6.Part 1 - Before you buy Pricing AI with Suzanne Valentine

Susanne Valentine: Foreign.

Host: Hi everyone, and welcome back to another episode of the Sweet Spot by Pricing Works, where we aim to inspire you to achieve pricing excellence through conversations with fascinating guests. We're kicking off a fresh season with an outstanding guest today. Her name is Susanne Valentine and she's the Senior Director of Pricing AI at PriceFX, a well known and, um, established pricing software company. Suzanne will tell us about her decades of experience in delivering applied analytics and pushing pricing into its next era. At, uh, PriceFX, she talks about the small but meaningful steps any organization can take today to prepare for pricing automation. And she helps us deal with the dread of getting started with it. We'll also talk about AI for pricing and agents, which are basically AI systems that can plan and take autonomous actions across tools. And if you're thinking that AI and agents are just for fancy, highly sophisticated pricing, think again. Suzanne tells us that they can be just as valuable for getting your data, foundations and basics in order. So let's dig in.

Susanne Valentine: To start with my background, I'll actually start at the beginning. I grew up In Los Alamos, New Mexico, which is a small town of about 20,000 people. And it has a great education system since it's home to a national lab. And it gained a little more exposure recently with the film, um, Oppenheimer. And my dad was actually computer scientist and he encouraged me to pursue something in science. And so I studied applied math and scientific programming in undergrad and then an advanced degree in biostatistics. And I actually started my career in clinical trials at Procter and Gamble. And about three years in, I had the opportunity to work on a project to build a trade promotion optimization system which would help brand managers figure out how much they should spend, for example, promoting Tide at Walmart or Pampers at Target, et cetera. And. And underlying the system were models of consumer demand that took into account price and promotional elasticity and seasonality and trends, and the cannibalization that occurs when prices change between substitutable products. And I found this super interesting, partly because this type of econometric modeling was evolving quickly. And I also enjoyed collaborating with the business people at Procter and Gamble to understand the questions they were trying to answer. So I did that for a couple of years and then I ended up leaving to join a small startup company called Demand Tech, which was an early cloud based price optimization provider. And I was the first statistician hired and started building up the team. Which was hard in the early 2000s because data science wasn't really a thing yet. But over the next decade or so, we grew the company, got to work with big global retailers like Walmart and Target and Sainsbury and Carrefour, and we were ultimately acquired by IBM. And then I spent six years at IBM primarily as a liaison, um, between IBM Research and the software group, finding research partners who could enhance and accelerate our innovation, which was fantastic. Then I did something completely different. For about five and a half years, I went to work at Facebook, now Meta, and I led a variety of data science teams in their ads business, which which spans Facebook, Instagram and now their messaging apps. And one of the biggest differences at Meta was the scale. They have so much data and so many customers, both on the consumer and business side. And I was very impressed by their formalized approach to data engineering and also how religious they are about defining goals and tracking progress with metrics. So I did that for a couple of years and then in 2023 I took a break from work. And during my time off I reconnected with a friend who worked at Price Effects, who are a, uh, leading provider in the B2B pricing space. So a little bit about Price effects. You know, B2B pricing is a relatively small space with respect to vendors, and I know you had Sophia Samaria on your podcast a while back and she gave listeners a great lay of the land for B2B pricing. We essentially help businesses set prices across their entire pricing waterfall, from reference or list price through on invoice and off invoice discounts all the way to their net price. And we provide extensive and flexible capabilities for price setting and quoting, as well as optimization capabilities. And over the past few years have really enhanced their product suite with a variety of AI capabilities which allow customers to query and interrogate the platform using their favorite LLM and define and deploy agents that can help monitor business opportunities. Similar to our peers, we tend to focus on a couple of key segments, namely distribution, discrete manufacturing and process manufacturing across many verticals, such as automotive, chemicals, high tech, healthcare, building products, and food and beverage, to name a few. Price FX is based in Europe, actually Prague, and many of the original customers were smaller European companies. But over the past five or so years, Price FX has greatly expanded both in terms of geography. I'd say about 70% of our customers are in the US now, and also customer size. So we're fortunate to have a number of very large and well known global brands as customers. One thing I'll highlight that's fairly unique to Price Effects beyond our excellent AI, of course, is that we're a of part partner first company for implementations. So we work with most of the big strategy consulting firms and some smaller boutique firms focused exclusively on B2B pricing, including pricing works, and we work with many consultants in between. And this gives our customers lots of options in terms of getting assistance with both strategy and change management, as well as really flexibility in controlling their implementation costs. So to close out my career story and tie this together, Price Effects was looking for a senior data scientist to consult on some projects. And I was actually looking for an outlet where I could hone my new Python programming skills. So it was a great fit. And after a year of getting to know the team and the culture and the tools, I took on leading advisory services for pricing AI, which encompasses sales and delivery of AI products. So in terms of how this whole journey has shaped my perspective, I've really gotten to see the evolution from static to dynamic data and models. And I'm thrilled that today's pricing teams can take advantage of a wide range of AI models and algorithms to provide insight into their businesses and customers. Having worked with both global enterprises like Walmart and small businesses on the Meta platform, I recognize that pricing innovation must really adapt to different scales of operation while maintaining its effectiveness. Uh, and this flexibility is something we really strive for within Price Effects in how we build our platform. And then finally, I've seen again and again the importance of being able to translate complex algorithms into tangible business outcomes. Pricing innovation isn't just about technical sophistication. It's really about creating meaningful business value that users actually adopt and trust.

