Saved $6.2 Million in Cost

For a Prominent Banking Corporation
‍

Impact
‍

Cut call handling time by 2 minutes per agent. This change saved $6.2 million in costs for the quarter. Improved efficiency and usability for customer service agents via a natural categorization structure. These insights created the new IA.This change made the customer service team more efficient.

Reduced Call Time

2 Minute Reduction

Learn How

5 Unique Groups Discovered

R and Optimal Workshop

See Research

Actionable Insights

UI Overhaul

View Results

Overview and Abstract
‍

Context: Customer service agents use Fare Rules, a feature within a larger banking application, to determine travel costs.
‍
Problem: The text interface made it tough to find information. It slowed down agents and led to more human errors since agents had to type commands manually. This cost the bank millions in operational costs annually.
‍
Role: As the Principal UX Researcher, I chose methods, gathered data, and analyzed it. I worked with product managers, help desk agents, and the design team.
‍
Skills Applied: Card sorting, information architecture, R programming, statistical visualization, enterprise UX research, quantifiable ROI measurement, dendrogram analysis, user personas.

Client Presentation Research Plan

Project Runway
‍‍

Opportunity: I talked with the senior product manager and clients. We found that redesigning Fare Rules is a big opportunity.
‍
Planning: Six months prior, I created a research roadmap with product managers and stakeholders, focusing on product and design priorities. This included research for Fare Rules later on.
‍
Challenge: Redesigning the text-based interface needed a grasp of how agents think about Fare Rules.
‍
Strategy: I picked card sorting to uncover how daily users naturally categorize items. It was the best way to find the best information layout for fair rules.

Methods and Process
‍

Schedule:
This was Tallwave and the client's first card sort, so I provided two months for education which occurred during group and 1-on-1 client meetings.
Other methods, like usability tests, can be done faster.
This approach ensured thorough data collection and analysis. I finished data gathering and presentation creation in 10 days.
I shared the findings with the client and the UX design team. We discussed them over the following months.
Method:

I talked to stakeholders to decide which cards participants would sort. Fare Rules is a complex feature. I also confirmed a research plan before starting the research.
The customer service manager scheduled agents with me. I used Optimal Workshop to gather data.
I clustered to see how participants grouped cards. This helped identify pages and sections for the new UI.

Highlighted (in yellow) are similar categories chosen for each group and the total number of groups.
This group concerns ticket types.

Customer service agents grouped categories about discounts and billing into the second group.

Customer service agents grouped categories that are closely related to ticket policy.

The group focuses on bookings and flights. They found that agents prefer to search by flight number and trip number.

Optimal Workshop analysis cut 51 fragmented categories down to 21, defining four major groups. This suggested the new design should have four pages or sections.

Further Analysis in R
‍

After looking at the Optimal Workshop findings, I saw I needed a deeper R analysis. This analysis took place after the report was released because:

The client needed clearer proof to approve big interface changes. These changes would affect thousands of agents.
The design team asked for specific guidance on which elements to group. They also wanted to know which ones had flexibility.
We needed to look at the statistics to check if the patterns were real mental models and were not random.
Optimal Workshop couldn't show how strong the agreement was between cards and groups. This made it hard to rank the most important groupings.

I wrote a program in R to analyze the data.
‍
I anonymized the data here to protect my client and the users.

This diagram helps to decide how to group the cards:

The team and I would use the groups to decide what goes on each page. The text in each group provides the relevant information for that page.
I identified five distinct categories of information through dendrogram visualization. This supports five pages, not four like the Optimal Workshop analysis suggested.
The dendrogram displayed groups but didn't measure user agreement on cards within each group.
From the dendrogram alone, I could not determine how necessary it was for some specific information to be on a page or page section.

I used the heatmap to determine the degree of agreement between users on whether or not a data field should be on a page:

The blue text image shows part of the program that created the dendrogram and heat map.

Outcomes
‍

The heatmap and dendrogram provided insights not visible in the Optimal Workshop analysis:

The analysis showed that there could be either four or five groups. The final decision was to use five groups, which was arrived at after speaking to stakeholders (UX designers and the Senior Product Manager).

We split ticket types from routing info. This fixed the confusion that slowed down searches.

We placed fare calculators with pricing rules where agents expected to see them. This cuts down on wasted search time.

We organized commands by how often people use them. The most important tools are right at the front.

In conclusion, we created categories matching how agents think about fare information, making the process more intuitive.