Digital Commerce and AI – Chris D’Arcy, EcomPricer

This is Textron TV. That's a great pleasure today of being joined by Chris Darcy. Chris is founder of Ecom pricer.

Welcome Chris. Awesome. Thanks Mitch yet we're talking with you.

Tell us a little bit about you and also about your company on pricer. Cool. Yeah, so I'm the founder of Ecom price and what essentially we do is we use AI to manage discounts on online stores automatically on a user by user basis.

So what essentially that's allowed for us to do is create a method to not only allow stores to tailor discounts to customers and then get higher profit, you know higher profit out of that. But all so ensure that customers who really need discounts and are unlikely to convert without them do get those discounts. So it feels like a really nice application that works out well for everyone.

I'm curious. Is it about getting the right discounts to the right people or is it getting the right products to the right people at? The discounted price or both the right discounts to the right people is really what we focus on so essentially we you know, our product is an AI that just manages the dishing out those discounts to the right people.

So when you use it comes in essentially a call is made to our service and then it goes and decides based on who showed up what discounts needs to be applied. In order for them to be most likely to convert, you know, and in practice what that ends up meaning is that uses with less means or who are going to be less able to afford a product end up being given discounts and those who are able to afford it without end up paying full price of full price and what that essentially does it boosts. We're all conversions and profit margins, but also in shows that people who otherwise would be priced out are able to purchase.

Okay, very good. I know from your background, you know through your your studies your PhD with Alan Turing Institute machine learning Behavior economics and and your exit to do a startup or do a company tell us. I've said AI is an important part of your whole career, you know from your educational background studies and then into to business and being an entrepreneur tell us a little bit about AI is a study, you know, why you chose that I feel to go into and then the role it's played and Ecom pricer.

Well, I've always been you know from when I was very young. I was always really interested in the in financial markets and how unpredictable they were. Yeah, and so when I went through to start University, I did mathematics and what essentially I found there was that, you know, there was a lot of interesting stuff in that but not a lot of it could be applied in the fields.

I really wanted it to so there wasn't a lot that I could do initially that would really work with stochastic systems and random systems and all the noise and heavy-tailedness that you kind of encounter day today. And as I pushed more and more towards statistics and then through to mathematical modeling it kind of gradually progressed towards AI is being more the venue where you know the best problems and the best tools for tackling those kind of real world applications with that healthy dose of Randomness sad. so essentially yeah being able to merge with behaviorally come did allow for you know, some really nice interaction with people who at the end of the day are very unpredictable, you know, and do have a lot of interesting behaviors, so Essentially it was there was always a wealth of tools but it was finding the tools best suited to problems in the real world.

Mm-hmm. Fantastic. What a great path timing is everything right?

Oh absolutely really Advanced and machine learning is kind of brought it to the Forefront for a lot of it. So, oh, yeah, I mean being able to just build a neural network on your laptop and set that going wait, you know, that's only really been capable. It's only really been something you could do within the sort of best five years of my undergrad, you know.

And I'm really very thankful for that, you know without the ability to do that and to experiment I really would not be here today. Fantastic. Well talk a little bit about econ pricer and the role that AI has played in with your formation of the idea of what you're doing and then and then the execution of it.

Oh, okay. Great. Yeah.

So essentially I left my PhD to stop say good which was a Oh still is still functioning it too. Yeah feeling work say data science and AI consultant business and so largely what I did there was you know, it was a lot of intake work so building out systems that could you know forecast default building out systems that could you know, algorithmic trading systems were a big thing. We actually still doing you know that takes up still a lot of my time.

Yeah, but what we yeah, well, they're essentially trying to figure Out we are Consulting doesn't scale as well as a product offering can. And you know looking through the really vast code base we built up. What I did find was that there was a lot of stuff we had there that could be well applied to pricing decisions.

So what Christ to show what person what discount to show what person in order to maximally convert and that's something that Doesn't really exist in Industry. Now there have been there are some ancient Legacy companies this ravionics for example, which is you know, again, very old and a few kind of companies of that ilk but they launched we you know, the technology that isn't necessarily similar Tech and in practice, you know, the you know tens or hundreds of thousands of dollars a month, you know, you get a team of Consultants that will come over, you know, they're not really accessible to smes. So what we did on our side is we brought over some reinforcement learning Tech that we had that was particularly.

Well suited to these kind of tasks and we built a system that could take it that could integrate with websites and swap out discounts and prices in order to maximize overall conversions and profit. You know, so that's been you know kind of where we got to where we were the reinforcement learning bent was really big for us, you know, and again this is you know, right place right time, you know, since again reinforcement learning is only been really practical for applications in the past few years. Instead of part a little bit more.

I'm familiar with expert systems. But that kind of the learning part of it. Tell us a little bit more about that specialty part of AI.

Oh cool. Yeah. So essentially what we have here is we have, you know, a website with customers coming in it is you know, a customer will come in they'll come in repeatedly and there's lots of customers who will come in frequently your actions with the customers.