Host: Right. Wow. What a track record. And I'm so happy you decided to join the dark side of pricing.

Susanne Valentine: And I couldn't stay away.

Host: I understand that. I myself haven't switched from investment banking to pricing. I think first, it's absolutely great to have you on. It's an amazing introduction. I have so many questions about your career and everything you've learned and absolutely resonate with what you say in terms of the tangible business outcomes. Right. I think a lot of our listeners, that is what they're interested in, they want to solve real problems that they have. Before we jump into how they can do that with technology that is available today, I just want to say, because I'm sure I'll make this mistake a

Susanne Valentine: lot, how cool it is that the

Host: names of the companies, that of our two companies both include the word price in it. So price pricing, pricing works and price updates, Branding.

Susanne Valentine: I know.

Host: Can you tell that we're into pricing? I think so. I think that's Pretty obvious because I think it was more than once that I, you know, wanted to say pricing works and I said price effects and the other way around. So just to avoid confusing our listeners, we can clarify that. But wonderful introduction. I think the world of pricing AI is amazing, but I think it's also, or it can be daunting for some people. There is a lot of talk today about optimization in general within pricing, within sales.

Susanne Valentine: Right.

Host: We all want optimal outcomes, we want roi. But I think it can be difficult if you are in the C suite for pricing, for sales, for product management. Can be sometimes difficult and overwhelming to figure out where you are on, um, that pricing and sales journey and where you should go next, how can you future proof your pricing and what kind of tools should you be using? And I think there's a lot of different questions that we go through as we go through that process. If we take an example of a business that is maybe the furthest away from AI, so maybe they're using Excel today to manage their pricing, Right. So some of the basic questions that they'll be asking themselves is one, do I need a pricing software? Although, ah, to some of us it might seem like, oh, of course there are still a lot of businesses, especially in manufacturing, who are using Excel for pricing. So they might be asking themselves that question. If the answer to that is, yes, I need software, then which software do I need? So which provider do I go for? What kind of capability do I need to look at? And then also, and maybe this is an unusual question to ask the head of pricing AI, do I really need AI and how do I need to use AI? And things are changing so quickly. There's new tech coming out, what do I actually need? How do I figure out what I need, where I should start and what my journey looks like? So I know that's a lot of questions packed into one, but how might someone start tackling these? Can you give us a little bit of guidance on that?

