Capital One
Reinforcing in-app support with search and help center
As a lead designer on the Capital One app team, I led the design of in-app support features from concept to implementation. I collaborated with UX researchers to gather insights on user habits, partnered with product managers to ensure designs balance both business and user goals, and worked with engineers on testing.
Role
-
Product design
-
User Research
-
Prototyping & testing
IMPACT
-
~$1M projected annual savings by reducing call center volume.
-
84% increase in Help Center engagement.
Problem
The only in-app support we had was a lousy Chatbot.
In-app support like help center and chatbot are ubiquitous in all mobile apps. But back in 2023, the Capital One app only had a chatbot. The chatbot couldn’t consistently interpret our users’ intentions, and was only able to answer some general questions. A JD Power survey revealed that only a third of our customers were satisfied with the guidance provided in the app. This has resulted in a huge amount of calls to customer support, increasing costs. So, a project was commenced to reinforce the support functionalities in our app.
How might we expand our in-app support so that users can find information they need without having to call customer support?
User Research
A diary study to lay the foundation.
I worked with UX researchers to conduct a diary study to understand on how customers engage with in-app support in various finance apps. 30 participants took part in the study via an online platform. We asked participants to make an entry about their experience whenever they were trying to troubleshoot or find information in a finance app. Participants were asked to make at least 4 entries in a 4-week period.
In general, users prefer not to call customer service.
Approaching a problem, users often have to decide between the level of effort needed to find the solution on their own, versus the potential time spent waiting for customer service.
Different users will choose different support features based on factors such as urgency, uniqueness of the situation, and their familiarity with both the problem and the app.
Top-level insights
Quotes about features
Help Center
"I first go to the help section because I find a lot of the apps provide you with all the facts, FAQs. Hopefully, somebody has had the same problem I had and we can get a fix."
CS, 57
Chatbot
"The chatbot was able to give me the exact answer l was looking for that expired cards would be sent out automatically and I should expect it soon."
KS, 34
Search
"So then I went to the search bar. That's just top right corner. Hit the search button, type in, "find checking account number" and voila.
BH, 29
Solution 1
Focus on Search in this phase.
We decided to prioritize a global search feature over upgrading the chatbot. Improving the chatbot would require a backend overhaul and integration of LLM to meet rising user expectations. Meanwhile, Search is something our engineers can quickly build and can connect users to information that our chatbot can't. There are also different types of content already available and searchable across Capital One and they are best presented in their existing state, rather than within a chat format.
Intelligent suggestions before you type
On the landing screen, we want to anticipate what a particular user might need at that moment based on time, location, account status, etc. For example, when it's 2 days before the payment due date, a suggested search might be “Pay bill on XXX card.”
Based on our chatbot data, we learned that 74% of our users are looking for a particular feature in the app. So it makes sense to place "Quick Actions" — jump links to key features — at the top.
Autocomplete that prioritizes Quick actions
On the input screen, autocomplete is a ubiquitous feature most users expect. To start, we decided to:
-
Show maximum 7 suggestions — we don’t want the list to scroll
-
"Quick actions" will show on top if there's a match
-
"Suggested searches” is a catch-all of other content types
-
In future tests, we can consider "Help topics" and "Account info" as their own sections.
Use case 1:
User doesn't exactly know what he's looking for.
Let's say the user vaguely remembers a Walmart-related benefit. He types in “Walmart”. Does he want to pay his Walmart credit card bill? Does he want to redeem his Walmart rewards? Or does he mean Walmart Pay?
When a user searches for something ambiguous, it’s faster for him to pick from a list of options, rather than talk to a bot to narrow down his intent.
This example also shows how different content types can be pulled up via Search. Depending on the search, a user can access direct links to features, help articles, information from our website, account information, promotion and offers all from one page.
Use case 2:
Going from global search to local search.
This is an example where a global search seamlessly transitions to local search. Let's say the user just got her latest statement, and the statement balance is way above her expectation. To investigate, she searches for the largest transactions last month.
From the search results, she can jump to the Transactions feature where she can refine her search using filters and sorting. As she uncovers a transaction she doesn't recognize, she can report it right there.
SOLUTION 2
A quick expansion on Help Center leads to a quick win.
Back in 2023, the Help section under Profile only had links to the chatbot and customer support numbers. The chatbot was lousy, and most users had to call customer support for even the smallest issue. From our diary study, we learned that some users prefer browsing a Help Center to find answers. However, a robust help center needs a lot of well-organized content, and it would require many teams across Capital One to align priorities and collaborate.
With our efforts focused on Search, we had to quickly expand the Help Center on the side. Based on user queries we collected from the chatbot, we had a list of most commonly asked questions. So for MVP, we added a section with answers to these issues.
Next, there are already help articles on the Capital One website. So, to start, we created another section and added links to these articles based on product type. However, most of these articles were not written for in-app experience. In the long run, our plan is to work with various LOBs to update the content and bring articles in to a self-contained Help Center.
Before
After
IMPACT
Meaningful improvement with measurable impact.
While I was on the team for just 15 months, the improvements we made to Capital One’s in-app support are already showing business impact. The expanded Help Center saw an 84% increase in engagement within the first month of launch.
We have data on the top 20 reasons our users call customer support within 2 hours of logging in. 90% of these can be resolved through information already available in the app or on our website, and our new Search should enable most users to find answers on their own. Based on monthly call volume, average call length, and estimated call cost per minute, we project at least ~$1M in annual savings by reducing call center demand.
Next Project
Capital One Website