So what prices in distance to display affect the behavior in the future Behavior? Um, so and what you're really trying to do on your end is your maximizing your reward, which is your overall conversions. So you have all the pieces there necessary or you know, really a standard reinforcement learning application.

Lots of people tend to think of these problems and HubSpot does a great example of this is a great example of this actually to publish a lot of work kind of in that Realm. Where they talk about contextual multi-amp Bandits is a very specific problem. And so in our case what we essentially did was we built some scripts that were taken data, but of course not too much, you know, we do stay gdpr compliant and with then very quickly process that make decisions about what discount should be shown and then learn from what it's done.

So we have conversion data for all our clients, you know, in order to make sure that the system trains, you know, so we can link up customers what we've done and what the outcome was and in doing that at scale and building means to be able to compare customers across different clients. We built a way that not only can we for any individual plan, you know, allocate discounts effectively, but new clients who come in we can also learn from that existing Corpus and effectively allocate those discounts off the bat, you know, so really it's been taking the risk out of things has been a key goal and upset which we've done very well. Yeah.

So are you are you building kind of on a large data base of you know past history that you applying machine learning algorithms too. And now let's personalize that to the individual who's at the site or you're really just is it more of a model that you've built that you're applying to their interaction while they're on the site. So yeah the model we so the model we built was, you know, very key but also the data that we've collected is a large part of what makes us unique.

I think the hardest part Getting this running was getting initial clients because essentially we were taking control of our clients' pricing with no data and just you know while they're running very expensive ad campaigns, but we were very fortunate that you know, our first few clients have been yeah, essentially really, yeah really cool about the whole thing. They just kind of this best ones when we didn't have data letters start running and we used a whole lot of you know, we had some prayers to begin with, you know, we had things like we could make assumptions about iPhone uses spending more and we could look at historical spend to go and make estimates of what would happen. So we set that up right but once we started collecting the data, we ended up having an advantage that no one else could easily replicate it is very Very hard to convince someone to let you run a pricing model without price data.

No, but now that we've collected that data and we have a you know, a nice pool of customers new customers as they come in. We can always make predictions about how different changes will perform and what will happen and what should happen. So we're getting you know all the time.

We're getting better and better with that fast startup, you know, and in particular, we you know, we've always been confident in our ability to make more money for our clients, you know, we essentially we guarantee it. Um, you know, the ROI for product is always been guaranteed at 200% Yeah. So yeah, we will never charge you if you're not making money from us.

Mmm. You know, one of the things about pricing models even product models, they can be extremely complex themselves. Oh, yeah, we can tell setting.

Well, I'm not even just retail. I'm curious if the fact that you're using AI technology and the data that you have that might also help you take advantage of some of those pricing models and maybe doing some unique things with it with AI. Oh, yeah, absolutely.

I mean we do want to eventually we do want to kind of expand outside of you know, like largely right now. We're good for setting up prices, you know on stores and you know for online sellers, so, you know, you you very much it does fall into the quite traditional do you put discounts on a sneaker? But we also want run with clients who do things like web hosting and marketing services on subscription.

So really anything we can click by button online but in practice we'd also be Keen to try and move that we know a lot of the tools can be applied in cases where you have customer data in a physical setting. Yeah, but yeah, and that is something we want to eventually, you know, kind of shift to that would be again a really fun thing to tackle. But for right now we're kind of focusing on really dialing this in as much as we can, you know, we feel very confident on Products and now what we're trying to do is make sure that we can get the message out that you know, the product like this does exist.

Yeah. Yeah, we can you know, cause yeah, we're essentially provide a great benefit with you know, no risk, you know, you know and then trying to make sure that people at least aware of us. So the education part is honestly the hardest of all of us, you know, that's actually something wanted to ask you just speaking of Education.

What do you think? It's what's the greatest misunderstanding people have about using Ai and a product as part of your strategy. Oh, honestly people are I you know, I find once they start talking about AI they start wondering if they have to go and do stuff for some reason seems to be a big thing.

The more we talk about the technical side the more people start trying to involve their CTO or the tech team or things like that and start asking about resources and what data they can add to it how our API calls work what we've largely found is that you know, just like any other, you know statistical technique. Yeah. It's probably best if people don't really know how the sausage is made.

You know, AI does say at the end of the day that we do have some kind of cool technique up and running but We found that giving customers too much information about the details of how it works rather than the outcomes for how it works just leads to confusion. Yeah, so we really know we've been trying to minimize that as much as we can. I think based on how we do work the people do need to know that we're ai-based simply to be able to explain the fact, you know, how our billing model works for example and how we're able to do attribution vote how much the system has made them.

But outside of that, we really try and simplify as much as possible because our Target customer isn't CTO. It's the Ciara, you know, all the CFO. Those are always who we really are trying to get in front of You know, you mentioned Finance you're interested in finance in earlier in your life.

