George Boretos has over 25 years of professional experience in leadership positions in the enterprise software market, a deep understanding of AI technologies, and a successful journey as an entrepreneur launching three startups and raising $9mn in Seed & Series A funding, working with Fortune 500 and other customers worldwide. His most recent endeavor, FutureUP, brings in this experience to help enterprises make data-driven decisions to optimize pricing, improve profitability, and accelerate growth.
In this episode, George shares about using AI for price optimization, particularly in SaaS and manufacturing industries. He explains how his company, FutureUP, helps businesses analyze customer and macroeconomic data to determine optimal pricing strategies, focusing on standardized products. He also discusses AI’s role in predicting pricing for new markets and the importance of taking the first step to review and adjust prices regularly.
Podcast: Play in new window | Download
Why you have to check out today’s podcast:
- Learn how AI is revolutionizing pricing strategies, particularly in SaaS and manufacturing sectors.
- Gain a deeper understanding of the key differences between pricing in B2B and B2C markets, including negotiation tactics, price variance, and strategic vs. deal-based pricing.
- Discover practical advice on how to approach pricing in your business, including why regular pricing reviews and value-based pricing can make a big impact.
“Just start doing something. Even if this something is, okay, reexamine your prices. Not once per two years or per year, examine them every quarter of a year, at least. Not just your cost plus your margin, do something a bit more sophisticated. Ask your customers, for instance, or your partners to establish some other benchmarks and interesting price points.”
– George Boretos
Topics Covered:
01:32 – How he got started in pricing and how he initially got into AI before pricing
03:39 – The differences between AI when he started versus today’s advanced tools
05:57 – Describing his AI model and how it integrates various parameters at multiple levels
08:06 – Clarifying the common perception of AI being synonymous with neural networks as a misconception
11:04 – Explaining that for his particular model, the formulas are pre-existing and universal
12:31 – Agreeing that AI today can incorporate both internal company data and external factors
14:07 – What his company FutureUP is built out for
17:42 – How extrapolating data from one country to another can be challenging if the data sample is small
19:10 – Explaining how his company, FutureUP’s model can handle both B2B and B2C markets
20:15 – Key differences between B2B and B2C pricing strategies
22:47 – How price variance is more common in B2B though in B2C there can still be significant price differences between list and actual price
23:35 – What company size does FutureUP typically targets
24:32 – George’s best pricing advice
26:18 – Two main reasons companies don’t often prioritize or experiment with pricing
27:55 – Why change management is crucial in pricing decisions
Key Takeaways:
“I do believe, and I do agree that the negotiation part and the discounting part is more complex and more interesting and more important for the B2B environment. The list prices are almost irrelevant.” – George Boretos
People/Resources Mentioned:
- FutureUP: https://www.futureup.io
Connect with George Boretos:
Connect with Mark Stiving:
- LinkedIn: https://www.linkedin.com/in/stiving/
- Email: [email protected]
Full Interview Transcript
(Note: This transcript was created with an AI transcription service. Please forgive any transcription or grammatical errors. We probably sounded better in real life.)
George Boretos
Just start doing something. Even if this something is, okay, reexamine your prices. Not once per two years or per year, examine them every quarter of a year, at least. Not just your cost plus your margin, do something a bit more sophisticated. Ask your customers, for instance, or your partners to establish some other benchmarks and interesting price points.
[Intro / Ad]
Mark Stiving
Welcome to Impact Pricing, the podcast where we discuss pricing, value, and the quantitative relationship between them. I’m Mark Stiving, and our guest today is George Boretos. Here are three things you want to know about George before we start. In 2016, he founded a company that used AI to do forecasting. In 2018 he founded QRM that uses AI to help healthcare companies find and win new business. And last year he founded FutureUP, an AI firm that does pricing. Oh, and he rides bicycles a lot. Welcome, George.
George Boretos
Hi, Mark. And thanks for the invitation. Great to be here.
Mark Stiving
Hey, thank you. So let me start with the question. How did you get into pricing?
