Claus Lauter [00:00:00]:
Welcome to episode 295 of the ecommerce Coffee Break podcast. Today we talk about new AI pricing technologies that can streamline your revenue growth. Joining me on the show is Simeon Lukov, founder and CEO of Dynamic Pricing AI. So let's dive right into it.
Voice over [00:00:18]:
This is the ecommerce Coffee Break.
Voice over [00:00:23]:
A.
Voice over [00:00:23]:
Top rated Shopify growth podcast dedicated to shopify merchants and business owners looking to grow their online stores. Learn how to survive in the fast changing e commerce world with your host, Claus Lauter, and get marketing advice you can't find on Google. Welcome.
Claus Lauter [00:00:43]:
Welcome to the show, and welcome to another episode of e commerce Coffee Break. Today we want to talk about pricing, pricing technology and how AI can help you with that. And we will try to find out or try to help you to take out the guess word out of pricing. Now, finding the optimal pricing is very, very difficult, and it's a very complex topic. So with me on the show, I have Simeon Lukov. He's the founder and CEO of Dynamic Pricing AI. Simon is a three time founder with two exits in 15 years of ecommerce. He began his career in consumer electronics.
Claus Lauter [00:01:15]:
He discovered the importance of pricing optimization. And spinning off that, he expanded a competition intelligence service to 37 countries. In his latest venture, Dynamic Pricing AI. He provides AI models, pricing policies and tools. And we want to dive into that right now. So let's welcome Simeon to the show.
Claus Lauter [00:01:33]:
Hi.
Claus Lauter [00:01:33]:
How are you today?
Simeon Lukov [00:01:34]:
Hi, Claus. Happy to be in the show. I'm great. How are you doing?
Claus Lauter [00:01:38]:
I'm very well. Pricing is a very complex topic. So while I was preparing for the show, I found out how many bits and pieces fall into that. And you're working on the pricing sector and finding the best prices for a product for a very long time. Now, what's the most common error you see merchants do when it comes to pricing?
Simeon Lukov [00:01:56]:
There are basically three things that merchants are thinking when they want to find the right price. First of all, if they have new product on the market, they want to test different prices. So they don't know what works in terms of pricing, and they use some kind of price testing solutions. Then when they know their pricing, probably they want to optimize for either profit or more sales, more revenue per category or on a story level. And the third thing is often they want to find competitors to track their prices and to see their promotions. How is going on in the competitive landscape?
Claus Lauter [00:02:44]:
Let's start with competitors. I found this very interesting because there are some areas in business, like airlines, like hotel booking engines, who are on the forefront on offering you the best pricing or their best pricing in the moment. Now, a lot of merchants, specifically small and medium enterprises, trying to do competitor research on a manual level. So from time to time they go to a competitor and see what's happening there. Obviously this is not the best way to do that. How do you do competitive analysis when it comes to pricing so that you're always up to speed what's happening in the market?
Simeon Lukov [00:03:16]:
First of all, we have scraping engine that is taking all of the data from particular website, scraping, promotions, availabilities, prices, descriptions of the products, and then we're trying to find identical or similar products. So identical products is a bit more easier. And it works for some industries, but most of the time, let's say for fashion, you cannot find the same product. And then you need to use new techniques to get descriptions of the products. And for example, use LLMs to find similar products based on the description and show, for example, white holidays dressed and their prices, their promotions, their positioning, all that stuff.
Claus Lauter [00:04:08]:
Now, AI obviously helps with that a lot. And with dynamic pricing AI, you came up with a solution on helpdet. Talk me through what AI can do for you as a merchant when it comes to pricing.
Simeon Lukov [00:04:20]:
I started with price testing, I will continue with that one. So, first of all, many people are using old techniques like a B testing to test several prices. But right now we have new frameworks that are coming and it's more efficient to use techniques like multi arm bandits or bandits, some kind of fast reinforcement learning, so that you're not splitting the traffic on 50 50, but allocate more users to the price that is going toward your objective, either revenue or profit. So you are not spending so much while you are testing. So that's the first thing that AI is bringing as a new technologist. On the second thing, for price optimizations, there are new techniques. In the old days, people need first to get data, to have historical data to fit that data, and then to see how the demand is going to make a forecast and on top of that forecast to make some optimization. Now, these things is made in one step.
