How AI and Consumer Behavior Can Reduce Shopping Cart Abandonment | Valon Xhafa

Matt Edmundson

00:00:07.040 - 00:03:09.010

Welcome to the E Commerce Podcast with me, your host, Matt Edmondson. The E Commerce Podcast is all about helping you deliver E commerce. Wow.

And to help us do just that, I am chatting with today's very special guest, Valon Jaffa from Bahamics about how AI and consumer behavior can can reduce shopping cart abandonment. Yes, we are getting into all of that good stuff, AI shopping cart abandonment. What is not to like?

But before Valen and I jump into that, let me suggest a few other E Commerce Podcast episodes that I think you will enjoy listening to. Check out why you should be using AI in your e commerce business with Shaniff.

That was a great conversation with him and about how AI is changing shopping product recommendations, specifically with Oliver Edholm who was just. He blew my mind with that conversation. So do check those out.

You can find these and our entire archive of episodes on our website for free ecommercepodcast.net on our website. Yes you can.

You can also sign up for our newsletter and each week we will email you these links along with the notes and the links from today's conversation with Valen directly to your inbox. Totally free. It's all amazing. Now this episode is brought to you by the E Commerce Cohort which helps you deliver e Commerce wow to your customers.

Yes it does.

Valen, I'm sure you've come across a whole bunch of folks stuck with their e commerce website or they've just got siloed into working into just one or two areas of their business and miss the big picture. Well, enter E Commerce Cohort to solve this problem. Problem.

It's a lightweight membership group with guided monthly sprints that cycle through all the key areas of e commerce, the sole purpose of which is to provide you with clear, actionable jobs to be done so you will know what to work on and get the support you need to get it done.

So whether you're just starting out in E commerce or if, like me, you are a well established e commercer, I will definitely encourage you to check it out. Head over to www.ecommercecohort.com for more information.

Or you can email me directly matcommercepodcast.net with any questions and I'll try my level best to answer them. But head over to ecommerce cohort.com Honestly, you're going to want to check it out. So Valon is a top bloke? Yes, he is.

He has previously worked as an AI scientist at Google and other research institutes, developing sophisticated AI techniques and algorithms. Vallon is always looking for innovative methods to use artificial intelligence to improve the online shopping experience at Bahamix.

Valen, it is great to have you on the show, zooming in all the way from New York. Thanks for being with us. Great to have you.

Valon Xhafa

00:03:09.970 - 00:03:12.530

Hey Matt, thanks for having me.

Matt Edmundson

00:03:12.850 - 00:03:38.990

Oh no, brilliant, brilliant, brilliant. So Valon, you have an impressive resume or CV as we like to say here in the uk. An AI scientist that would make a really interesting business card.

Do you know what I mean? Where you've got your name? I'm an AI scientist and the conversations that that would create at dinner parties.

So you've obviously done some interesting stuff. You find yourself now at Bahamix. So what is Bahamix? What does it do?

Valon Xhafa

00:03:40.350 - 00:04:45.840

Yeah, I started Behemix or I founded Behemix right after I left Google as a way to apply AI in E commerce. So what I saw is that AI is widely applied in other industries like anonymous driving, healthcare and finance, but not a lot in E commerce.

And I expected that AI could make a huge impact in E commerce too. So I saw Behmix as a way to apply AI in E commerce to reduce shopping cart abandonments. Like shopping cart abandonments.

It's a big deal and 70% of cards are abandoned and there is no feedback or clear facts or statistics. Why do we have shopping cart abandons?

So this was as a challenge for me and that's why I started being to understand why do we have shopping car abandonments? How can we prevent shopping car abandonments?

Not just like recover them through emails and all these retargetings and everything that you usually do, but also be able to understand before they happen or predict before a car is abandoned.

Matt Edmundson

00:04:47.360 - 00:05:21.930

So you want to predict shopping cart abandonment before it happens, which is, if I can put it bluntly, sounds almost a little bit like witchcraft. Right?

It's that kind of, it's this sort of, it's this black hole of stuff where I think people like me to sort of get lost in our thinking a little bit. But I has made some interesting advancements.

But before we get into that Valon, let's just for those that are new to E commerce, what are cart abandonments? What do you mean when you say car abandonment? 7% car abandonment.

Valon Xhafa

00:05:22.890 - 00:06:14.490

Yeah, there is a kind of like a definition of the car abandonment, a simple one which is like when a user visits your shop and they add a product to the car and then they leave without buying that product. So that's usually, it's called car abandonment. And because of that you have like 70% of users, they abandon their Cards.

So we're talking about like 70% of users who added something to the card, you know, and not 70% of your whole traffic. Right?

Because there are a lot of users who just come because they're just like, yeah, just want to kill some time and just like enjoy themselves, like just looking at products and everything. You know, like everyone, all of us pretty much, we do that.

But Car band is like this group of consumers or users who add something to the car and then they don't end up buying them, but they just leave. So that's the car abandonment.

Matt Edmundson

00:06:16.230 - 00:07:03.890

So people have added stuff to their shopping cart but they've just not purchased. And this is, this is a big problem, isn't it? I say it's one of these things that I've heard talked about off and on for a long time now.

You know, it's one of the sort of the key areas that people like to focus on. It's like what is stopping the people once they've added it to cart? What stops them paying me money? And I want to understand that process.