Susanne Valentine: Of course, yeah. So, you know, we get the question of where to start all the time from prospects and customers and the first thing we tell them is take a deep breath. You don't have to solve everything at once, and frankly, you shouldn't try to. And the companies that we've really seen succeeding with price transformation are the ones who start with an honest self assessment of where they are today and not what the industry heights tells them they should be doing. So let's talk about that spectrum that you mentioned. Yes, Excel is not the most sustainable pricing platform for all, but very Small companies. But that doesn't mean you need to jump straight into optimization on day one. In fact, I'd argue that's actually the wrong approach for most companies. So think of it this way. If you're doing everything, ah, in spreadsheets today, you're probably dealing with version control nightmares and limited visibility and pricing decision logic that lives in someone's head rather than in a system. And the first step isn't let's build a complex optimization model. Let's get our pricing data centralized and governed and transparent. And that alone is transformational. For many companies, it's important to figure out where to start with some sort of assessment. At, uh, price effects. We spend a lot of time in the early stage just understanding where a company is on that maturity curve. And we often work closely with partners on this assessment piece. Some companies are still pretty reactive, so they're making pricing decisions based on gut feel or simple cost plus algorithms. And others are more informed. They have pricing analytics but limited systematic processes. And then you've got companies that are genuinely strategic with mature pricing functions. The key is that where you start should match your maturity level. So if you don't have clean data or clear pricing principles yet, jumping straight into AI optimization is like trying to run before you learn to walk. So you'll, you'll end up with a black box that nobody really trusts and then you haven't really gained anything. And I think this brings up a really interesting question. Do you need perfect data to get started in building that data foundation? With all the headlines about how imperfect data can lead to bad results from AI, it's easy to feel paralyzed, like, well, I can't get started until my data's really good because the AI won't have a good foundation and might make a mistake. My experience is that you, you know, while you need to continually strive for the highest quality data that you can curate, sometimes it isn't until you surface and start harmonizing various data sources that you realize what really needs work or is potentially missing. So it's a vicious cycle and if you don't get started, then you'll never get started. So I'm a big advocate of getting started with harmonizing whatever pricing data you have and creating a data roadmap for improving those initial data sources and curating additional sources that will unlock additional capabilities down the road. For B2B pricing, a great place to start is with transactions, A, uh, product master and a customer master. That alone can build like an MVP data foundation. And you'll want the transaction data to capture relevant price points, such as list price, if you have it, and net price, and then also try to integrate relevant cost and margin information. And actually I said don't jump straight to optimization. But I would like to introduce agents at this point in the conversation. You know, while there are of course many sophisticated insights you can get from agents, they can also be quite valuable in the data discovery phase. So for some of our customers, agents are being used to identify inconsistencies and unusual trends that may or may not be real. For other customers, agents can actually help them with setting rules and strategy. So we've essentially at uh, price specs built. We've become agent first in our implementation methodology. And one side bonus is that many of our prospects have budget for AI experimentation when it's harder to get budget for like full blown IT pricing implementation. So by starting with the deployment of a few agents, we can actually help businesses better understand their data and potentially make the business case for a broader project.

Host: That's amazing. And it's so interesting what you've said about don't think that, you know, agent is probably the most sophisticated thing that we have today, but don't think that you need to be ready for an agent. As in the agent can help you get ready. So it's coming uh, at that level and maybe just a follow up. So a lot of the times we also at pricing works, work with clients to assess where they are on that maturity journey. And data is obviously a challenge, we all know it. But even getting different data sources to talk to each other.

Susanne Valentine: Right.

Host: And connecting information across data sources that can, you know, that can be a challenge. Sometimes we can't get cost data or it's not clean or it doesn't add up. So that kind of reconciliation is something that takes a lot of time and is one of the biggest impediments or one of the biggest, I would say detractors from like high quality analysis. Right. It's where we spend a lot of time trying to clean that um, up and it's usually still not very good. Can agents help with that process as well? Can they help us if we give them different files and different information? Can they find ways to connect that? Or is it mostly just general insights?

Susanne Valentine: Yeah, this is a pretty exciting area. So we've seen cases where you can take agents and feed them a number of data sources and have them give their best intuition at what, like creating a data dictionary, for example. You know, if you maybe you have years and years of collaboration on data files and it's not really clear what some of the fields mean. And you have sort of archaic labels or descriptions of the different data types and agents can actually give you a starting point. And it's, it's easier I think, for a human to be presented with like, here's how we think your data fits together and then go in and correct it. Or like, you know, help the agents clarify it. But it gives you a starting point, whereas, you know, otherwise you're, you're stuck with this blank page of how do I get started? This, this seems overwhelming. The agents never get overwhelmed, they're very confident and so they are going to, they're going to bring you something to review and then uh, from there you can kind of figure out how to iterate.