Simeon Lukov [00:05:27]:
You do both demand forecasting and optimization in one step, which is huge improvement. The third thing is the help of large language models which can find similar products so that you know the competitive environment easily.
Claus Lauter [00:05:44]:
Let's talk about stock optimizing. Just talk me through how does that work, which kind of different factors are considered when it comes to optimizing the price on your stock levels?
Simeon Lukov [00:05:55]:
Let's say you are selling winter jackets and you have like four months period to sell your goods. So in the beginning of period you have different pieces on stock. Like you can have 20, you can have 200 pieces. And you put this inventory data within the model and several prices so that the model every day optimize based on that time horizon and that time is going on. And the model knows how many pieces have been already sold. So if the sales is going well, the model will be conservative. It will try to sell them on a higher price to get more profit, more revenue. But if the sales is not going well, the model will try to lower the price to be a bit more aggressive so that you don't need to think about changing prices, you just need to set up the model.
Simeon Lukov [00:06:50]:
And up on top of that you can include additional data points. You can tell the model what's your spend for marketing for another category, so that the model is getting all the expenses. The model knows how many pieces you have and try to find the optimal policy towards revenue or profit or mixture.
Claus Lauter [00:07:12]:
You mentioned the marketing aspect for it and I'm a marketing guy so that's always interesting for me. It's like obviously the prices for your meta, Facebook, Instagram ads, Google Ads are going up and down also like demand. Are there other factors that you can slot into the calculation when it comes to marketing, like your website or other marketing?
Simeon Lukov [00:07:33]:
Sure, you know it's difficult to get the marketing for an individual product, but you can put the marketing for one category or for another category. You can put also the spend for storage. So if you are selling refrigerators, this will take huge space. But if you are selling some small stuff, it won't cost a lot. So you are giving the marketing spend, you are giving the storage spend. And the models knows what's the cost of having this product and based on that decides what's the policy, what's the price that will optimize for profit or for revenue for more sales.
Claus Lauter [00:08:16]:
You're doing this for a very long time. You're working with some big companies. Give me an example of what kind of results do your customers see. Maybe a real life example before and after implementing price optimization.
Simeon Lukov [00:08:28]:
I can give you an example with pharmacy company. They used to have six stores when we started and in one year and a half they went in top three pharmacy in the country, only having good pricing, good process for that one. So they had like 30,000 orders per month. So 1000 order per day, which was huge improvement for them from one side. You can see the improvement shortly while playing with tactical tools. But the good thing is that pricing is not you are not doing it day to day, but it's a strategy. You decide in advance if you are going to be like discounter or play high low or play with dynamic pricing. This is long term decision.
Simeon Lukov [00:09:23]:
And after that, when you decide which path you take, the policies, the structure and the entire process is set up and you more or less are crushing the competition. So everybody who does not use systematically pricing is getting someday out of the game. So that was one example. The other example is when people are using us for making promotions, they're reporting that their sales is increasing by 15% or more. That was the number. And they're saying that after the promotions, people are still willing to buy the product. So it's having the huge impact. So automated promotions is working.
Simeon Lukov [00:10:09]:
You don't need every day to decide what to promote. There are tools out there that can decide instead of you. And after that, when you see how some products are selling on promotions without human interactions, then you can put human that pay a little bit more attention, put more marketing money on that and the product can skyrocket.
Claus Lauter [00:10:33]:
Hey Claus, here, just a quick one. If you like the content of this episode, subscribe to the weekly newsletter at newsletter ecommercecoffeebreak.com. I score and create 50 news sources so you don't have to saving your hours of research. Grow your revenue with ecommerce news, marketing strategies, tools, podcast interviews and more, all in a quick three minute read. So head over to newsletter ecommercecoffeebreak.com to subscribe as said, 100% free. Also you will find the link in the show notes. And now back to the show you touched on. You don't want to give too much power to the AI assistant.
Claus Lauter [00:11:05]:
And you also said there is a strategy behind that. So how do we bring this together? How do you make sure that your pricing AI assistant does not go crazy and start selling underprice? Basically, how do you teach them the strategy behind your brand, behind your company? And how does the implementation look like?