George Boretos
Oh, it’s a long story. Do we have enough time?
Mark Stiving
We got time. Let’s go, get started.
George Boretos
It started when I was in the MBA, in my MBA back in 1994. Our marketing professor there said a very interesting case study about Shell. Shell was at a price war back then, and they were all decreasing prices, losing money, what always happens in price wars. And there was a brilliant brand manager that decided to measure price elasticity, level price elasticity, and understood that he had some space to increase prices instead of decreasing them. And he did that. And eventually, he ended the price war. And as you can understand, they increased their market share and they ended up being the kings in the specific market. And when I heard about this, I was excited. I said, okay, so mathematics and this kind of principles in pricing really work in practice. And I decided that I wanted to do something with that. Eventually, I entered the marketing space afterwards, after I finished my studies, and then I was responsible for pricing as well. And this is how it all started.
Mark Stiving
Okay. So then I would’ve, that’s not the answer I would’ve guessed if I was guessing. Based on your background, did you get into pricing first or AI first?
George Boretos
I would say AI because I studied AI when I was in university before my MBA where I was studying software and computer engineering. So that was first, but then it was a parallel thing because I did AI things in parallel with pricing, and then I combined them with the model that I’m still using. By the way, the model that I’m using right now, I developed it before 20, 25 years. So it’s a combination of AI and pricing back then.
Mark Stiving
Nice. And so, boy, we’re hearing all about AI nowadays, right? So for the last year, we’ve got chatGPT, OpenAI, all of these companies with large language models. and it’s fascinating for all of us, but you’ve been using AI for 25 years. What’s different between what you started with and what we’re doing today?
George Boretos
Back then, it was primitive. So you had to discover and reinvent the wheel in every step of the process. Right now, it’s not like that. You have Python, you have R with all these great libraries of great tools, and now you have chatGPT, LLMs, all these great tools that are really and truly amazing. So back then you were on your own and you had to do a lot of things yourself. And this is why I developed my own model back then. If I were to start again, now, most likely I wouldn’t do that. I would just benefit from the big libraries out there with great tools. It was quite different. I would say it’s quite different.
Mark Stiving
Sorry for asking what I think could be a really hard question, but if you developed a model 25 years ago and today, you would start over using the LLMs or at least the databases that they’re all using, couldn’t we say that your model is behind the times then? I mean, how are you keeping it updated so that it’s still competitive with other models?
George Boretos
Oh, no, no, no. It’s the other way around. It’s far ahead, but I would have to spend a lot of time to develop it. Now, I have 30 years in order to develop it, to test it again and again. But if I knew then that I had to do something so big, it was a big endeavor, a big project that took three decades to be not concluded, but to a certain point, I wouldn’t have done it. It’s quite painful. but it’s much better because if you use, let’s say chatGPT or if you use any of the standard general purpose tools, usually they don’t have an econometric aspect. They don’t have business econometric intelligence, which is important, especially for pricing, but also for strategic forecasting in businesses. And these guys now, I think they are missing a lot, but again, I wouldn’t say to anyone to start doing what I did back then and invest three decades of their lives to do what I did. This is what I meant.
Mark Stiving
Yeah. Nice. And so describe your model to us, because it seems like you’ve used it now in at least three different fields, right? One was forecasting, one was healthcare, and now pricing in general. So what’s common across those and, what’s the model like?
George Boretos
Yeah, in two fields, I didn’t use it in the second startup company because they are, I invented, let’s say, completely different models because it was a different case. But my first, and this startup, I’m using the same model. So the model is, again, an econometric model that takes into consideration several different parameters at different levels. So it’s macro indicators. First of all, inflation, sales rates, GDP, all these things. then it’s market indicators, competitors, prices or whatever the market size, all these things. Then at the macro level, it’s market segments, customer segments, customer characteristics, and also product characteristics. Now, the difference with general purpose models, that might also take into consideration some, not all, but some of these parameters, is that the model understands the mechanics and how all these different parameters influence sales performance, probability to buy, willingness to pay, pricing, all these things.
