RBC to use AI to understand customer intent and recommend products

rbc-slideshow-size
Kevin Van Paassen/Bloomberg

The notion that a consumer would seek retail product recommendations from their bank might have seemed far-fetched a few years ago, but banks now want to play a central role in helping consumers make purchases initiated directly from the apps and platforms banks own.

Retail media networks are brand-owned advertising platforms that allow partner merchants to reach the brand's customers through ads or targeted offers. Royal Bank of Canada, which operates the Avion rewards program, is the latest major financial institution to expand efforts in this area, with plans to roll out prescriptive product recommendations based on customer intent data next year, according to Avery Miller, vice president of shopping and Avion digital products at RBC. 

Prescriptive recommendations are based on an analysis of customer intent, like searches or browsing and shopping history, and how customers engage with apps, as compared to predictive offer recommendations the bank currently pushes out that are based only on transaction data. 

Predictive recommendations are based on past transactions, whereas prescriptive offers are based on a combination of shopping intent data, along with behavioral characteristics, including how customers might engage with RBC's apps and digital banking platforms. A predictive model "is still trying to use past behavior to put something in front of you," Miller said. "What I'm looking forward to is something that I'm calling more of a prescriptive set of behaviors, where instead of taking transactional data, I'm taking intent data."

Intent data is gathered from a browser extension called Shop Plus, which finds cash-back deals and discounts for customers as they search for products online. Some customers can also earn more Avion points through these offers. The bank's research arm Borealis AI is currently running experiments on the use of machine learning algorithms that analyze customer intent data and behavioral data to push tailored, prescriptive product discount and points redemption offers, Miller said.

"When I can collect browsing data and website hits and searches … I have an idea about your intent, and now I can prescribe a behavior," whereas RBC's models currently push partner offers that are only based on a backward look at transaction data, he said. 

For example, the bank might send an offer that may be redeemed at a different retailer in the same product category. A prescriptive offer, by contrast, may suggest products from entirely different product categories based on intent. A prescriptive offer also takes into account how a customer has interacted with offers in the past.

"What those models today don't take into consideration is whether or not you've seen those offers before and didn't interact with them, and therefore aren't interested in them now," Miller said. "You might be interested in something that we haven't even predicted for you yet, but your intent shows that you're probably going to make that purchase decision."

A Toronto-based customer, for example, might be looking at ski slope conditions in Whistler, British Columbia. As a result, they may be interested in flight deals, or they may be open to being served up deals at a ski apparel retailer.

"I want to be using that intent data, that shopping data, the pre-transaction stuff, as a way to prescribe a set of things that I hope a user will do," Miller said.

The bottom line

Acting as a retail network for partner brands serves several objectives for a bank: It encourages card transactions; it encourages points redemption; and builds longer-term brand loyalty. 

RBC's Shop Plus browser extension, according to the bank, was developed in partnership with fintech Wildfire Systems, which works with several banks and fintech firms, including Citi, LendingClub and Acorns. Banks and fintechs the company works with can take a cut of merchant-funded cash-back offers, which can be 10% of the price of an item or higher, Wildfire said in February. RBC, however, said it does not take a cut of the transactions generated through the Shop Plus platform.  

Other large banks and financial firms have jumped on the creation of retail media networks in an effort to diversify revenue streams beyond card-swipe fees. Last month, JPMorgan Chase launched Chase Media Solutions, a platform that will help the bank develop more card loyalty programs funded directly by merchants instead of card network transaction fees. 

Chase said it will connect the bank's 80 million customers with deals from tens of millions of shopping, dining and travel merchants. Because its platform provides clear attribution, marketers will only pay the bank when a specific offer triggers a purchase, the bank said. Meanwhile, Citi announced earlier this year that it's partnering with Wildfire Systems on a browser extension called Citi Shop that searches for coupons on eligible merchant checkout pages and suggests applicable codes to apply at checkout.

Banks are "seizing on the opportunity to generate revenue from troves of proprietary customer data, and really being able to turn that into an asset for advertisers," said Tiffani Montez, principal analyst at Emarketer. "Part of that has to do with cookie deprecation — we know that at some point that cookies are not going to necessarily be an effective way to target people — and then we also know there's an attribution issue," she added, noting that it's challenging to know who looked at an ad and tracked it all the way to payment.

Global retail media ad spending, estimated to reach $137 billion this year, is expected to rise to $280 billion by 2028, according to Emarketer.

An avenue for generative AI?

Asked whether RBC is using generative AI as part of this effort, Miller said it's not being used to analyze customer data, but the company is exploring the possibility of using it to generate text for offer copy, which could be customized depending on the type of customer. He gave no timeline as to when this might be adopted by the bank.

"We have looked at using generative AI as a way to quickly produce a lot of different versions of an offer tile," Miller said. Banks' tech stacks aren't always optimized for rolling out thousands of different versions of an offer — someone would have to type them out, he added.

Looking ahead, banks have to think about relationships with customers that go beyond transactions, and retail partnerships are one way to move deeper into the product purchase road map.

"There are many players in the payments and banking ecosystem that are exceptional at the rest of the user journey, whether that's search and discovery, inspiration, consideration, and banks are not playing in that space in a meaningful way," said Miller. "In order to maintain our position with our customers … we need to be better at the rest of the consumer journey."

For reprint and licensing requests for this article, click here.
Artificial intelligence Technology
MORE FROM AMERICAN BANKER