So why, why is this and, and the other thing that I've noticed and Valen, correct me if I'm wrong here, is the percentage of cart abandonment seems to be going up. It doesn't seem to be generally falling. It seems to be it's becoming more and more common.

Is that a fair reflection or have I got that entirely the wrong way around?

Valon Xhafa

00:07:04.770 - 00:07:34.050

It is, it is a, it is a. It is a statistic. It is a trend.

So car abandonments are going up and connected to that also product returns because these two concepts are tightly related to each other.

So these are statistics or KPIs that are unfortunately going up because of a lot of issues, a lot of reasons out there that we are going to explore them during this podcast. But yeah, that's the reality.

Matt Edmundson

00:07:35.010 - 00:08:15.240

It's interesting actually that you've linked there product cart abandonment and returns something that I want to come back to. So I've jotted it down. Whenever I look down, I'm making notes.

By the way, I'm not checking my email, just being clear, checking the football scores or playing sudoku. So what are some of the main reasons that you have found for cart abandonment? Why did we still partake in this? And let me be frank, right?

I catch myself doing it, adding stuff to cart and then not buying. What are some of the reasons we do it?

Valon Xhafa

00:08:16.520 - 00:12:00.500

So they're like external and internal factors and we're talking about like the soar. External factors regarding the soar and internal factors.

At Behamix, we mainly focus ourselves on the internal factors because these are like factors that we can actually impact and we can improve and prevent car abundance. And, and these factors usually account for the majority of the reasons.

So if we're talking about some of the main reasons, we group them into two categories like behavioral issues or behavioral reasons and technical issues. Technical issues within your shop, right, so behavioral issues are simple.

Put like users, as we, as humans, we have part time making decisions when we have a lot of options to choose from. Imagine for example, for you or for me or for anyone else who had a lot of options to go to good universities, you know, or got a lot of offers.

It was a really hard decision to make, you know, but when you have only two offers, it's easier to make a decision. It's the same story everywhere, right?

If you want to buy a car or jeans or T shirts or whatever you want to buy, the more options you have to choose from, the harder it is to make a decision.

And going back to the shops, usually within a shop for a T shirt, you have like 20 different t shirts within the page, you know, which look pretty similar. And you only want to buy one T shirt. So you're going to be like, okay, which one should I pick here?

Because a lot of them, why is it different from this? Why is the price changing?

I mean this looks a bit better, but the price is higher or the price is lower, you know, all these different confusion points.

So what happens here is that what usually users do is that they shortlist some of these T shirts, let's say out of 20, they pick like two or three, they add them to the cart. But here's the catch.

Most of the brands, most of the people think that the user has intention to buy this product just because they added these three T shirts to the cart. Which is not true because users actually, they think the car as a wish list. You know, the wish list that you usually have that you don't usually use.

And most of the users, they don't use, you know why they don't use it? Because the card itself, it's actually wish list, it's a short list for the users.

So when a user adds three T shirts to the car, they're not adding because they want to buy all of them, they're adding because then they have a short list out of which they can make it easier decision. And then when they go to the car, they have three different T shirts and then they can think about it. And here's where returns are related.

Because if the user cannot make a decision to pick one T shirt out of these three different T shirts, they will say, okay, maybe I can pick two and then buy them and then make the decision at home. So are you following me here? How the users actually push the decision making as much as possible into the future?

First they say, okay, let me just, I'm going to make the decision when I go to the car. And then no, maybe I'm going to make the decision when I have the product at home. This is a very human nature. We like to push the decision making.

We like to procrastinate. We like to push decision making as much as possible in the future. We like to push the hard decisions as much as possible into the future.

You know, that's the same plan that happens here in the soar. So that's kind of like related to the behavioral part. We have a lot of options to choose from. It's called also paradox of choice.

You have so many options to choose from, you just. You get suffocated from the decision making process.

Matt Edmundson

00:12:00.740 - 00:13:07.990

Yeah, I've heard about the paradox of choice.

I'm trying desperately to rack my brains of the book that I came across it in where they talked about the famous jam experiment where in a supermarket they put six jams in a suit. You know the one I'm talking about, they put six jams in a supermarket island. They monitored sales. So you could go and you could try them, right?

They put six flavors out, you could try them and they monitored sales and then they put, I think it was something like 36 different flavors out and monitored sales after that. The idea being that the more choice you have, the more likely you are to find something that you want, therefore sales should increase.

The paradox of choice was actually the more options you gave to consumers, the. The more they didn't buy, the more they put off, like you say, making that decision with jam. And I thought it was a fascinating experiment.

And so what you're saying here actually on your e. Commerce website, a lot of car abandonment is down to this paradox of choice. You've got a lot of options and what that does is.

And I think you're right. You know, we've all got. What's that phrase? Decision fatigue.

Valon Xhafa

00:13:08.790 - 00:13:09.270

Yeah.

Matt Edmundson

00:13:09.350 - 00:13:43.420

Do you know what I mean? We're trying not to very hard not to make any more decisions. And so we just keep putting them off and putting them off and putting them off.

And of course by not making a decision, you make a decision. And I'm laughing because I purchased two coats off a website recently, and I purchased two because I wasn't sure which one I wanted.

