Subscribe
Share
Share
Embed
Our flagship series will propel you to the forefront of the global ecommerce revolution. From analyses of breaking current events to the intricacies of navigating cross-border sales and regulations, Business over Borders entertains and informs any audience who wants to learn more about how international ecommerce works.
Welcome back to another episode of Business Over Borders. I'm your host, Leo Tucker, and I'm joined by Lohrasp Seify. Welcome to the show again. Nice to have you back.
Lohrasp Seify:Nice to be here, Leo.
Leo Tucker:I think the last time we chatted, we talked about all things data and mostly how routing plays plays into that. You know, connecting the dots between where a customer is, where the acquiring banks are, BIN routing, things like that. I want to dig a little bit deeper into data. So let's let's take a store, for example, someone who's kind of already expanded into the markets they want to work in. How can they best, you know, without accessing new markets, sort of optimize payments in their existing markets?
Lohrasp Seify:What data helps you understand is whether the actions that you're taking are having a positive or negative effect on your business. And then trying to use the data points that you've collected to vector yourself correctly. What should I do? Where should I focus on? If you've already expanded globally and you have storefronts and volume coming from many different countries, Obviously, there's a set of holy grail data that every merchant is after, which is authorization rates, what percentage of the transactions that consumers want to put through actually go through, transaction costs, and basically things like how many payment, how many different kinds of payment methods are being used for from a specific country, which payment method is successful, which one isn't, etcetera, etcetera.
Leo Tucker:We'll call those the holy grail, the big points.
Lohrasp Seify:The big points. Yeah, those are the standard things that everyone collects. I think depending on the maturity of your business, you can zoom in on any of the areas that I was just talking about, depending on how much time you want to spend and how detailed you want to get. For example, there's a lot of things that could contribute to authorization rates. Maybe your website has a problem and when people put in their card number it crashes.
Lohrasp Seify:Maybe you have people from a country routed to that website where the cards are not getting approved. Maybe the cards are getting declined at the issuing bank or the acquiring bank is refusing to do something with them. It's not just as simple as, well I just got to improve my authorization rates. If your authorization rates are not at a specific level or any of the data points that I talked about, it's not at a specific level, you have to decide how much you want to double click, how much you want to zoom in to a specific area and how you're gonna deal with that.
Leo Tucker:Well, maybe let's just do a single double click
Lohrasp Seify:Sure.
Leo Tucker:If you will, sort of beyond just, say, my authorization rate is 82%. Well, that's good. I'm happy with that, or I'd like a little more. What's the next level of granularity you'd go down to? What's our first double click?
Lohrasp Seify:Yeah, I think there's a lot of different ways to look at it, right? I don't want to set out it's my way or the highway. I'm just going to say how I would approach it or how we approach it at Reach.
Leo Tucker:Okay.
Lohrasp Seify:The very first step is out of the transactions that were declined, how many of them were correctly declined, and how many of them were incorrectly declined?
Leo Tucker:Now, when you say correctly, incorrectly, let's talk about that for a second.
Lohrasp Seify:Sure, yeah. So what percentage of my transactions are fraudulent transactions, which then later on would result in refunds or ultimately chargebacks. How many of them are legitimate transactions of a consumer trying to buy one of my products but then their card or whatever payment method that they're using gets declined and it doesn't go Obviously you could, let's just use a theoretical situation. If you could go to the bank and say I will underwrite and I will accept the risk for every single transaction that is getting processed. As in, if the consumer can't meet their obligations for the money movement, I will pay you.
Lohrasp Seify:There will be nowhere in the world that will say you know you're not allowed to do that. They say absolutely, 100%. So you can have a 100% authorization rate, but the price that you pay for increasing that KPI to 101% is good, right? You look at that number and you're like, yes.
Leo Tucker:That's true. I love 100%.
Lohrasp Seify:Compare me to my competitors. We have an extra 15% auth rate. But the cost of the portion of that volume that comes back and hurts you is a lot higher than how much you gain by that 15%. The overhead of chargebacks, the programs you might go on, the cost of defending it, etcetera, goes very high. So usually one of the first things that we do is, and this was one of the examples of unwanted transactions, but we try to figure out out of what was declined, how much of it should have been declined.
Lohrasp Seify:Now out of the ones that were declined but shouldn't have been, there is no evidence whatsoever for it to be fraudulent. We double click on the reason why it got declined. Was the reason because the issuing bank said do not honor Was it because of insufficient funds? Was it because the time expired, they didn't pass 3DS? There's just a lot of different ways that a card could get declined.