This reminds me of you know, there's a time where you know, Wall Street would someone would take a great model that they've developed, you know, there's statistician or someone who's doing analytics and create a model for portfolio management or whatever. It might be some aspect to the market and kind of build a company or an offering around that in similar way. This is an evolution of that right creating an ai-based model that leverages that and applying it with data and it doesn't, you know, fortunately the data feeds itself into your model.

You don't have to constantly, you know, come up with a model and make sure that it's correct. It'll tell you because you'll see it in action. Exactly.

Yeah, I think yeah, it does actually You know, I would say, you know doing stuff back with in that kind of financial realm. It's gotten a lot more saturated that yeah, it's and again it is, you know, don't get me wrong. We do build yo, you know trading models, you know, we are familiar with you know, just there is a lot of fun.

We've just pumping huge amounts of data into a model and then kind of letting it go But in our case, um, you know for this it's just so much more. Scalable and there's so much more room, you know for this to grow that we you know, we feel that this is you know much more fun. Yeah.

Mmm and you're very even a very measurable outcome, right? You're trying to increase the ROI and exactly and after succeeding you're not yeah, that was honestly really that that was one of the things that Honestly, I think the benefit of you know, like I must I'm a solo Panda. Yeah, so I have to kind of fill the shoes of the technical and then the you know, the more marketing side of that but what I found was that from a technical standpoint it is possible for us to You know be able to do you know statistical attribution.

So say, you know how much we've made for you with some aerobics, you know, because every time we allocate a discount to someone and we buy it's it would be disingenuous for us to say, oh we've made you the whole amount of that sale because that sale might have occurred. Anyway, when we actually have to do is we have to do yeah, we have to run small comparison groups that don't use that have the service turned on versus the service turned on and then meet estimates based on those changes between those groups to be able to say with some certain see how much money we've made people. So we really, you know, but in doing that we've been able to set up a model where we charge People based on that returns.

Yeah, which I think is a really nice model and really it should be you should be yeah should be the ideal. You know, you only pay if you've made money and in our case, if you check for example out terms and conditions you will find out billing model is in that which so it's all a lot of legal and then all of a sudden several pages of very complicated code. Yeah, but the end result is that we do a very good estimates from that side of what we are eight actually able to do for our customers.

Great. Well just a little bit about folks can go to your site. How do you engage with people?

What's your someone says I'm interested in this and is it sitting down and plugging it into woocommerce or whatever you're using or is there something they can you know sandbox on your side or some way that they can kind of get a feel for how this works. Oh, absolutely. So basically you go to Ecom price.

Yeah. com. You just you know, you hit sign up take, you know about you put an email address and a password.

And then the fall setup involves you putting two. You know JavaScript Snippets into your website one in the header, which will change the prices and collect data and then the other in your checkout, which will go and look at. Based on which customers came in and purchased.

How did they how much did they actually spend? That whole process for a Shopify takes about a hundred and twenty seconds, you know, so it's very fast and then once the Snippets are in there. You essentially agreed it with a bunch of dashboards that will give you all this data coming in, you know, so our customers always get all the data.

We collect, you know, you have the option to download that. And then you go into setup billing and give us different versions of your product discounts. So we used to run a model where the system could choose whatever discount it wanted.

Which what great but we do we you know, we found that it probably be nice at a give our customers some sense of control. So essentially you say you give a set of examples of what price what prices and discounts you would like to be available particular products and then the system swaps it out. So full setup is honestly doable and About three minutes, you know, if you're speedy and then you can kind of just you know stand away and let it run.

Yeah, it's gonna just do its thing for the however long you want it to it'll send you know. It'll invoice you eventually when you start making money and then you know, it'll send you reports on how much you've made. So we try and really keep it very hands-free.

We don't want to create a product like Google and you know, like Google ads with people are sitting down and specializing in it because the end of the day that's not helpful for anyone. You know, we really try to make this product as user friendly as possible. So I think the world needs another Google ads we're good so ecommer, I'm sorry.

com, correct. Exactly. Yeah.

Yeah. se, which is a large Swedish company for Ecom prices showing up first in the Google results. But if you Google as welcome at best now, yeah.

com and then yeah, it's easy setup. You can also contact us and we'll walk you through if you have any special, you know cases we do occasionally get people trying to set up with very high ticket items. So people sell cars and Realtors and stuff like that.

But we do actually have solutions for them as well. Yeah, and of course, yeah, no, no upfront charges or anything, you know, that's you know, we only like to charge once we definitely done something for you. Well, it proves the value or not, right?

That's awesome. Exactly. Well, thank you very much Chris great to have you joining us Chris Darby who's foundy founder with Ecom pricer and I hope you come back keep us in informed and updated as things progress.

Awesome. We'll do thanks so much Mitch. It's great.

Yeah, you bet. Thanks Chris. Yeah.

Have a good