And I thought, I'll get one and I'll return the one that I don't want. But I need to see it in person a little bit. And I wasn't quite sure on the sizing and the fit and all that sort of stuff.

So everything that you have just said, Valon, I am guilty of. Right, so everyone is guilty. Yeah.

Valon Xhafa

00:13:43.420 - 00:13:59.100

So it's, It's. It's a human nature. I mean, we're like.

I mean, decisions we make, and the way we make decisions in the real world outside the online shop, it's the same process in the online shop. Right. So it just. We just get suffocated and it's hard to make a decision. So.

Matt Edmundson

00:14:00.540 - 00:14:13.640

So we've got the paradox of choice. We push decisions to the future. What else sort of affects cart abandonment?

Is that the key thing or are there other things that we need to think about? You said behavioral and technical.

Valon Xhafa

00:14:14.440 - 00:17:20.910

Yeah. So regarding the behavioral part. Regarding the behavioral part, everything is.

I mean, there are different layers of issues that can happen, but everything is tightly related to this.

Having a hard time making decision, being unsure about making decision, or making the wrong decision, you know, that can actually lead to a product return, you know, so all of this is. It's about making decision, and this relates to the behavioral part. But then the second category of reasons why we have carbon are technical issues.

Like you have a bug or you have an issue with the payment, or you have this or you have that. So the thing here is that there are so many tools out there that can actually track all these things for you.

The issue, though, is that technical resources are scars for online shops. Not only for online shops in general, especially for online shops. Right. So you won't be able to fix the issues right away.

You won't be able to understand, okay, why do I have how big of the problem this is? Like this technical issue. And the most important thing is that you need to actually prioritize these issues.

You need to understand, okay, how big of an issue this technical bug is so that I can go and fix it.

For example, there are bugs and issues or errors that can happen on the payment page, and they only have to, like, 10 times a day, you know, but these 10 bucks can actually cause you a lot of revenue loss compared to some other issues that happen, I don't know, like on the product detail page that appear like thousand times a day, but they don't cause you as much issues, you know, so there are so many breaking points within the shop that can happen. And it's really important to understand, okay, how are these issues tied to car abandonments? Because here's the thing.

If I add a product to the car and I go and want to make the purchase, and I have issues with a payment, and you as a shop owner, you didn't recognize that I had an issue with a payment, and I ban the other car because of that. You can send me whatever email you want, Matt, with a coupon or something. I'm not gonna come and buy because it's not.

Because usually, like, coupons are the way to recover card abandonments, you know, with emails and usually get a coupon when you abandon a car.

But I'm not gonna come and buy on your shop because I had a technical issue, I couldn't pay, you know, so instead of sending me a coupon with an email, why don't you just contact me and provide me help to actually make the payment? So what I want to summarize here is that there is no, like, overview of what's going on and how are these issues related to car abandonments?

Because everything is pretty much related. Like the latency, like the page or the product image doesn't lo. This is. This can cause a huge issue.

You know, maybe it's like, oh, maybe just the product is unloading.

But yeah, this is interconnected with the user experience and everything, you know, so, yeah, bottom line, you have behavioral issues or reasons and technical issues that actually lead to car abandonments.

Matt Edmundson

00:17:22.350 - 00:18:31.630

So actually, I'm listening to you talk, Valen, and I'm going in my head, I'm thinking, I get things like the paradox of choice. I understand that.

I get things like, if I don't clearly tell you what my shipping rates are and you get to a page and it suddenly surprises you that they're 400 bucks, you're going to go, well, no, I get the logic of that and I get the logic of what you're talking about. But the more you're talking, the bigger and bigger the web sort of seems to go, if that makes sense. Sure. And so there's.

Cart abandonments are a big issue for E Commerce owners. And the reasons for cart abandonments can be behavioral, they can be technical, they can be obvious, they can be really unobvious. And there's.

There's a. There's a.

There's a lot going on here that's going to require a little bit of attention, which is where I'm assuming Valen, if I can sort of preempt the conversation a little bit, I'm assuming this is where I actually can actually make a really big impact. Right. If you could apply AI somehow in this to be your 247 eyes and understand what's going on and reasonings, Is that right?

Valon Xhafa

00:18:32.670 - 00:21:56.360

Yeah. So that's actually the contribution of AI here in this space.

Because AI not only predicts if the consumer is going to ban the car, something that is common, something that is general knowledge, but we have actually also pushed it a bit further by also building an AI that can explain to us why is user gonna abandon the car. The why, the reason. So imagine you, Matt, come to my shop and we predict you're gonna ban the car.

But then I'm also gonna understand why you're gonna ban the car. You know, like this specific reasons before you even do it.

And through this process I can actually intervene while you're still on site and help you and support you or whatever you need.

If you're gonna ban the car because of the technical issue, then I can assist you saying, hey, we see that you're having a hard time with payment or you didn't see this image loaded or something. Can we help you somehow?

Or if we see that you have a hard time making decisions because of this paradox of choice, or you have so many options or you're not finding the right size and we can predict that you're going to ban it because of these issues, then we can actually reach out to you while again you're still on site and try to provide support according to the reason why you're going to ban the car. Right.