Lohrasp Seify:Out of that, there are things that are within your control as a merchant, and things that are outside of your control as a merchant. So again, if your consumers all have insufficient funds, and that's the reason why cards are getting declined, there's there's nothing for you to do to increase
Leo Tucker:Maybe market to a different audience, I supposed .
Lohrasp Seify:Maybe. Yeah. Exactly.
Leo Tucker:Maybe. Outside of that, you can't give your customers more money.
Lohrasp Seify:More money. Oh, the I I wish some merchants did give me more money. But generally speaking, you're right. That's generally the thought process. You know, I'm trying to not get super specific.
Lohrasp Seify:But figure out what you want and what you don't want, what's within your control, and the things that are within your control, which party do you work with to try and increase those? So for example, the do not honor, the acquiring bank issues, all of them are things that Reach does help with and does try to optimize to increase authorization rates.
Leo Tucker:Right. Because if you know what you're looking at, you can do something. If you're like, well, I have, you know, an 82% auth rate, and, you know, I don't love it, but that's all you have is that raw number, it's not particularly useful. But if you know that, you know, a certain percentage of those are are insufficient funds and, you know, you've got a bunch of, 3DS issues or, you know, do not honors, you can sort of do something with those buckets is what I'm understanding by simply collecting the data, you know which areas to attack.
Lohrasp Seify:Correct. Absolutely.
Leo Tucker:So we've we've bucketed everything. We say, okay. We've got of these, you know, of the declines, we got the ones that should have been declined and the ones that, you know, maybe shouldn't have been or and in there, we break those further down into our different categories. Let's talk about the ones we can do something about.
Leo Tucker:What can a store do to take a look at those, you know, denials and say, okay. We can fix this. Like, where should they look?
Lohrasp Seify:This this have a has a very large domain where there's entire companies whose job is to consult and help you understand how to increase your authorization rates. But just to draw a picture of the spectrum, it could be anywhere between simple checks that you put in place on your website, on the credit card numbers that are getting inputted on your checkout page to make sure that no one fat fingers a card number wrong.
Leo Tucker:Okay. So simply just validating the data you get. Like, is this actually a credit card number? Are we trying to send some junk to the payment processor? That kind of thing.
Lohrasp Seify:Right. And or if you fat fingered a one into a two. If you've noticed there are websites that then some text pops up and says, hey. This is not a valid card number. There are things that you can do as a merchant if you have control of your own checkout page.
Lohrasp Seify:Obviously, if you're with platforms like Shopify, they take care of things like that for you. But you can go as deep as changing every little configuration on your merchant account that you have with any payment processor that you can change, changed your classification of your merchant, to putting systems in place that builds trust with your consumers over time. What do I mean? This was a very vague way of saying, for example, if a consumer saves their card and has the same purchase, like in digital and subscription platforms. If the consumer subscribes and every month for three years they have managed to pay their bill, the thirty seventh time that something is going to get charged, the chances are it's going to get approved and it's going to pass.
Leo Tucker:Probably pretty good. Yeah.
Lohrasp Seify:Trying to make sure that the UX, the user experience, promotes things like subscriptions, save your cart data, instill trust in people so that they don't think, oh, if I save my data here, it's gonna get stolen, etcetera, etcetera. So you see some of it is technology, of it is process, some of it is building trust. There's a very large domain that you can tackle to try and make things better in terms of auth rates.
Leo Tucker:Sure. Some of that seems simply down to maybe not auth rates specifically, but conversion rates
Lohrasp Seify:Yeah.
Leo Tucker:Come pretty heavily with, you know, just taking out friction from a checkout. Do I have to click a thousand buttons on three pages to to buy these these loafers? Yeah. I'd rather not. Maybe I'll just buy them somewhere else where I can click.
Lohrasp Seify:Yeah. Yeah.
Leo Tucker:So I like this. I mean, data is all the talk right now, especially, you know, with AI being at the forefront like it is and all the news and everything. I'm often kind of reminded that collecting data is really important, but being able to do something with that data is is really where the rubber meets the road. In your day to day, you know, you deal with a, you know, a lot of a lot of merchants, a lot of customers.
Leo Tucker:Where does that collecting data and then knowing how to use the data and, you know, have it be actionable? Like, where do those tie together interesting ways?
Lohrasp Seify:There might be a lot of rhetoric about people that understand data, data scientists, BI analysts, data engineers, etcetera. They will come and revolutionize your business. And even though that's not completely false, that is also not completely true. Understanding the context of your business and being able to coherently explain what you are trying to achieve. I am trying to increase the authorization rate of wanted transactions.
