Optimizing payment forms for conversion with AI-driven credit card scanningJanuary 26, 2021
The payments ecosystem is vast and complex.
A single card payment a user makes as they’re cooking dinner goes through a large network of stakeholders before it’s processed and settled.
But despite being at the heart of this ecosystem, users take little account of any of this.
They don’t care about the stuff happening behind the scenes. All they want is a smooth, secure way to pay for their subscription, business trip, or a piece of clothing.
Meet BlinkCard: a credit card scanner built around users
A few weeks ago, we released BlinkCard v2, an AI-driven credit card scanning solution for mobile and web apps.
BlinkCard eliminates data entry in any mobile or web app by letting users scan the payment card they wish to make a purchase with.
There are two reasons credit card scanning should be a top-of-mind strategy for businesses looking to drive better experiences and more revenue.
One, as an online payment method, credit and debit cards are still massively popular, with digital wallets being their only major contender. (1)
This means hundreds of millions of people are still retyping their card information by hand.
Two, the majority of checkout flows out there are loaded with friction. In Europe, for example, 56% of e-commerce stores have at least three basic errors in their payment forms. (2)
A Stripe analysis shows businesses are literally making it hard for users to give them their money. Nine out of ten lost sales in Europe fail on the checkout page.
Exactly why so many users abandon their checkouts will depend on other factors, too, but poor UX is unquestionably among the main culprits.
Field validation helps users fill in their payment details on the first try. BlinkCard is the first SDK we’ve shipped with a built-in, customizable form.
When data entry issues precede multi-factor authentication and other mandatory checks, the entire process can drag on for way longer than it should, driving users away from making a purchase.
With credit card scanning, we can eliminate this friction from piling up early on in the flow, saving users both time and effort.
What’s hindering a widespread adoption of credit card scanning?
If the easiest way to pay with a card is to scan it, why are we still filling in our payment details by hand?
The answer to this question may be that it’s still not easy to scan your card in real life.
In fact, we found some of the world’s best banking, ecommerce, ridesharing and food delivery apps offer credit card scanning, but it’s far from ideal.
Using a mid-range Android smartphone, we took a few of them out for a spin, then compared their performance against BlinkCard.
Our test data consisted of three genuine credit cards with challenging designs; one with faded, printed characters and the other two with embossed characters that blend in with the background.
Here’s what we found:
- In a popular ridesharing app, we were repeatedly unable to scan the card with printed characters. The two cards with embossed characters made it through, but the app only returned the card number — we had to type in the expiry date and CVV ourselves.
- A peer-to-peer payment app with 1M+ downloads scanned all three cards but didn’t extract the cardholder name on either of them. Also, it could only recognize Mastercard and Visa cards. Interestingly, the app has scanning as its primary method of adding a new card to the account.
- Finally, an e-commerce app with more than 100 million downloads. This one gave us mixed results. Three out of four times, it was able to scan the card with printed characters on it, but it kept returning the wrong expiry date. It also extracted the card number and the expiry date from the other two cards (with one exception presented below), but we had to manually enter the cardholder name.
In comparison, BlinkCard’s machine learning-based engine consistently returned complete information on all three cards.
BlinkCard extracted the cardholder name, card number, expiry date, issuer, CVV, IBAN and an image of the scanned card, all in under a second.
Easy to use, difficult to produce
We don’t blame the apps (or the SDKs they use) for not supporting all card types and only returning partial data on some of them.
It’s very difficult to get credit card scanning right because you don’t know what to expect. Unlike, for example, government-issued identity documents, there’s few and often disregarded guidelines for the way payment cards should look.
Tricky by design: Payment cards come in many shapes and forms and some people choose to personalize them, which can introduce another level of complexity.
To address this drastic lack of consistency and make data extraction possible at scale, we’ve had to re-engineer the workflow behind BlinkCard.
In its upgrade, BlinkCard received an AI makeover similar to the one BlinkID had in 2019, when we made a switch to a machine learning-based data extraction.
We started to train our models to each deal with a specific task, such as detecting the card issuer and identifying pieces of information on the card. We then feed their combined outputs into a separate algorithm to fill and validate the results with parsed data.
This new pipeline allows BlinkCard to read cards issued by Visa, Mastercard and ten other networks in real time, on the device itself, keeping cardholder data protected at all times.
Aim to serve, not to swerve
Technology has made online payments a lot easier but there is still room for progress.
By simplifying checkout forms with AI-driven credit card scanning, we can lower abandonment rates and improve user experience, and that sounds like a win for everybody.
If you’d like to see how credit card scanning works in your iOS, Android or web app, you can try BlinkCard for free.
In case you need any help with your integration, let our support team help.