Simeon Lukov [00:11:25]:
After having the data from the model, we always put business rules like margin guards, like roundings, like shipping costs that needs take into account. So the managers are always sure that the model will not undercut below some margin level. So on the recent days, it's this trend that you go on pricing competition and when you reach this minimum level, you can go a bit higher. So you make like a loop. And if there are guys on the market that are following you, they're going after you and they do the loop with you. So imagine you're selling headphones and everybody's undercutting with one cents or two cent. So everybody is undercutting. And usually people squeeze their margins, so one of them is rising the price within the night when there is no demand and everyone is following the guy.
Simeon Lukov [00:12:30]:
So that's another trick that managers are using not to undercut and to be sure that prices are in a range that they're comfortable with.
Claus Lauter [00:12:39]:
Some people fear AI. I see AI like an empowerment tool for a business. It just streamlined things, get more data results, or you get more results from the data that you have. So how long does it take to train the AI before it really can start optimizing prices?
Simeon Lukov [00:12:57]:
If we can pull your orders in for like two, three, six months, we can train the model on historical data. And when we have that, when we have your orders, you can immediately decide how to play for some categories. So you can either play for more sales or for more profit, or to make a mixture. And when you are giving good input to the model, some reasonable goals and targets that you want to achieve for the next period, the model is taking into account this target and is giving you back the prices that will make this target, this revenue or that profit. So having the data, it's quick learning. When we don't have data, it very much depends on the traffic on the page visits that you have. So let's say we put several prices for a new product and depends on the visits. If you have 100 or thousand visitors on the page, the model will change the price based on the conversion rate and will decide what price worth more for your objective.
Simeon Lukov [00:14:10]:
And it can be adjusted via metadata, I mean parameters. If the model should be aggressive or a bit conservative, so its experimenting stage could be regulated. So if you want to see if you can sell on a higher price, and the model tries to prove that one on first date, so it will then sell the product on a high price and you get nice revenue. But if you want to be sure that your experiment is kind of conservative, you need to spend several weeks to test several prices. You can go with a bit conservative policy. The model will still find depends on the market, depends on the goods. The model can learn quick or be kind of conservative based on the parameters of the model.
Claus Lauter [00:15:07]:
Talking about markets, are there any specific industries, niches where it works very well? Or on the other hand, are areas in business where dynamic pricing might not be a good idea?
Simeon Lukov [00:15:19]:
I think dynamic pricing is working very well in the fast moving consumer goods, grocery, online grocery, it's working very well. The models are working well. When you have some idea how to execute your strategy in terms of pricing, revenue, profit, and let's say you want to be affordable for some categories and you want to be profitable for some categories. So imagine you want to have transaction people to come on your store and speak that you are affordable. You need the model. Many people have that in mind. Managers want to have such policies, but they could not execute it optimally. So this is where the power of the models are coming.
Simeon Lukov [00:16:09]:
You install the model, you give some data, and the models are recommending price, so you can take the prices, think for yourself and change your store. It's not a black box. So those kind of models that we are using are very transparent. So you know everything. You have several prices, a range of prices you're giving to the model, asking the model what would be the next best price, and the model is giving that one for you. And the model knows the context. So if you are giving competition, data marketing, data seasonality, stuff like that, the model is taking into account all that data points and helps you not to be biased people. We are kind of biased.
Simeon Lukov [00:17:01]:
Sometimes we think that someone might be competitor, but it actually does not hurt ourselves. This is where the models are showing us where we are biased and give us some new ideas.
Claus Lauter [00:17:14]:
Yeah, I like the idea that you have a neutral assistant who's not biased and gives you the facts and not the gut feeling on your pricing structure. Let's talk a little bit about the onboarding, the installation, and what does the day to day life of a merchant working with your tool look like?
Simeon Lukov [00:17:33]:
Depends if you have custom store, or let's say Shopify, Magento, salesforce, cloud commerce, or prestoshop. That was the drivers that we have. The time is different. So with custom shops, it might take two weeks to get orders from your system to see how you are positioned in terms of profit and revenue. But for Shopify and the other commerce providers, we can immediately take hundreds of thousands of orders and see where you are, how you are positioned. And then we're making this actionable, goal oriented dashboard. So like week two, you can see what the models are recommending to you, what actions you may take based on the model. So in terms of implementation, I may say this one is very quick.