So it understands that inflation is something different than exchange rates or GDP growth or than competitors pricing. All these are important, but they have a different way, different formula actually, that they affect sales performance. The model was not built in one day, obviously, so I had to build the different components, the price elasticity component, the market component, the competitors component, and all these things step by step. And then all these were glued together into one model. So this is how it works. Obviously, it’s not as simple as I described. It’s a set of differential equations, some complexity embedded in it with some statistical processing and some AI modeling on top of it. But more or less, this is what the model does and how it works.
Mark Stiving
Okay. And so I could be naive. In fact, I am naive about AI, so please educate me for a second. When I think of AI, I don’t think of a formula. I think of it more like a neural network or something that pops out by itself. Nobody programmed in the formula. Am I wrong about that?
George Boretos
Actually, yes, you are. Because this is just one part of AI. This is neural networks. Neural networks is one of the many, many methods that we have in order to do AI things. Machine learning, for instance, which nowadays is a subset of AI, has many different methodologies in order to do modeling in AI. And it’s all about formulas, even neural levels at their no level. They do have specific formulas. Sometimes they are linear in it’s more simplistic form. Usually it’s non-linear. But again, there are some formulas. So it’s not exactly like that. This is actually a usual misconception. Most people think that AI is neural networks. It started like that in our attempt to imitate human intelligence, but then it was broadened as a term. Individually, it encapsulated everything, machine learning statistics, practically everything we do to simulate things, not just human intelligence, but other things as well, or to forecast or to optimize, fall into the AI space, which is a broader space right now.
Mark Stiving
Okay. So by the way, George, you don’t know this. I have a PhD. I understand econometrics relatively well, probably nowhere near as well as you do. But if I were trying to say, okay, how does competitive pricing influence demand? I would start crafting models and testing models and saying, which models do my data best fit? and so which ones give us the best predictive results on the data that we’ve got? And so that’s me doing all the work. How is AI different than that?
George Boretos
Well, fundamentally, it’s not different, but it’s more structured and it has specific methods and algorithms behind it. But fundamentally, AI does the same thing. It tests different models, algorithms within the AI space. There are some error metrics and some methodologies to test that in and out of sample testing and so on, so forth. Or price experimentation, things like that. And then eventually you end up with a model that performs better than other models, and you have accuracy levels, error levels, and you see if you can proceed with this model or not. But fundamentally, what you described, and of course you have a PhD, so , this process, this is exactly how it works, is trial and error. And this is why I said that what I did with my model was a painful process because I had to try and see and check out things for three decades. So I cannot recommend to someone to do the same thing and wait for three decades. For me, it happened. I wouldn’t do that deliberately. It happened and thank God it happened because it was a nice thing eventually.
Mark Stiving
Right. And so if AI is going to do this, do you still have to feed it the formulas or does AI just come up with formulas on its own nowadays?
George Boretos
Not for my model. For other models, it might be like that. For my model, it’s the parameters that need to be computed again and again through model training. But the model is a universal model that I have. I customize it per case, obviously, and I might activate or deactivate some of the elements that are relevant or not relevant per case. But eventually the model is there and they don’t do anything to alter that. But in other cases, in other models or in other use cases, yes, obviously the AI could also change the model itself.
Mark Stiving
Okay. So if we go back five years ago before LLMs existed and I thought about AI this way I would’ve said companies like Pros and Vendavo, these guys are using AI mathematically. And what they use is whatever data the company has. So I’ve got win-loss data, I’ve got price data, I’ve got competitor data, I’ve got customer data, and I can pull all this data in and help it make decisions. I get it. I think the really powerful part of AI today is I can now use that data, but say, tell me about the weather. Tell me about the GDP of the country at the time, or the state at the time or the economy. And so it’s pulling in all these data sets that are not relevant, directly relevant to that company, but could be relevant to the decisions that are being made.