Lohrasp Seify:Is a fundamental first step for you to not spend hundreds of thousands, if not millions, tens of millions of dollars on just collecting data and then not using it well.
Lohrasp Seify:The data points are many and understanding which of them actually help you and which of them don't is a challenge. That is a little bit of a science and a little bit of art. So just because you collect extra data, that doesn't mean your business is going to do better. You have to be intentional about what you look for and what you're trying to solve for.
Leo Tucker:Well, yeah, kind of in our previous example that ties right in like, I don't know, just knowing your authorization rate is a data point. But what is it where does it get you? Yep. Not not super far. Well, you know, it could be more, I suppose.
Leo Tucker:But simply breaking those down into little categories seems like a really a really good way to understand and use data. But you have to have that, you know, data from the payment processor, for example.
Lohrasp Seify:In the example that we were using that we kept double clicking and zooming on stuff, on the authorization rate specifically, at no point were we using data concepts, right? I didn't talk about facts and dimensions, databases, structured and non structured databases, data mesh, nothing.
Lohrasp Seify:We weren't talking about any of those things. We were talking about, logically, if you step through it, what do you want to separate from what? And why is one domain of data more important to you than the ones that could be successful, the authorization could be successful but it's not? You got to be careful to not get overloaded, especially because a lot of the platforms tout pitch that they have more data points than the other people. Well, they getting collected well? Are they actually useful for the stage of business I'm in?
Lohrasp Seify:What am I going to do with it? Just because you have it, that doesn't give it value.
Leo Tucker:Sure. So so we've we've talked about one. We've double clicked in pretty deep on authorization rates. We've given some various examples, declines that should, declines that shouldn't have happened, etcetera, things like that. Let's let's not drill down, but let's back up and take a look at another data point and maybe not one specific, but for a young business or a business that's just starting to take a look at their data in a larger way.
Leo Tucker:Where should they start? Like, what what other data is interesting, is worthwhile, is the most bang for the buck, something like that.
Lohrasp Seify:I think it depends on the context that you're looking at. And just to not be super specific into payments, I'll give a non payments example, then I'll do a payments example.
Leo Tucker:Okay, great.
Lohrasp Seify:If, as a business, you're trying to grow your business, obviously understanding who your client base are, where they're coming to your website to if you're an e com merchant, how many times they click so that you can quantify interaction, and how many clicks does it take for them to go to the checkout page are all good indicators. So if your main interface with the consumer is an online website, obviously something like Google Analytics or anything that a platform can give you stats on clicks and page views is a good indicator on how things are today. So we call that descriptive data. Your first step is as any business, you want to understand the state of things as they are. The payments version of that is authorization rates and how many different kinds of payment methods are being used to purchase your products from which country.
Lohrasp Seify:Again, that sort of data would be available on your ecom platform or if you're trying to manage and own your own checkout flow, you can gather that data as part of the checkout flow. Like I was saying, the descriptive data does not tell you what to do. It tells you how things are right now. Now you, as an expert of your own business, probably have ideas on what you think is going to make things better. So for example, you think that a marketing campaign targeted at 24 to 30 year olds is going to have the highest impact.
Lohrasp Seify:Since you have the descriptive system, descriptive data set up, you can iterate and implement some of the ideas that you had and measure their success using the descriptive data points that we've had So now you know, and this is by the way what we call hypothesis testing, that's at the core of any data science work or any data driven project, where you hypothesis test what you thought is going to happen versus what actually happened. And by adding a bit of rigor to it, for example, by not doing 10 different changes all at the same time, so now you can't attribute your success failure to one thing.
Leo Tucker:Isolate your variables a bit.
Lohrasp Seify:Exactly. You can you understand what has a positive impact and what has a negative impact on what's going on. If you accumulate enough of these data points, so hey, every time that I do a marketing for a specific age group, let's say it's 23 to 40 year olds, my sales go up, suddenly now what was descriptive and a measurement to the changes you put in place becomes prescriptive. So now suddenly you look at the same pattern in your data and you say, Oh, I've seen this in this other business. I think what we need to do is to market to 23 to 40 year olds for our sales to go up.
Lohrasp Seify:So that's what I just described as the maturity stages. If you want to dip your toes, you start from the descriptive. And the more sophisticated you get, the more you let the data inform the actions you want to take.
Leo Tucker:Oh, yeah. So as you as you collect data on, I'm selling this market campaign done to this age group and these regions, I've noticed, and we've seen time and time again, author rates go here, you know, sales go here. We sort of have a a description of a particular action instead of just a snapshot of your sales in real time.