Simeon Lukov [00:18:28]:
Regarding the competition, intelligence is a bit slow. We need to gather information that might take a bit more time, like a month. But if you are only in pricing optimization, that's kind of quick.
Claus Lauter [00:18:43]:
Okay, on that. You mentioned a couple of numbers before. Who's your perfect customer? What's your perfect size? Of a customer, I will start who.
Simeon Lukov [00:18:52]:
Is not the perfect customer. So we are not for dropshippers, we are not for garage based ecommerce guys, but more for established businesses. The one that are making like eight, nine figure digits in revenue, the guys that have transactions, they have solid customer base and they want to grow. So that's our ideal customer. We're not focusing on big enterprise, but something in between.
Claus Lauter [00:19:27]:
Who's the one working with the system? Is this the marketing manager within the organization who uses the system?
Simeon Lukov [00:19:34]:
Usually those are the sales team, the guys that are selling the goods. That might be product managers, that might be decision makers that are planning strategically what to do. It might be product marketers, because they can see similar products on their own. And we save a lot of time because without having automation, they have to click and try to find local or other products similar to theirs. So they are also using the system. Also, if someone is pricing on an international level, they are setting policies for their subsidiaries in the other countries from the headquarter level. So mainly pricing and marketing guys and the strategy guys within the store.
Claus Lauter [00:20:27]:
Okay. I think these guys probably have much more experience when it comes to pricing than I have because I see that pricing is a very complicated topic. There are so many elements in there, and a tool like yours definitely helps in that. Before our coffee break comes to an end, is there anything that you want to share with our listeners that we haven't touched on?
Simeon Lukov [00:20:45]:
We will have a release of our compute unit that will come in two weeks, and we're partnering also with IBM for that one. So we'll have like cloud unit that is pulling in data, calculating new prices and sending back to you based on the model. Small but efficient engine for store owners. And we think that that will be the thing that we will focus on next period.
Claus Lauter [00:21:17]:
Tell me a little bit about the pricing. What needs a merchant need to calculate with when they want to use dynamic pricing?
Simeon Lukov [00:21:24]:
So first thing is how many products you are selling, then how often you want to reprice. We have people that are pricing one twice per day, and we have others that are pricing like 30 times per day very quickly. That's the second thing. The third thing is which model are you using? Are you using price testing or some of our AI models? So the cheapest is $29, the most expensive one is $640. Depends if you put more dashboards, analytics, the price might go $1,000 or more per month. And that's for the pricing. For competition, intelligence depends on the scope. It might be from like 20k up to 100K, something like that per year.
Claus Lauter [00:22:19]:
I understand you have a Shopify app. Where can people find more about that?
Simeon Lukov [00:22:23]:
Yes, we listed the app several months ago. It's called Price Explorer AI. It's only price testing. So right now several conversion rate optimization guys are using the models and several shops, some Shopify plus guys are using the tool to test price. And we're going to put some more models inside and it's work in progress.
Claus Lauter [00:22:52]:
Okay, where can people go to find more about your solutions?
Simeon Lukov [00:22:57]:
People can find more about us on dynamicpricing AI. We have documentation for all the models there. They can even try some of the models, see if they're working or not, which is the first thing. Clyde, if you don't know if dynamic pricing is for you, you can make a simple test like having static prices and having dynamic pricing. You test that one. If dynamic pricing your path, you are welcome.
Claus Lauter [00:23:26]:
Okay. I would definitely recommend that to go to your website and figure that out. Pricing is so important. And we mentioned before, a lot of people do this by just their gut feeling and being biased. And I think there is a lot of potential in doing it the right.
Simeon Lukov [00:23:41]:
Sure, sure.
Claus Lauter [00:23:42]:
Thanks so much for your time, Simon. I think that was a very interesting take on how you can optimize your business, and I hope a lot of our listeners will go to your website and check it out. Thanks so much.
Simeon Lukov [00:23:52]:
Thank you, Claus.
Claus Lauter [00:23:54]:
Hey Claus here. Thanks for joining me on another episode of the e commerce coffee Break podcast. Before you go, I'd like to ask two things from you. First, please help me with the algorithm so I can bring more impactful guests on the show. It will make it also easier for others to discover the podcast, simply like comment and subscribe in the app you're using to listen to the podcast. And even better if you could leave a rating. Thanks again and I'll catch you in the next episode. Have a good one.