George Boretos
That’s a very nice point for a couple of reasons. First of all, yes, apart from the internal data of the company, you need to obviously use this and process this, but apart from that, there are all the external factors. It could be weather in some cases, it could be natural disasters, it could be just macro market indicators and things like that. In my case, I do take into consideration all these things and it’s very interesting and it’s extremely important in some cases to identify the right price or how the company goes. But there was something else that you said about the way that you interact with the software. And this is something that LLMS can do even for standard software pricing software or forecasting software or whatever. The way that we use, or still in most cases interact with software is we have a fixed menu and we select options and we get results and so on and so forth.
With LLMs, and for me, this is a roadmap by the way. I don’t have it yet, but others do have it in other fields. You can interact in a physical, natural way. Just think about the software, what you want to extract from the system. So I want to optimize prices, taking into consideration the weather, GDP, I don’t know, whatever, please process and give me a chart indicating this or that. That’s a great way to interact. And I think this is going to revolutionize the whole software market, not just price optimization or pricing software, but the whole software market, I think.
Mark Stiving
Nice. So tell me what your company does. What does FutureUP do?
George Boretos
Well AI company for price optimization, I’m targeting mostly SaaS companies, Software as a Service companies, plus manufacturing companies. The going factor between those two segments is standardized products. So the product needs to be standardized. Even if it is configurable, that’s okay, but it needs to be a standard thing, not built to order for instance or custom services, things like that. And what we usually offer is we get customer data plus macro data, all these things. We analyze it and we offer some optimum price suggestions or some what if scenarios for different market circumstances, competitors decrease their prices, what will happen with our market share or our sales. The same with macro indicators, et cetera, et cetera. In some cases it’s new product launches. In this case the system doesn’t get us input actual sales data because in the new product loans, obviously we don’t have new existing data, but we get pricing or market research data and also international expansion. That’s interesting, especially for companies which are already operating in several countries, but they want to expand to additional countries. So practically we take all the data from the existing operations in existing markets and we extrapolate using macro market indicators for the other countries to the other countries where this company hasn’t sold yet. We give some open price suggestions plus some penetration expectations, revenues, profitabilities, things like that.
Mark Stiving
Boy, that just leads me to another question that I don’t understand about AI. In the world of statistics, we would always say we can make predictions within the range of the data, but we can’t extrapolate outside the range of the data. And it feels to me as if I’m a successful company inside the US and I say, let’s not want to move that to Europe. How do you extrapolate those results from the US to Europe? That just feels really hard to me.
George Boretos
Yeah, you are absolutely right to what you say, but actually it’s within the variables that we’re measuring, monitoring, and processing. So if you are monitoring and analyzing, let’s say GDP, inflation, unemployment, macro indicators, different per countries, but not for all countries, then that’s a standard in and out of sample testing case. You have some countries, not all of them with their macro indicators or other indicators. You analyze them and then you extrapolate, obviously you do that a few times with in and out of sample testing in order to understand how accurate or not the model is. And if you feel comfortable, then you move on and extrapolate to other countries where literally you don’t have anything right now and you need to understand what is happening there. So this is how it goes, and you can extrapolate to new things, but you need to have some gluing factors, something, some type of information that is common for all these countries or whatever you want to extrapolate or whatever you want to forecast.
Mark Stiving
Yeah, so, I’ll give you the really weird example. So imagine I’ve got a software product that’s here in the US that it’s really successful and I want you to predict how well it’s going to do in France. And I don’t change the language to French.
George Boretos
Okay? First of all, in this case it is one country. You have one country and you want to predict another country. This is usually not how it goes. The sample needs to be bigger usually than the out sample. So in this case, we couldn’t do that. But if you have more countries where in some you sold the software with the local language in some other cases not, and you had this structure as an indicator and you’ve seen how successful or not the software is in different countries where the language element was there or was not there. In this case, you could extrapolate, let’s say you have 10 countries for the actual sample to train the model. And then based on that extrapolate in France, I don’t know the UK or whatever else in this case, you might be able to do it. I’m saying you might, because always you test the waters and you see the accuracy of the model. Sometimes it’s great, sometimes it’s not. So you drop it or you use a different model.
