploring
Back to all work