So it's all about like the biggest issue is that when you bring a consumer or user to your shop, you already paid a lot of money, you know, so if you leave this consumer to leave your store and try to bring them back to emails and retargeting, this is going to cost you additional money. The solution here is like, you need to react while this user is on site and be able to intervene right away and prevent things before they happen.

We know this concept. If you prevent something, it's more cost effective than trying to recover or repair or something.

Like it's with cars, it's even with health, you can predict diseases and prevent them and you can treat them. This will be more efficient, more successful and more cost effective than just leaving it go big.

So that's the same concept that we're actually applying here.

Predict, prevent and try to keep the user on site as much as we can and not say, oh, I can send emails and bring the consumer back because it's going to cost you additional money. Especially these days. The biggest issue these days is that the cost of ads have gone so up that it's very costly to actually bring.

Or it costs you a lot of money to bring consumers or additional traffic to your store. Until a year or two years ago, the only way to grow was to bring more traffic to your store because it was cheap. You could just get more traffic.

You're good to go now, even if you have the money, it's gonna cost you so much money to actually just bring additional traffic.

So you have to go back to your side and see, okay, what else can I improve that the traffic that I have right now, I can actually increase my profit and margins, but also increase customer satisfaction because it's really important.

Matt Edmundson

00:21:57.880 - 00:22:30.270

So how I. It's. I mean, it sounds wonderful, you know, the sort of, the key problems and that you have.

So how does AI, If I go back to the sort of list that you gave earlier on. So let's look at the paradox of choice. How does AI help a consumer deal with this paradox of choice? How does that, how does.

Well, have you, have you figured that out? Right? Because that sounds, that sounds like an easy thing to roll off the tongue, right? I can just say that. But actually, how. I'm really curious.

How does that. How does it do it?

Valon Xhafa

00:22:31.470 - 00:24:58.950

Yeah. So we can take a very simple example to maybe explain the paradox of choice. It's more simplistic for the audience.

So one of the reasons why consumers abandon the cars is because they just add too many products to the car, which is maybe a misconception, because people or brands think that when consumers add products to the car, they have a higher chance of buying. Yes and no.

Because after a specific number of products in the car, your probability of buying goes down because you're not gonna buy 20 products or 15 products, you know, so you're just shortlisting them. So this is the case when consumers abandon the car.

And what our AI, for example, does here is that for every product you add to the car, we get a probability score of what is your probability of buying?

You know, and if we see that, okay, your probability of buying is increasing with more products, and then suddenly it starts to decrease, then we know that maybe it's the right time for you to remind to you, Matt, that maybe you can have a look on your car because it seems that you're ready with the car. It's like a gently Reminder.

This kind of like you know when you go to the shop, to a physical store and you have this shop assistant helping you with advices and maybe try this or maybe try that or maybe this would look good or this or that, right. So this is the touch and the feel that you don't have right now on online shopping experience. It's just like one way interaction.

You have to go, it's self service pretty much. You have to go and find the right products and everything. And what happens here is that most of the consumers, they just lose themselves.

They just like by trying to find more product and then spending more time, you know, instead of like instead of focusing on these three, four products that they already added to the car and see if they can make a decision here. So what our AI does is that or AI predicts if the consumer is going to add too many products to the cart and then they're going to ban the cart.

And before they do that, we just provide a gentle reminder to the consumer on site, you know, like a small note or something. Hey Matt, it seems that your card, maybe it's ready or something. Why don't you have a look on it?

Because we know if you continue doing that process, you're just gonna add too many products and it's gonna be harder for you to make a decision and it's just like end up leaving and not buying at all. So that's for example one case. But there are a lot of reasons.

Matt Edmundson

00:24:58.950 - 00:26:11.780

Out there that's really clever.

I mean that it's interesting that I mean forget all the, the AI, just the psychology of saying actually no, now's the ideal time for you to check out, but we're going to help you remember to do that because we want you in the shop but not too long. I'm kind of reminded of the supermarkets.

You know, they, I don't know if they still do, but years ago they, they used to play music at different speeds depending on how long they wanted you in the shop. Right. So if they wanted you in the shop and out because it was busy, they would pay, play faster paced music.

And if they wanted you to hang around the shop for a while and just browse and add a few more products to your basket, they would play slower, calmer music, just chill out, just take your time. And it's kind of what you're describing reminds me of that.

Valen, you know that actually what you're doing is you're going, no, the optimal time for this person buying these types of products is like I need them out of the store within 4 minutes and 5 seconds.

Which kind of goes against the logic which says actually we want them to hang around in the site, we want them to read stuff, we want them to get sucked in and drawn in. Actually you're going, well, that's not always the case. And you're using this sort of AI magic to fix. Figure that out.

Valon Xhafa

00:26:11.780 - 00:27:40.230

Right, exactly. So the magic here is that not, not every single customer is the same. So there is no average consumer. The average consumer is not, is not real.

It's idealistic, you know, so it doesn't exist. So every single consumer has their own time frame of how much time they want to spend or they're going to spend.

And we can actually predict that, you know, so if we predict the consumer is going to spend four minutes, then you can just waste your time trying to have them more on your store. Right. Or longer in your store.

Which means that if you see that the consumer is ready to buy, if you predict that and they don't have enough time, then the best possible way here is like, remind them to go to the car and see if they are willing to buy any of these products. Instead of like trying to confuse them with like coupons and like showing a pop up asking for their emails and promotions and this and that.

