In a world where mail is electronic, newspapers are digital and banking can be done on your phone, the continued existence of paper receipts is a bit of an anachronism. They accumulate in forgotten shoeboxes, get discarded accidentally and go through the wash in the back pockets of jeans, losing useful information in the process.
While the old-world receipt may not be disappearing any time soon, a Toronto fintech is making it easier than ever to bring the information from a receipt into the digital world.
Founded in 2013, Sensibill Inc.’s platform uses machine learning, or artificial intelligence (AI), to identify and extract information from receipts. It then reorganizes it for easy use in any number of applications, including expense management, filing warranties and taxes, or simply getting a better picture of one’s financial health.
“What we sought to do is … fetch all of those receipts that you’re getting in the form of a paper receipt, e-mail receipt and so forth and bring them all to one safe place,” says Corey Gross, the company’s co-founder and chief executive officer.
Mr. Gross points to other benefits of easily digitized receipts for consumers, such as fraud detection, easy long-term (non-shoebox) storage, and the enormous time savings involved in not having to pore through piles of paper to extract information.
Sensibill’s algorithm also takes a step beyond simply gathering and reorganizing data, matching it up with products and vendors indicated by the consumer’s spending preferences and producing useful information for the user.
“So now instead of seeing this kind of garbled description of what you bought at Loblaws with a bunch of numbers next to it, we can actually tell you, ‘oh, that’s a can of Green Giant peas’ or, ‘hey, you’re buying Robin Hood flour three times a month and maybe you should find an opportunity to save on a cheaper brand,’ ” Mr. Gross says. “We’re actually enriching that data with AI and intelligence.”
Many financial players see the value proposition: Sensibill’s list of financial clients and partners has grown over the past few years to more than 150 – including Bank of Nova Scotia, Royal Bank of Scotland and JP Morgan Chase, to name a few – that offer its receipt-capture and analysis tools through their own banking apps.
The platform is currently available to some 70 million users, Mr. Gross says, who can upload paper receipts by taking a picture of them or have them directly e-mailed to the Sensibill-powered banking app from an online vendor.
But the company’s latest offering – and the one where Gross sees the most potential for helping transform financial services – aims to help financial institutions use the technology and the intelligence gathered through the app to forge closer ties with clients and integrate their product offerings with individual needs.
“We take all of this enriched data, and we ultimately create these profiles: product profiles, merchant profiles, customer profiles,” he says.
With this information scraped by Sensibill’s AI algorithm, banking clients can tailor product offerings to their customers. For instance, a banking customer whose activity indicates that they’re an expecting parent could be offered information on setting up a registered education savings plan.
“Being able to educate the customer based on their preferences and their lifestyle is exactly how we serve these institutions in the customers they have,” Mr. Gross says.
Adithya Sreekumar, an investor at AI-focused venture fund Radical Ventures, says providers such as Sensibill are helping the banks play catch-up to the customer experience now common in the consumer tech space.
“As anyone who has glanced at their Netflix recommendations knows, customers leave behind signals in their digital journeys,” he says. “When captured and understood, these signals can be used to create superior personalized experiences.”
Radical is one of a group of venture firms that have poured money into Sensibill as its technology has gained traction. It led a US$31.5-million financing of the company in 2019.
While some recent fintech success stories have produced platforms that allow customers to take more control of their finances at a lower cost than typically offered by the banks, Mr. Gross sees a new fintech frontier in the use of AI to produce a differentiated and more human bank customer experience.
Indeed, it’s hard to call Sensibill a disrupter when its bank partnerships mark it more as a collaborator. What has changed recently, according to Mr. Gross, is the banks have realized they were falling behind in how they engage with customers.
Jason Pereira, partner and senior financial planner at Woodgate Financial Inc., says a key reason the banks haven’t taken it upon themselves to streamline payments information such as receipts is a perception that the cost of doing so wouldn’t be worth it.
“Think about what it would take to get rid of paper receipts,” he says. “The billing system would have to connect to the point-of-sale device … and then that would have to connect to you somehow.”
“It’s still a very broken segmented market, so paper is the point of least friction,” he says.
Looking ahead, Mr. Gross sees the potential for further bank-related applications of the AI technology in breaking down the traditional business silos at big banks to give each line of business a complete view of the profile of the customer it can act upon.
“Our goal is to start connecting the dots and truly be a holistic customer data platform for financial services,” he says. “We think we can win that because we already have that basic trust from the biggest banks in the world and we have a whole whack of data that tells us that we can add differentiated value.”
How to teach a machine to read a receipt
Anyone who’s used a good banking app knows that you can teach a computer to read a cheque, as a simple photo uploads the key information and the money magically appears in your account.
But doing the same with a receipt is a whole different matter, says Sensibill’s Mr. Gross.
“What you realize very quickly is all cheques are kind of formatted the same way,” he says. “We’re talking about structured document to highly unstructured document in the form of a receipt.”
Without any established standard for receipts, their layout can vary greatly. Start with the fact that there are millions of merchants around the world with information written in different languages and financial information in different currencies under different tax rules.
Beyond that, there are varying merchant categories, such as grocery or coffee shops, some of them chains, and other independents that may use basic receipts with very little information at all. Even receipts from a chain or franchise can present problems.
“Every Tim Hortons that you go to in Toronto could have a different format of receipt,” Mr. Gross says.
“So, you’re looking at receipts at the merchant level, millions; at the category level, hundreds; and then within a specific merchant brand, you may have hundreds of variations of the same merchant receipt,” he says.
So how did Sensibill teach a computer to sort through all this to the point it can reliably read a scanned receipt? They had to get their hands on different types of receipts – a lot of them.
“We’ve collected receipts from over 230,000 merchants worldwide in 32 countries and currencies. We’ve extracted six million unique product SKUs [barcodes], of which more than 250,000 are regularly seen by our platform,” Mr. Gross explains. “We’ve created more than 6,000 individual product categories.”
He says the receipt database has been built through its proprietary data over the years, which means that its receipt-reading technology has improved as its customer penetration has grown and it has received more examples to work with. Sensibill doesn’t buy receipt data from third parties as such purchases often have conditions attached about how the data can be used.
“So, how do you teach a machine how to read receipts? Data. And that’s what we do really, really well,” Mr. Gross says.