Plaid – A Modern API for Card Data
Our work in payment data cuts across many areas. Who has what data? How is it best monetized? How can it be put to use for better decision making? How can it be augmented and enhanced.
It’s in this last area of “enhancements” that we’ve become fascinated with a new startup company called Plaid. They are making the first serious attempt we’ve seen in a while that rethinks the quality of the transaction data.
Every consumer knows how confusing it can be to open up a statement or view their recent activity and find a bunch of transactions for purchases that are not obvious at first glance or maybe not even obvious after a detailed investigation. What is this? Is this a gas station? Was that me or somebody else using my card? And that’s with human reasoning. It gets a lot worse when you try to couple that level of transaction data quality with analytics and third party apps.
The root cause is the minimal amount of data that is available to describe the merchant in the card systems. The merchant name is often truncated to fit the fixed formats of card messaging. Magnolia Home Theater becomes MHT. Then there are merchant category codes (MCCs) that are defined by the card networks but actually set by the merchant’s acquirer. The codes standardize the category descriptions, but it’s the acquirers that force fit each merchant into exactly one of the categories that’s supposed to define the merchant’s primary line of business.
Think about that lovely meal you had eating out last night. From the perspective of your card issuer, you were spending money at a truncated merchant name located in a zip code that has been assigned to one of two categories:
5812 – Eating Places, Restaurants
5814 – Fast Food Restaurants
Your card issuer also knows how much you spent with that merchant. From this minimal information the issuer (or their analytics provider) can figure out how many times a week you eat out, how much you spend, how often you frequent this merchant, etc. Service providers like Yodlee can provide third parties with access to this data as long as the card holder provides permission and their online banking credentials. You see this Yodlee service used all the time by companies like Personal Capital, BillGuard, and Outright.
Plaid’s big idea is that it should be possible to take this bare bones transaction data and enhance it so that it has more relevance and meaning for third party apps. Essentially, to transform the data to be more usable in today’s market.
The basic model is the same as Yodlee’s in that third party app developers utilize the Plaid API to access statement data once the consumer provides the app with permission and login credentials to access their account. But instead of simply raw statement data, Plaid returns a “cleaned up” merchant name, a hierarchical category mapping, and assorted meta data about the merchant.
The category mapping replaces the notion of flat merchant category codes. Instead of “5812” the app sees the merchant category as a three-tiered structure of “Food & Drink -> Restaurants -> Spanish” for example. In addition to the Plaid categories, the API calls also return the FourSquare category mappings, the AmEx category mappings, and the factual MCC data.
The meta data is perhaps the most interesting enhancement. Here Plaid will make a best effort attempt to provide the merchant contact information (phone number and website) and location data (both street address and geolocation). Because this is not always possible for every purchase, they also provide the app with a meta data confidence score.
It will be interesting to see how shopping apps and offer targeting services make use of the Plaid service. Enhancing transaction data for contemporary use in today’s market seems to make a lot of sense. But it wouldn’t be easy.
Plaid’s service is currently restricted for use with accounts held at American Express, Bank of America, Chase, Citi, US Bank, USAA, and Wells Fargo. This mostly likely indicates that Plaid has gone back to square one with a bank-by-bank scraping approach to get the raw statement data. These are certainly the major credit card issuers, but we live in a country with 14,000 plus financial institutions — each of which issues debit cards.
That said, Plaid’s focus on developer tools and needs should help accelerate the pace of innovation with how payment data is used in the industry.
If you are interested in these issues, please join Scott Loftesness and myself at our next Data in Payments Workshop being held July 23nd in Mountain View, CA. We’ll be exploring who intrinsically knows what, how that knowledge is monetized, and the various techniques that are used to gain access to payment data.
This post was written by Glenbrook’s Russ Jones.