You know, like really have a lot of time. Let's say I'm on the subway going to my work and I see an ad and I click a product and you know, like on subway you have a lot of time.

You have like three, four minutes time spent and I add that to the car and suddenly I get like bombarded with like, oh, here's the promotion, here's this. No, I saw an ad, I added this to the cart. I want to finalize this purchase.

So why don't you help me focus on this instead of you trying to upsell and then lose me completely.

Matt Edmundson

00:27:40.310 - 00:28:38.560

Yeah, I like the message that you're preaching, Valen, I really do. I think it's so wise.

I wrote down this phrase you used earlier and when you said that when you were talking about how there were too many options, you called them confusion points.

I don't know if that's a deliberate phrase you use, but I thought it was a really interesting phrase that actually what we're doing is we're on E Commerce at the moment. We're creating a lot of confusion points and some of them you've hit upon like, give me your email address, do this, do that, do the other.

And it's like, I Just want to get this product and go, dude is what I want. Right, so what sort of things then does AI look at? So you talk about predicting the behavior and so what sort of does it look at everything?

Like what, whether you're on mobile, you know, whereabouts in the world you are, what time of day it is, what sort of things would AI look at and sort of start to take into consideration, start to calculate behavior.

Valon Xhafa

00:28:39.520 - 00:31:30.540

Yeah, so they're almost like behavior based data, like clickstream data.

In our case, we don't use personal data because we don't, you know, we're not going to send emails, we don't, we don't need to know what the user did like 3 months ago or something. It's pretty much completely behavioral data. Like which ad did you click when you came to the store?

Or which device are you using, which type of device? And then where did you land on the shop? Did you land on the product detail page or on the home page?

Which products did you click, the sequence of the product clicked, and all these different variables. So we use around 200, 250 different variables to actually be able to get a better understanding of the consumer.

And with every click the consumer does, we actually predict the behavior and we measure the probability score of buying. Because this is the most important thing here. So what happens here? Is that the confusion point that you mentioned?

No, this is something that I actually mentioned and I attended it because we have seen a lot of cases when the consumer was ready to buy, high probability of buying, and then the brand showed the pop up asking for an email and they just lost the consumer. The probability of buying dropped to like 50% or something and they just left.

Or there's also another big confusion point which is related to upselling. Upselling is great. You know, upselling is something that can actually make you money. Increase the average over values and everything.

But after a specific point in time when the consumer has added enough products to the car, don't try to upsell too much because you can lose the consumer. If I go the car and then I have three products that I want to buy and one of them is a T shirt and then I'm ready to buy. High probability of buying.

And then you provide a recommendation on the bottom saying like, hey, these are also some other T shirts. So this is a confusion point because maybe now I like one of the other T shirts which are being recommended to me. So now I'm gonna change my mind.

It's like, damn, which one should I pick? Should I go with the One that I was ready to buy or not with this one that I didn't see before. And then what happens?

Like maybe you add this new T shirt and then you just like, no, I don't know which decision to make. And there are like two options you're gonna do here.

The first is that you're gonna leave the store without buying or you're gonna buy both T shirts and then return one of them, which is not good. Returns are not good.

You should not do trade offs between returns because 1 percentage increase or percentage point increase in return is not a 1 percentage decrease in revenue, it's 4 or 5 because you have all the cost associated to returns.

Matt Edmundson

00:31:30.620 - 00:31:31.180

Yeah.

Valon Xhafa

00:31:31.420 - 00:31:48.500

So upselling is a confusion point. These pop ups like asking for emails and everything are a confusion point.

Like users are different and you have to be able to actually target them better if the user is willing to buy. Why are you asking for an email? You're going to get the email in a way. Yes. You know.

Matt Edmundson

00:31:48.500 - 00:33:49.760

Yeah. So that's clever. It's interesting because what you're doing with AI or what I'm understanding, Valen, let me put it that way.

Rather than tell you what you're doing, here's what I understand. What I'm understanding is you're using AI and AI is using hard real data, right?

It's using 200, 250 data points to try and predict what is going on here. And it's, I'm assuming the AI is very specific to that store, the behaviors for that store.

Because it's going to be different to the store, you know, further down the road, isn't it? It's going to have a different set of behaviors attached to it.

And you're using data to figure that out rather than what tends to happen is I will go to a website, someone's written a blog post which says you must have three upsells per product to maximize average order value. Right. And so in my head I go, I must have three products to upsell.

And actually what you're using is data which says no, actually for your website you should have a maximum of 2 based on this data. And actually the products which you upsell, they really matter as well. So it's not just the fact that you upsell, it's what you upsell.

And if you upsell that, the knock on effect is this.

And I sit here and I listen to the T shirt example and I go, well, from a logical point of view everything that you have just said makes sense, but I would never have thought about that in a million years. Which is why I guess when I go to the T shirt websites, they're trying to upsell me yet another T shirt.

And so I like the smartness of this and I like how you're using the data. Can I ask you a question? How much data do you need? As in should I be plugging in your AI if I'm just starting out?

Or should I be plugging in your AI if I've got more than 200 product SKUs? Should I be plugging in your AI if I've Got more than 10,000 customers?

I'm really curious how much data you need to start building this kind of predictive model.