Host: That's amazing and that's really interesting because a lot of people might find it difficult to commit to a full on project when it comes to implementing a platform. But doing a little bit of work with AI to figure out like where you stand and, you know, what your data is looking like and how to kind of put it together, that could be of interest for a lot of people because basically anywhere. So if you're with Excel today, anywhere you want to go from there will require you to clean up your data and to try and, you know, get together and make sense of it. So that's going to be the next step for a lot of companies. Right, really interesting. We don't have to answer this now, but I think a lot of our listeners when hearing this, they'll wonder the kind of budgets, et cetera, that they need. So you know, what's the kind of minimum investment that I need to get started with AI agents? Again, we don't have to answer it, but if you've got something on top of your mind and you want to float that, what would that be? Roughly how would that compare to a full on implementation?

Susanne Valentine: Yeah, so we, I'm not sure I want to quote exact numbers, but we're talking a couple thousand dollars, for example, to get some agents up and running with some of our customers. We will do a pilot where they bring us a couple of data sources, we integrate them into the platform and we can get some agents up and running and they can figure out where they want to go next from there. So it's really, we're talking small budget basically and some data to get started. Whereas historically I think a pricing implementation took, you know, months if not years to even plan out and then, you know, obviously a much larger budget to execute.

Host: Right, that's wonderful. I think A lot of people dread that very first step of like taking a hard look at their data and trying to make sense of it and then explain it to a software provider or giving it to a software provider to make sense of and, you know, to try to work with that. So, so I think that that's super valuable. Okay, so to the question of do I need pricing software? I think we're pretty much aligned that.

Susanne Valentine: Yes, you do.

Host: Yes, you do.

Susanne Valentine: Good.

Host: And then, you know, you said, you know, AI agents, a good way to start doing a little bit of work around this. What are the other options? What are, uh, the other ways you could go about this?