Mark Stiving
Understand, I’m just asking you really hard statistics questions, George. I’m so sorry. It just gets mad thinking.
George Boretos
It’s getting scientific. Usually I don’t have the chance to do this type of discussion, so I’m intrigued.
Mark Stiving
Let’s go back to your company for a second, FutureUP. You said you focus on SaaS and manufacturers. so I assume manufacturing is mostly B2B and SaaS as a B2B or as a B2C as well?
George Boretos
It could be both. For the model at least theoretically it’s an industry agnostic in size agnostic model and system. So it could be both, but there are many differences because between B2C, it’s fundamentally a different thing. The most important difference is what you can optimize. So if it is a B2C case, then you optimize the direct end price, which is visible to the customer. But if it is a B2B, then you have negotiations, intermediates, you have direct discussions with the customer, you have discounts, you have rebates, you have all these things blur the picture, but eventually you can do things there, you can optimize there as well.
Mark Stiving
God, I love that answer. I was actually thinking the exact opposite as in, in B2B, I need the answer because I want to know what price I should be shooting for in the negotiations each and every negotiation. And in B2C, I can kind of test prices to see what works. I don’t need AI to tell me what works.
George Boretos
That’s an interesting point and something that I’m not doing. Because what you’re referring to in the B2B environment is deal optimization, the case where you have a specific deal with a specific customer, this is what CPQ does. For instance, if it has the price optimization thing this is what my previous company did by the way, and this is not what I’m doing. I’m doing strategic pricing. So it’s about lease prices or negotiation discounts, but at a strategic level, you give guidance for Germany versus France versus whatever or versus different market segments. But I don’t give a suggestion, I mean, the system for a specific customer specific deal, how to do things, these are two different things. It’s day-to-day pricing versus strategic pricing. I mean the strategic pricing world.
Mark Stiving
Okay. Any other differences in B2B versus B2C that you can think of?
George Boretos
Yeah, the people that you discuss with are different. So, in B2B, usually you have different negotiation paths, which are more lengthy, and this blurs the decision making, therefore pricing decisions and the pricing process even more. In B2C, statistically, it’s a more robust exercise. Usually you have bigger numbers in B2B, you might have a few big customers in B2C, usually you have many smaller customers. Statistically, this is more robust and easier to statistically simulate. In B2B, I would say it’s more challenging in most cases and more customizable, you need to customize more things. In B2C also, in most cases, without talking to the customer, you can guess which variables are the most important. In B2B, it’s almost impossible. You have to make a big discussion with the customer and understand which variables, macro indicators, market indicators, I don’t know, whatever competitors pricing or whatever is really important in the pricing and decision making exercise. So complexity, I would say complexity and lengthy decision making process.
Mark Stiving
That makes a lot of sense. The other difference I would think is that in B2B you get much more price variance, especially if there’s a direct sales people negotiating deals. And in B2C, there’s much less price variance, even though you’ve got more data points.
George Boretos
You are true, you are writing that. However, especially if you see list prices in the B2C environment versus the actual prices in e-commerce places, for instance, in web and websites, you will see a difference of even 40%, 30%, which is standard for B2B. This is what you usually see when you give us a discount. So it’s not a surprise to see big discounts and differences from list prices even in the B2C environment. But I do believe, and I do agree with you, that the negotiation part and the discounting part is more complex and more interesting and more important for the B2B environment. Actually, it is the most important. The list prices are almost irrelevant.
Mark Stiving
Yeah. What size companies do you usually sell to?
George Boretos
I would say mid-sized. I mean, not very, very big companies because usually they have data scientists and departments, so internally they do things. Small companies, it’s a bit difficult to discuss pricing because they don’t have the right mindset to discuss this and understand the benefits. And so usually it’s midsize companies. Let’s say, companies around I don’t know, 100 million revenue 200, something like that. I don’t have an exact number. But I would say something like that, perhaps even a bit more.
