Banks model site search upgrades on Google

Register now

Google long ago raised the bar for search engines, and only now are banks starting to catch on.

Instead of keying in words and sorting through a hodgepodge of results, customers want query boxes that correct their spelling and intuitively complete their thoughts — at minimum. Even better if the search tool can tease out the intent behind a question the consumer poses in natural language, and deliver answers within the initial search results rather than requiring extra clicks.

“When Google implements something on Google.com, consumers expect the same kind of functionality everywhere,” said Kamran Khan, global lead for search and content analytics at Accenture and managing director in Accenture’s Applied Intelligence practice. “That’s why you see this move towards providing answers and not just results.”

Although a poor search engine isn’t a deal breaker for customers, “the harm is perhaps if someone can’t find the product or adviser they were looking for, and goes somewhere else to buy that same product or find a financial adviser from a different firm,” said Khan. “It’s a question of converting consumers and sales.”

Rather than just focus on website functionality for customers who have already logged in, banks are realizing that they need to improve the search tools on their website to sell products, said Khan. And with more tasks becoming digital, the ability to easily navigate a website is vital.

BBVA USA in Birmingham, Ala., uses Yext Answers, a site search product that understands natural language and provides answers. The experience is built on a company’s knowledge graph, or repository of information that helps the program understand queries and link facts to return a contextual answer. Meanwhile, U.S. Bank in Minneapolis is taking natural language search a step further with the recent launch of a virtual assistant embedded in its mobile app.

Questions and prompts

A little over a year ago, the search experience on BBVAUSA.com was heavily fragmented. The website for the $102 billion-asset bank had separate search engines or lookup tools for website content, branch and ATM locations, and mortgage and business banking officers.

“This was not an intuitive experience for our customers, so we put a lot of effort into developing a unified search where customers could easily find whatever information they needed,” Steve Creel, senior vice president and director of online and mobile websites, said by email.

Steve Creel, senior vice president and director of online and mobile websites, BBVA USA
Users can pose questions to the BBVA site's search engine as they would with Google, said Steve Creel, senior vice president and director of online and mobile websites at BBVA USA. Like Google, the answers are sometimes embedded in the page of search results.

It took two months to revamp the system with Yext (BBVA USA was a beta user of Yext Answers), and the new tool launched in October of 2019. Now, the entire search experience is pulled together onto one bar and uses natural language processing, or NLP, to interpret questions like, “Where is a branch near the Houston Galleria?”

“This eliminates the effort of trying to think of the right keywords,” Creel said. “People are also beginning to expect natural language search as they get comfortable using it with popular services like Google, Siri and Alexa.”

Search results are organized into sections and tabs, including Products, FAQs, Locations and Officers. For example, searching for a branch near the Houston Galleria mall summons a map marked with relevant locations and blurbs listing the hours, contact information and links to get directions. Scrolling down further leads to a list of mortgage banking officers in the area, with photos and contact details.

Visitors' questions guide which categories appear at the top of the results page. For example, asking, “Is my business eligible for PPP?” prioritizes frequently asked questions about Paycheck Protection Program eligibility.

BBVA also takes its cues from customer questions to enhance website content.

After noticing queries about how to find a routing number, Creel and his team created a webpage to elaborate on the topic as well as a Routing Number Lookup tool. As entirely new questions rolled into the contact center and through web searches during the pandemic, surrounding stimulus checks, loan payment assistance and more, “we could write and publish these new FAQs as fast as the new questions would arrive,” Creel said in a webinar that discussed BBVA USA’s collaborations with Adobe and Yext.

Creel and his team update their knowledge base daily to ensure branch information and frequently asked questions are current. They also have been studying voice technology.

“Banks are still in the early stages of implementing NLP on their sites, but as tools like Yext come along, we will see the evolution of what you call questions and answers on websites,” said Khan.

From natural language to actual language

U.S. Bank’s Smart Assistant is built into its mobile app and lets customers bank by speaking or texting requests. But the microphone button is dominant.

Voice search "is a twist on natural language processing, where your voice is converted to text and being interpreted using NLP techniques to match you with the right answer,” said Khan. “It will be a while but voice search is the way of the future, driven by companies like Google and Amazon.”

Besides transferring money, inquiring about upcoming bills or completing a variety of other tasks, customers can use Smart Assistant to efficiently locate any of the 300 features in the app.

“The holistic adoption of voice across the board has been dramatically increasing, especially in the context of search,” said Ankit Bhatt, senior vice president and chief digital officer for consumer services at U.S. Bank.

One plus of searching by voice is the customer can skip steps to find the information they need. Customers searching for their routing number with Smart Assistant, for example, can instantly locate that information rather than landing on a page that lays out their next steps. And instead of guessing at the right keywords or search terms, customers can ask the question in their own words.

Smart Assistant uses natural language processing technology from Clinc to understand customers’ speech.

When building the virtual assistant, U.S. Bank leaned on its team of accessibility consultants. Those who are blind, in particular, “have decades of experience using voice technology,” said Bhatt. The result, he said, is “a fantastic experience not only for people with disabilities but a better product for all of our customers.”

For reprint and licensing requests for this article, click here.
Digital banking Websites Virtual assistants Google Natural language processing Digital Banking 2020 BBVA U.S. Bank
MORE FROM AMERICAN BANKER