Masterclass with Tony Wimmer

[MUSIC PLAYING]

Hi, my name is Tony Wimmer. I'm head of data analytics for JP Morgan's Wholesale Payments Business Unit. And today, we'll be discussing payment performance optimization.

[MJUSIC PLAYING]

In the data analytics team at wholesale payments, we oversee all of data analytics for business lines of JP Morgan for merchant services, treasury services, trade, as well as commercial card. And they all relate to payments, and there's a tremendous amount of data in these business units. And the focus on helping our clients run their payment operations better and to grow their business. So they have about 150 data scientist, machine learning experts, and data engineers that I work with, where we create software and insights that help our clients make better decisions.

[MJUSIC PLAYING]

When we think about a data driven approach, there are two key things that I always focus on. The first one is transparency and the second one is a framework for systematic opportunity identification. So on the transparency side, you think about what optimization means is typically you have two cohorts of transactions and you're trying which cohort performs better. The challenge I often see is that instead of creating an apples to apples comparison, there's a lot of apples to oranges comparisons that lead to not very meaningful results.

So what we typically work with merchants on doing is those two cohorts of transactions we normalize them for very important factors such as, is it a first transaction, or is it a retry transaction, or about normalize it for the tenure of a merchant's customer base. Because for example, if you have one cohort of transactions, where you may have relatively new customers in it, and a second cohort of transactions, where you have a lot of existing customers in it that's maybe shopped at that merchant 10, 12, 14 15 times, the new customers will tend to have a lower approval rate. Because from a credit card issuer side, if you as a customer has shopped only once or never at a specific merchant, you look riskier or the transactions looks much riskier. So that's transparency.

And on the framework for systematic opportunity identification, we at JP Morgan, we have a framework where we systematically scan nine levers for payment optimization, six on the authorization side, and three on the cost of payment side. And the key is that you have a framework and you systematically keep going through it, and you periodically continue looking at it because the payment ecosystem constantly changes. So at JP Morgan, we have about 40 data scientists that are just dedicated to that that create algorithms, that look for opportunities, and keep updating them.

[MJUSIC PLAYING]

It is a lot of very valuable information about customer behavior in payments data. Take JP Morgan, for example, we are also very large credit card issuer. And we understand in great detail where customers shop, how far they are traveling to a shop, where they shop, when they shop. And we use this information in de-identified and aggregated way, because our highest priority is customer privacy. But in de-identified and aggregated way, a merchant can understand how am I doing in a specific customer segment relative to my peers, where am I doing better than my peers, and where I am in doing worse than my peers.

And if you track that over time, so against your peers and over time, payment data can be what they call it your cannery in the coal mine. So if you see a trend in a very important segment that is accelerating, you'd rather want to know that sooner rather than later to take corrective action. A second use case of this type of data is its geographic insights of your customer base. So you can use payment data to get a better understanding for how far our customers traveling to your stores relative to where they live that has helped a lot of merchants make better location decisions.

[MJUSIC PLAYING]

I would say COVID has actually amplified the valued area than payments data. So the effects of COVID and social distancing became very quickly evident in payments data, because consumers changed the way they shopped. So by understanding these type of trends, a lot of merchants looked at a type of data to guide them through the economic recovery. And on the payment operations side, what we also noticed is the rapid shift from physical card present transactions to online card not present transactions surprised a lot of merchants. They never had to deal with, how do I optimize authorization rates, how to optimize disputes. So it made clear that a data driven approach to how they organize your data and make fact based decisions in times of uncertainty was actually really, really helpful.

[MJUSIC PLAYING]

There's a lot of data in payments, but the challenge is data alone isn't that useful. You have to translate it into actionable insights. So and at JP Morgan, where we spend most of our time and investments is, how do you take the headache out of that process for our clients? How do you collect the data, clean it, organize it, and then use modern technology to extract signals so your accounts payable professionals can make better decisions? And we talked about a couple of these decisions, right? We talked about how do you optimize authorization rates, how do you optimize cost of payments, how you grow your business by understanding your customer base better.

And an additional area where at JP Morgan, we focused a lot of our energy on is, how do you help your accounts payable in conceivable professionals better forecast your cash flows? So when you talk to our clients in the treasury departments, one of the biggest problems is actually accuracy in cash flow misforecasting, how much money is coming in and how much money is coming out. Because it's typically operating on imperfect data. Data is never clean. They operate these spreadsheets.

So what we have done is really invest in new technology that now helps our clients. The push of a button understand what has happened in the past and then runs at the click of a button machine learning models that help merchants create a forecast based on the best possible algorithms to understand what may happen in the future. And then that forecast professionals can start using it as a very solid foundation to then adjusts for their needs and make better decisions on cash flow forecasting.

[MJUSIC PLAYING]