Valon Xhafa

00:33:51.040 - 00:35:37.980

We do have this actually a really, really good question. Like our AI is very agile in this way because we have our own pre built models and AI systems and everything.

So it's not that when we go live with your shop or when we install our AI in your shop, we're going to start from zero. No, we already have knowledge collected from different stores, from different categories, from different products.

Because the most important thing here is that you said that maybe for your Shop2Products upsell is good, you know, which is partially true because again the thing here is that there is no average score, there is no average user. Average user doesn't exist. You know, in most of the cases every user is different.

And the most important thing is that, okay, let's say for example that you two products to upsell is good for your shop. But this can also change within seasons.

For example, when you're like in winter season you won't be able to sell to upsell to two products because the price is high. You have like jackets and everything else which are expensive. So if you can upsell one product in winter season that's, it's good.

But for example, during summer you can maybe you need to upsell three products. Let's say if we talk about the average.

But again the average is not, it's not a good, it's not a good way to measure how many products you have to upsell to consumers and everything else.

So like going back to the topic of like how much data it's usually as soon as you install BMX we can actually make predictions right away within 1, 2 hours of data. We just need to reiterate and readapt everything because we already have like baseline systems capable of making predictions.

Matt Edmundson

00:35:38.860 - 00:36:13.280

And I guess BMX is always learning, isn't it? It's always kind of adapting to what's going on and that's the beautiful thing about AI. It's this constant learning machine, isn't it?

Can I ask you say you've collated all this data from websites that you've been on, and obviously you've seen stuff as a result of this.

What have been some of the most surprising things for you, as in you've done this on the car abandonment, you've looked at the stats, you've looked at the data coming out of stores with your AI. What have been some of the most surprising findings? I'm curious.

Valon Xhafa

00:36:16.800 - 00:36:44.240

I mean, the findings could be like interesting findings or funny findings, which are findings that maybe they don't have a huge impact. It can be also findings that actually they're not that interesting, but they have a huge impact. So I can start with funny finding.

One of the funniest findings that we found is that users these days actually also sometimes abandon the cards on purpose because they know they'll receive an ebook with a coupon.

Matt Edmundson

00:36:44.480 - 00:36:44.880

Yeah.

Valon Xhafa

00:36:44.880 - 00:38:38.930

And we're able to actually, we were actually able to see this pattern. We saw the consumer, they abandoned the car, even if they had a high chance of buying, because our AI can predict that they abandoned it on purpose.

And then after a few hours, we saw them coming back and adding the coupon that they got from the emo and coming back from email.

You know, so this is a pattern that we actually saw because recovery emails with coupons were great in the beginning, but now actually consumers also sometimes abandon the car on purpose and hopefully to get a coupon. So this was a really funny finding that we found, but the one that is interesting or that can have huge revenue loss.

Our findings were brands, they know they have issues within their shop. I mean, we've been talking like, about like brands that are like $100 million in annual revenue or something.

They know they have an issue, it's visible, and they can go and fix it.

Which is like you have, for example, filtering options are not working or image images are not loading or something because they know that this is an issue. The challenge here is that they don't know how big of this issue is.

So this is why it kind of like led us to think maybe we should tell them, hey, this image that is not being loaded is costing you 20 grand a month, so that maybe they can actually go and fix it.

And this is something that is actually across most of the clients we work with and that we have seen in the industry, there are issues that are clear they are not being fixed. Issues that can be actually fixed easily. And can actually make them more money and generate more money.

And because these issues can actually lead to car balance directly, if the filtering option is not working, then this is another reason for me to leave the store.

Matt Edmundson

00:38:40.290 - 00:39:04.770

So is this before we hit the record, but we were talking about a new development that you guys have got called Flow, which you're excited about. Is this what you're talking about about here? The ability to put a financial value on, in action, for want of a better expression on your website.

The stuff which is wrong and not working. It's costing you 20 grand, 30 grand. The image not loaded. Is that what Flow does? Your new. Your new product?

Valon Xhafa

00:39:05.410 - 00:41:52.570

Yes. So we. We took the findings from the psychology part.

So we're like, price is something that can have a huge impact on everyone, you know, and it's the same story also for brands. So if brands see that, okay, this error is happening 200 times a day, that really.

It's not really urgent, you know, but if they see that this error is causing them, like, 20 grand a day, this is urgent, you know, like, price is really important.

And if you really want to prioritize things, you really have to kind of, like, be able to see and prioritize and rank them based on the revenue that these issues are causing so that you can go and fix them and even escalate the issues. So that's what we kind of, like, did here.

We applied AI again, as always, to be able to explain things that are happening within the store and help us attribute every single issue to a revenue loss. Which means that if you fix that issue, you're gonna get that revenue loss right away, pretty much.

And we run the tool where actually we launch it with our existing clients and everything, and we see, like, astronomical values in terms of loss because of small issues. We see, like 50 grand a day, which is like 15% of the whole revenue, you know, like 20 grand a day because of small issues.

And these issues are either, like, on the payment or they are, like, on product list or images or whatever. These are issues that can be fixed. But the thing here is not just fixing the issues, because you can fix the issues and then you're good to go.