Lohrasp Seify:Correct. Yeah. Exactly. Okay.
Leo Tucker:Yeah. So that's really interesting. I like the the descriptive versus prescriptive side of things. I think that's that kinda moves from from just observing to actioning something. But those are still pretty obvious.
Leo Tucker:You know? That makes sense to me. Like, I need to know my authorization rate. I need to know my decline rate. Probably should know fraud and chargeback and stuff like that.
Leo Tucker:Those seem pretty obvious. And if I run a marketing campaign, sales should go up. Yep. That seems pretty straightforward. Are there any less obvious data points that businesses might overlook that have real value?
Lohrasp Seify:The answer is obviously yes. And depending on your maturity stage, you might think of some data points as obvious versus not obvious. But I think the crux of the thing, just like any other problem, you can achieve 90% of your outcomes by putting in, what is it, 20% -25% of the effort in. So if you have 20%, - 25 % of the data, you can achieve 90% of the things you want to achieve.
Leo Tucker:Is that kind of our eighty-twenty rule?
Lohrasp Seify:Eighty-twenty rule, yeah. Whatever the standard way of saying it is. Thing that will trick people is what we call overfitting in the data science and machine learning world. So overfitting is when your theory, your idea about how your business, your payments, data, etc. Works is so complex and convoluted that it explains 100% of the things that happens underneath the surface in your data.
Lohrasp Seify:And you think for your theory to actually make sense, everything needs to be described. That is a very common trap that you fall into even if you're early in your career in data analysis. It's a very simple trap that you fall into. And when you overfit, your hypothesis, your understanding of what's going on in your business generally is just wrong. It's only correct for the data set for the time period that you collected it for.
Leo Tucker:Right. Is that just because there's too many variables like we talked about earlier, not isolating sufficiently?
Lohrasp Seify:It's not just too many variables, but you have to tune every variable exactly to be able to explain 100% of your data. But human behavior, data pipelines, the way that the transaction bounces around between so many parties, there is just error that's introduced into it. Sometimes there isn't a great hypothesis on why a transaction got declined. Maybe it got declined because the banking system is still on COBOL and it has a character in the name that it didn't like, you know? How are you supposed to know what it is exactly?
Lohrasp Seify:Rather than trying to get more specialized and more niche data about things and focusing there, my suggestion to small and medium sized owners, small and medium business sized owners is focus on the bulk, on the themes, on the bulk of your data and the themes that come out of it, and don't overanalyze what's going on. When you see a theme, track that theme down. Just because there's a data point, one data point that's so far away, and that that happened once in the last twelve months, You don't have to track it down.
Leo Tucker:Let's give an example here. So, you know, in your time here, you know, we're kind of talking about themes and maybe outliers as well. Yeah. Yeah. Talk about that a bit.
Lohrasp Seify:Let's say that you are trying to figure out what payment methods you need to use and which geography. So let's say you want to start sales in Norway, and you go and buy some Norway data to figure out what payment methods do people in Norway use.
Leo Tucker:Okay.
Lohrasp Seify:Let's just come up with a payment method called payment method X. It doesn't matter what it 'Payment method X' statistics show that the Norwegians, 25% of the time, when they are given the option to pay with 'payment method X', they do pay payment method X. So when you go to set up your store, find the correct payment service provider, you think to yourself, okay, given the report that I got, it looks like I need to have payment method X for something to work for me to be able to sell into Norway. And then you go and assess how much does it cost to integrate 'payment method x', what does the process look like, etcetera, etcetera. So you spend some money and that's when you go and implement it.
Lohrasp Seify:Then sometime down the road, let's say you figure out that credit cards cost less than 'payment method X', and you're trying to answer the question, if I remove 'payment method X', how much of my volume am going to lose? Given everything that I told you, if you've just looked at the surface level data, you think to yourself, well, my sales are going to drop by 25%.
Leo Tucker:Yeah, 25%. Twenty five % of people who have it offered to pick it.
Lohrasp Seify:Pick it, yeah. That's where, if you practice, and we have put that in practice at Reach, that is almost never the case. And I'll explain in just a second, but let me say what pitfall people fall in. When they see that it doesn't go down by much, and I'll explain why in a second, They think that they just have to collect more granular data and more data points from the shopper to be able to understand what happened. And technically they're not wrong.