Mark Stiving
So, somebody who has desire and data, but probably doesn’t have their own data scientist yet. You’re the outsourced data science team.
George Boretos
Let’s say. Sort of, yeah, something like that.
Mark Stiving
So solving a specific problem. Excellent. Yeah. So, George, I’m going to get to the final question here. What’s one piece of pricing advice you would give our listeners that you think could have a big impact on their business?
George Boretos
Pricing advice, not AI and pricing. Okay. I will go a bit like Nike here, just do it. I mean, the biggest problem that I’ve seen with many companies that I’ve worked for or with is that they really don’t do even the first step with pricing. They are either afraid or they don’t understand how important it is, but for any reason at all, they don’t deal with pricing as much as they should do, and they don’t prioritize it high enough. So I think that’s the most important thing to understand that this is an important part of your business. For me, it’s one of the most important elements of success actually in business. But okay, I’m biased. It is extremely important for a thousand reasons. And you need to do something. So just start doing something, even if the something is okay, reexamine your prices, not once per two years or per year. Examine them every quarter of a year, at least not just your cost plus your margin. Do something a bit more sophisticated. Ask your customers, for instance, or your partners who establish some other benchmarks and interesting price points. Start with these simple things. And I’m pretty sure that the appetite will increase internally in the organization to do more things. Eventually they will reach value-based pricing. They will reach AI price optimization. But you need to start from something. So start doing something now.
Mark Stiving
Okay. I got to tell you that. I love that answer, but I want to ask you, take your best guess on why companies don’t do pricing. Why don’t they play or test or change their pricing very often?
George Boretos
Two reasons. First of all, they don’t understand that it’s important. They think that it’s just another task that they need to do. It’s like when you have a product individually, you need to build their package. And you say, okay, that’s the last thing. It’s not so important. But if you are a marketer, you know that this perhaps is even more important than the product itself is the same with pricing. They leave it in the end at just one checkbox, another task, which is not the case. Another reason is some guys really understand that pricing is important, but they know how to start because they don’t have the skillset, they don’t have people within the organization, and the economy’s perception. Something that I had when I started with marketing was that okay, I’m a marketeer. Marketing is all about the four pieces of marketing. One of them is price. So I’m the pricing guru so I can do everything myself. This is how I started doing things in pricing back then, and I had some successful stories to share there because it’s not like that. Pricing is a big field. It’s a big premise and you need to learn a lot of detailed methodologies, approaches that are beneficial for the organization. A common marketer, as I was back then, really don’t have the skillset to do that. The same for most companies. And I think this is why they don’t start doing something really interesting with benefits for the company.
Mark Stiving
Nice. I love both your answers. Let me toss out a third one that I think is really accurate, too. And that is, so many people care about pricing. It’s just a huge political mess to go change it.
George Boretos
Definitely. Yes. I couldn’t agree more. Change management is key. Even if, I mean, it’s a multidisciplinary thing. So even if you have the best answer, best pricing, best monetization model, whatever, if you don’t have access and good negotiation, collaboration with all the different departments, you’ll get nowhere. Also, if you don’t have access to CEO and COO commitment, you won’t get anywhere. So it’s a very difficult thing to navigate yourself to success with pricing. Why? Because it affects everyone. Practically, everyone.
Mark Stiving
Everyone. Yeah. George, this has just been a lot of fun. I hugely appreciate it. Thank you for your time. If anybody wants to contact you, how can they do that?
George Boretos
I think the easiest thing is to reach out through LinkedIn. It’s George Boretos.
Mark Stiving
Perfect. We’ll have the URL in the show notes. To our listeners, thank you for your time. If you enjoyed this, would you please leave us a rating and a review? They are the lifeblood of our podcast. And finally, if you have any questions or comments about this podcast or pricing, feel free to email me, [email protected]. Now, go make an impact!
[Ad / Outro]