Which is not true, because the whole idea of the Zoom is to actually prevent these 50 grand in the future, because whenever you push new things and you change new things to your store, you constantly add new stuff, and the chances of issues appearing are high. So let me give you an example.

So we have this client who told us that, look, we added additional payments options, and it actually we lost revenue, we'll lose in revenue. How is that possible?

And then they told me that they took them a month to figure out that some of the payment options were not working all the time for some devices for some days.

You know, so pretty much it backfired on them and backfired because adding more payment options works or increases conversion rate or whatever that part because they didn't have a solution in place to actually tell them, hey, you're losing money because of this and this is a new issue.

So if they would have our tool in place or tool would be able to tell them right away, hey, look, you have a revenue loss because of this issue and then they would be able to go and fix it. So that's kind of like the direction that's.

Matt Edmundson

00:41:52.570 - 00:42:04.780

I mean, that's incredible. That's a product in its own right. I see why it's a standalone product. And that's a, that's a beautiful piece of technology.

You know, it's like a constant monitor of your website. This is wrong. It's costing you this much money. Fix it. I mean, that's.

Valon Xhafa

00:42:04.780 - 00:42:05.220

Yeah.

Matt Edmundson

00:42:06.020 - 00:42:10.500

For anybody that's running an E commerce store, I'm like, that's a beautiful thing.

Valon Xhafa

00:42:10.500 - 00:42:10.700

Right?

Matt Edmundson

00:42:10.700 - 00:42:39.950

Where do I sign up? Because I mean, that just sounds great.

So you've got this technology which monitors that, you've got this AI which you know, figures out when the best time to check out is and a few other bits and, you know, helps with the behavioral side of things. Can AI also help customers make the right choice about a product so that they don't return it?

So that, you know, it's not like me going, oh, I'm not sure about this coat or this coat. And then returning one. It helps me make the right choice in the first place. Can I be that clever?

Valon Xhafa

00:42:41.150 - 00:43:27.970

Yeah. So we apply AI for like two specific, let's say, goals.

First is to predict if the consumer is going to ban the car, because if the user bans the card, then there won't be any returns because they won't buy.

But if they are going to buy a product, and if we predict that they are going to buy and purchase a product, then we also predict and try to explain if the user is going to return that product, individual product, and why before they even buy it.

So pretty much we can actually, with 95, 6% accuracy, we can predict for every individual product if they are going to be returned in few weeks or not as soon as they are bought.

Matt Edmundson

00:43:28.290 - 00:43:30.130

Well, that's quite a high percentage.

Valon Xhafa

00:43:30.690 - 00:45:50.620

It's A high percentage, yeah, because it really shows that there are patterns. Because otherwise if there are no patterns, you won't be able to predict.

This just shows that there are a lot of patterns out there that you can actually take into account to do these predictions.

But the cool part here is that we can predict if the user is going to return the product before they even buy the product, just by having them on the card. And then we can explain, okay, why is the consumer going to return this product?

For example, one of the reasons why consumers return product is that they really want to have a size, a specific size of that product, but that size is not available right now. So what they do, they go and add the closest size to that, like a smaller or a larger size, which is not good because again, it's not the right size.

And we can actually predict this. And in that case, we would actually try to recommend them to swap this product with a similar product that has this size available.

Because again, it's really important that we prevent returns before they actually are purchased, before they actually are returned, because that's kind of like the most important thing.

But there are so many other cases, like, for example, users, they just go and add like three, four different, or three, four different products of the same category or something so that they can make decision back at home because they're not conscious of the whole environmental mess and the shipping costs and everything that can actually lead to that.

So we kind of like gently remind to them that, hey, if you add like four products and we predict that you're going to return three of them, you know, we sometimes add a gentle reminder that, hey, these products will be returned and this can cause environmental damages, you know, because of CO2 and everything else that is related to that, you know. So these are some kind of like gentle reminders.

Everything is gently suggested, like reminded to the consumer that, hey, your decision has a consequence. You know, it's not like recommending them directly, hey, you should buy this product because three other users bought this product too.

You know, that's kind of like sounds like too much to push it here, you know, but we're trying to tackle it from like gently psychological perspective.

Matt Edmundson

00:45:51.020 - 00:46:58.480

Yeah, I like that. It's a bit like when I go to a hotel room and they put on the towels, don't they?

You know, if you want us to wash these, put them here, otherwise we'll leave them alone because it's going to save the planet. And it's that sort of gentle reminder.

It gives the consumer a bit more choice, which Seems to work very well from a psychological point of view because you don't come across as preachy, but you do remind the consumer of what's important to them, which I think is really clever. You know what, Valen, I'm aware of time, right?

And I feel like as I say this, if you're a regular to the show, I'm really sorry for the amount of times I say. I feel like I'm just scratching the surface. It's a genuine thing. I feel like we're just getting into the conversation here.

Question for you, right, bit sort of left field question. As you know, this show is sponsored by the E Commerce cohort, right, which is just, it's like a mastermind group.

So imagine you're in a hotel somewhere and everybody from cohort is there.

You've just delivered your keynote speech, you've talked about how AI is going to change their world and rock their lives and they're like, yeah, it's amazing. Best speech.

Valon Xhafa

00:46:58.880 - 00:46:59.400

Well done.