Lohrasp Seify:They could spend months figuring out what data points need to get gathered. A lot of different ways to solve the same problem. If you have the context and the experience, you know that if you go to a website and your preferred payment method is PayPal, let's say, but they don't offer PayPal, you always have another way to buy something. Whether you choose to use the other method or not is your choice, but it's not like it's PayPal or nothing at all.
Leo Tucker:Right. Right. Right. You're taking it as, you know, 25% of the people use it, but only when it's presented. But, you know, a much larger percent of people will pay for it regardless.
Lohrasp Seify:Exactly. Exactly. So they have a preferred way. But the better way in English to ask that question, and this is why it becomes so important when people look at stats in the news or whatever to pay attention to the phrasing. The way that we should have phrased it at the beginning to get the data that they want was: What percentage of the consumers will leave your website if payment method X was not offered?
Lohrasp Seify:That is the true number of consumers that you're going to lose. I didn't collect any new data. My hypothesis didn't match the data collection, the data that I had bought that informed how I'm going to expand my business. Rather than getting lost in how do I collect more data to answer this question, I just tweaked my hypothesis. I didn't make my hypothesis super duper complicated.
Lohrasp Seify:Oh, that's because, well, when people that use payment method X are not using it, that means that they do reverse marketing with the same class of people people ages 17 to 19 are the ones that use Payment Method X. So when it becomes not cool to come here, then older people are going to come here more often. Know, like, you can just spin things.
Leo Tucker:Like complicated hypothesis explanation for it
Lohrasp Seify:Whereas the simplest answer was, hey, if we add APMs, they will cannibalize each other's volume. Yeah. So how do you isolate the performance of an APM to understand how much that APM is needed. Can you cut it? Is it costing you more or is it costing you less?
Lohrasp Seify:Did your GMV grow because of marketing or because of the APM? That's a very, very good example of how we look at stuff even. When we try to suggest APMs, we try to understand, well, this consumer that's using this APM, what other APMs have they used within our system? Do they have a credit card? Could they have used the credit card?
Lohrasp Seify:Does this user have all payment methods or just one payment method, as in if that payment method doesn't exist, that's it, they're gone? And what type of impact would that have on ours and our merchants' bottom line?
Leo Tucker:That's right. We had Melissa Pottinger on as a guest a while back, and we talked about the payment method buffet Mhmm. Which was a a good tie into that. So if you haven't seen that episode, check it out. Alright.
Leo Tucker:So, Loa, I just wanted to summarize a little bit here and and fill in the blanks just to make sure I understand it. But, you know, if we're talking small to medium sized businesses that are just starting to collect data, you know, obviously, your your holy grail data points, your authorization rates, your payment data, you know, take a look at the big ones, but drill into a little bit. And when we're talking about authorization rates, we wanna know, you know, payments that should have been declined, payments that shouldn't, and really dig into those ones that shouldn't. And also maybe just don't go while collecting a lot of data is good, maybe, you know, attributing every change in your business to a small little data point might not be, it might not be the way forward. Am I summarizing that correctly?
Leo Tucker:How would you what would you add to that?
Lohrasp Seify:I would say there's a lot of hype around it and for good reasons. But don't stress if you don't have the most granular and the best data possible. Your problem solving needs to be led with context and business knowledge and supported and measured by data. If you lead with data, you will get lost and you can keep going deeper and deeper into a realm of extremely complex and not useful territory. Figure out what you want to achieve, what you want to change, what you want to make better.
Lohrasp Seify:Describe your system using data. Make changes to your business as you see fit, the whole gut feel thing, but measure it. Figure out if your gut feel is in line and things are getting better or not. And when you do enough of these and you have data points, now you get to the, okay, the data tells me that I gotta do A, not B. Don't get lost in the jargon.
Lohrasp Seify:Don't get lost in data concepts. Think of data as a tool in your tool belt. You still have to design the house. You still have to architect it. The data will just help you achieve your goals.
Lohrasp Seify:They shouldn't be your goal.
Leo Tucker:Well put.
Lohrasp Seify:Thank you.
Leo Tucker:Loh, thanks again for being one of my favorite guests. You come back soon?
Lohrasp Seify:I bet you say that to all of your guests, Leo. Yes. Absolutely. Whenever you need me.
Leo Tucker:Absolutely not. Just don't check the end of any of the, sign offs on the previous episodes. Well, thanks, everybody. That's been Business Over Borders. Everybody, thanks for joining us for this episode.
Leo Tucker:If you enjoyed what you saw today, head over to our website, with reach.com. And we've also added some additional content and related links down in the description. So check it out when you get a chance. And as always, don't forget to like and subscribe and hit that bell if you wanna hear when we got something else coming up. Thanks.