Susanne Valentine: Yeah, well, so, so agents can help you get started. But I think a really important part of the foundation is rules. So, you know, people get excited about optimization, but optimization that doesn't take into account business guardrails isn't, isn't particularly useful. So I'm, um, actually a huge advocate for starting with rules based pricing of some form, even if your end goal is sophisticated optimization. And here's why. If you build that rules based system, it forces you to articulate your pricing logic explicitly. You know, what are your markup rules, how do you segment customers, what's your competitive positioning? And how do you respond to cost changes? And when we work with customers on this, it's kind of a diagnostic. We help them codify their pricing principles. And honestly, that process usually reveals gaps or inconsistencies that they didn't even know they had. And maybe they say they have customer segmentation, but when we try to write the rules, we find that it's actually very ad hoc, or they think they're pricing to value, but the rules reveal that they're still just doing Haas plus with some sort of negotiation on top. You know, when I think back to my days at Demand Tech, we definitely learned the importance of rules early on. We built our optimization solution first with all kinds of fancy math and assumed our customers would adopt it because the science was so cool. But we quickly realized that our optimizations needed to operate within the feasible space available once business guardrails were applied. And over time, we actually made rules based pricing the best practice starting point so that we could better understand the true opportunity available for optimization once all of the business rules were in compliance. So one great thing about a modular solution is that you can start with something as simple as price setting, you know, getting your prices published consistently across channels and then layer on more sophistication over time. You're not locked into a massive implementation where you have to boil the ocean right out of the gates. You can get something valuable up and running quickly and then add to it over time. My recommended progression is really data agents to do some data discovery and help with that harmonization, uh, rules and then optimization and, and then more agents. And you, uh, know, once you've implemented some basic agents and have the rules based pricing working, and by working I mean, you know, your team's trusted, you're seeing business value and your data quality is really solid, then you can start asking, where would optimization have the biggest impact, given how I do pricing. So for example, distributors have many products and customers and offer different deals depending on the situation. So cleaning up and optimizing the floor target and ceiling deal pricing quarter may yield the biggest value for them. And this type of optimization, often referred to as negotiation guidance, can be applied to multiple objectives such as margin and discount depth. Um, and an interesting, uh, extension of that is to also look at likely win rate at various price points. So getting this analysis requires that quotes are classified into wins and losses. But for many businesses, that extra step is worth it if they can understand both margin opportunities and win probability. So that's one example where we've taken into account the business process, the guardrails and applied optimization on top of it. On the other hand, there are some businesses that don't really offer deals, but are very focused on getting their list price right in an environment where they may have a lot of competitors or there are all sorts of cost changes going on. So right now their prices might be managed with a cost plus approach. And adding even one more dimension such as competitive price introduces more complexity than a spreadsheet or other small tool can manage. And this is a classic problem for optimization. And modern approaches can take into account dozens or even hundreds of business guardrails as constraints and optimize for multiple objectives. So, you know, two very different use cases, but also ones that we see commonly in our space. And then the great thing about optimization and agents is that they can have this symbiotic relationship. So agents can help clarify where the biggest opportunity is for optimization. And then once that optimization has been set and deployed, the recommendations from the optimization can be monitored by the agents. So one of the things that agents can help you understand is compliance. So businesses, you know, really need to understand whether the optimized prices are actually being leveraged in practice by the sales team. You know, if you're optimizing your prices coming out with these new guidelines, but they're not being adopted, it's crucial to investigate why not. And these conversations could yield additional context that needs to be fed into the optimization to make it worthwhile. And you could also have agents, for example, that are monitoring your competitive data or general market data to understand when constraints should be adjusted in the optimization. You could even have agents that automatically adjust constraints and then run several scenarios on your behalf. So this is the more sophisticated application of agents. The agents understand your base scenario and then they make changes to the constraints as market conditions change, run the scenarios for you and come back with several options, and then the human can actually decide which one they want to implement.

Host: That sounds amazing. And I mean, there's clearly a lot that can be done in this space. And before I ask you follow up questions, I think it's worth perhaps clarifying some of the semantics around this and some of the terminology, because different people might have different understanding of what we mean by optimization. What an agent is, what rule based is. Right. And I really love what you said about rules based. And I'm going to give you a parallel with how we manage our maturity assessments, which is in very plain and simple language, we ask clients to articulate how they manage pricing today, uh, and what are the principles that they follow consistently when they set prices, when they set discounts, when they set rebates, et cetera. And so usually we start out by being very confident that of course we know what those are because we've done it for years, et cetera, it's well established. But then the more we dig into that, the more we realize that either we have a very different understanding of the principles and we're all kind of doing the same thing, or we have the same understanding, but we don't apply it consistently and there's no way to track how we apply it. So it's a little bit of, uh, a mixed bag. The great thing about rules based. And would you say rules based is kind of the mvp, the smallest kind of smart pricing type of smart pricing that you can do once you move on to pricing software? Is that how we should understand?

Susanne Valentine: Um, yeah, I mean, I think, I think rules, rules based pricing, you know, I have this background in, you know, applied math, and you would think that I would be pushing hard for optimization advocates, but it helps. I think articulating the rules and documenting the rules which really represent the business guardrails, is foundational to the pricing strategy. And here, like I loved what you said about, you know, everybody thinks they know what the pricing strategy is and what the principles are that they should be applying. But if it hasn't been really formalized and documented, people may have different interpretations of it. So I also liked what you said about, you know, sitting with users to understand what they do today. You know, we call that a day in the life. And it's so valuable just to watch what what people are doing today and their thought process and how much of that may be captured formally in data and how much of it is just in, you know, the heads of the most experienced pricers. And only by truly understanding what customers are doing today and then also helping them to document the pricing principles that are going to be the foundation going forward is really crucial.

Host: If you've enjoyed this first part of our interview with Suzanne, don't forget to check out part two where we pick up the thread and take it further. And if you have any questions or comments for us, we'd love to hear from you. So please go to pricingwords, uh IO and get in touch. See you next time.