Matt Edmundson

00:46:59.400 - 00:47:21.930

Violin. Go, go, go. And you, come on, you take about and you take that opportunity to say thank you to various people. You know, I wouldn't be here without.

I'm curious, who do you think is it what people, what books, what podcasts, you know, what who's kind of impacted your life and got you to where you're at?

Valon Xhafa

00:47:23.210 - 00:49:01.930

It's, it's actually at the end of the day it's more like just hands on experience, to be honest, because you just said there are like so, so, so many articles out there that just provide misconceptions, you know, like three products are the best one, two products upselling are the best one. I think at the end of the day it's all about.

It comes down to like just going into the trenches yourself and seeing things and measuring things and doing a lot of things. Because I come from a science background and even to this day, because we build a lot of new things in terms of AI and technology within behmix.

We read a lot of articles, we constantly see what's up there and we try to combine different things and you rarely can, let's say get the results that you see out there for your own experiments or tests or something.

So everything is kind of bubbled up pretty much to have the best results, the greatest results, you know, like to provide this traction, to provide this buzzwords, you know, as you said, like free product sound is the best one, you know. But at the end of the day the most important thing is the experience. You just have to go into the trenches yourself.

Test things yourself, measure things for yourself and be able to test new things pretty much. So that's something that I think is really important even to this day.

We read a lot of stuff out there, but not everything is like completely true and. Or completely false, you know, so you have to take everything with a bit of grain of salt here.

Matt Edmundson

00:49:02.250 - 00:49:21.540

Yeah, no, fair enough, fair enough. That's very true. You have to get into the trenches. I like that phrase. I do like that phrase. Valen, how do people reach you?

How do they connect with you? If they want to find out more about B Hammocks, if they've got questions for you about AI, what, what's the best way to do that?

Valon Xhafa

00:49:24.260 - 00:49:55.650

They can definitely go to our landing page, behemx.com they can get more information into the behemings.com or if they want to reach out to me directly, they can just connect with me on LinkedIn. And yeah, I'd love to, love to talk and see what we can build and how we can actually apply AI more into E commerce.

Not just say like, AI is going to change the world, but like actually applying AI in E commerce, which is a completely different story.

Matt Edmundson

00:49:56.290 - 00:50:00.530

Yeah, yeah, I like that. I don't want AI to change the world. I just want it to make my store better.

Valon Xhafa

00:50:01.890 - 00:50:04.050

Yeah, because you have to start somewhere.

Matt Edmundson

00:50:04.050 - 00:50:04.930

You know, you do.

Valon Xhafa

00:50:07.090 - 00:50:20.370

Just let's start step by step, you know, like baby steps first. Let's just start first, like to implement AI a bit because we cannot go to like change the world without actually even starting.

Matt Edmundson

00:50:22.300 - 00:50:29.660

No, I like that. It's brilliant. It's absolutely brilliant. Yeah. Solve the world, fix my store. That's.

Valon Xhafa

00:50:33.420 - 00:50:39.100

Yeah, yeah, yeah, yeah. Maybe I should use that.

Matt Edmundson

00:50:40.460 - 00:50:40.940

Yeah.

Valon Xhafa

00:50:40.940 - 00:50:46.060

See it on the Dynamics website if I have the copyrights for it. Yeah, yeah, yeah, sure.

Matt Edmundson

00:50:47.690 - 00:51:08.090

That's brilliant. Absolutely brilliant. Well, Valen, thank you so much for joining us. Really appreciate you being with us. Genuinely great conversation.

I'm super excited about your products. I'm really excited to see where they go and what kind of benefits they bring to people.

So do stay in touch with us and bring us the case studies because genuinely, really, really curious. Thank you. Thank you so much.

Valon Xhafa

00:51:08.890 - 00:51:13.220

Hey, Matt, thanks a lot. Thanks for having me. Yeah, take care.

Matt Edmundson

00:51:13.460 - 00:53:00.760

Absolutely. Well, there you go.

We will of course link to Valen's info in the show notes, which you can get for free along with the [email protected] or direct to your inbox if you have signed up for our newsletter. So thank you so much for joining me.

A huge thanks again to Valen and a big shout out again to Today's show sponsor, eCommerce Cohort.com head over to Ecommerce Cohort for more information about this new type of community that you can and probably should join.

Be sure to follow the Ecommerce Podcast wherever you get your podcasts from, because we have even more great conversations lined up like today's with Vallon, talking about all kinds of stuff which is going to help you deliver E commerce well, and I don't want you to miss any of them. Oh, and just in case no one has told you yet that, dear listener, you are awesome. Yes you are. It's just a burden you've got to bear.

I've got the same problem. Valen's got the same problem. We just have to bear awesomeness. Oh well. The E Commerce Podcast is produced by Orion Media.

You can find our entire archive of episodes on your favorite podcast app. The team that makes this show possible is sad. AF Bain on Josh Catchpole, Estella Robin and Tim Johnson.

Our theme song was written by Josh Edmondson and my good self.

And as mentioned, if you'd like to read the transcript or show notes, head over to our website ecommercepodcast.net where, as I said, you can also sign up for our weekly newsletter and get all of this good stuff directly to your inbox totally free. Which is amazing. So that's it from me. That's it from Valen. Thank you so much for joining us. Have a fantastic week wherever you are.

I'll see you next time. Bye for now.