The Evolution of FP&A: Building Financial Models for Success with Paul Barnhurst

Randy Wootton (00:04):

Well, hello everybody. This is Randy Wootton, CEO of Maxio, and your host of SaaS Expert Voices, the podcast where we bring the experts to you to talk about what’s happening in SaaS today and what is on the horizon. Today, I’m delighted to have Paul Barnhurst join us, The FP&A Guy, who has been doing FP&A for a long time but has come at it from multiple different venues and tracks. And we’ll get into a little bit of his background, and we’ll talk about what he’s seeing play out in the CFO role in FP&A writ large. He also is spending a bunch of time in AI and finance and how that is manifesting, so we’ll spend some time there as well. Paul, thanks for joining.

Paul Barnhurst (00:43):

Thank you for having me, Randy. I really am excited to spend some time with you today.

Randy Wootton (00:48):

Thank you. And let’s get into your background. We both started in the Navy; you were a civilian contractor in procurement at China Lake, which is a really cool base. I had spent 12 years in the Navy, had flown in and out China Lake a couple of times, but from there, it looks like you leveraged that background in that procurement into becoming an FP&A expert. You were at Amex and a couple of other gigs. And so why don’t we talk a little bit about if you would like to go back to China Lake and how you made that change from that too, I think it was Amex was your first company, and then what I’m really interested in is I love talking to entrepreneurs and people that are starting off facing the void and saying, “I’m going to create something.” And so, leaving a large company and moving into your own gig, I’d love to talk about that as well. So what are the career shifts that you’ve had, and as you look back on them, what were the inflections or the catalysts for those shifts?

Paul Barnhurst (01:41):

Yeah, so I’ll tell a few there. So I started my schooling with an entrepreneurship emphasis in undergrad, and I always thought I wanted to start my own business, but the risk side of it being very risky and never seeing a good opportunity, I’m like, “Okay, I just need a job.” A job came up. So I went to the government, which was always the last place I said I’d work out at China Lake, great experience for about four years, but pretty quickly realized a career in government contracts was not for me. So I made the decision to go back to grad school, and in grad school I was originally going to do supply chain, kind of be a business type analyst, but I’d always enjoyed numbers; I loved my finance class, and that was kind of inflection point. I switched and said, “Let’s do the finance degree,” and I ended up doing a dual degree program in my MBA, did a master of science information management as well.

(02:26):

And when I first worked for American Express, I was really doing more report writing, a lot of SQL, a lot of back-end stuff. It was a finance analyst role, but I also managed a cash forecast. With our AR team and worked a lot with our FP&A team, and an opportunity came up for an FP&A role and I really didn’t know much about FP&A, but it was a promotion like, “Oh, they’ll pay me more; I know the guy who’s hiring.” I know all the people because of the role I had. And so it made a natural fit, and I applied and got hired, and that’s how the FP&A journey started for me, and I loved that I had a background one in procurement, also a little bit of a technical, I knew some SQL, I knew report writing. It really helped with working with the data, understanding data structure, all those things that, if you don’t, it can get really messy sometimes.

(03:13):

Then the next inflection point was for me. I moved back to where I’m originally from with Amex and realized that unless I was going to New York to get promoted, I pretty much had to wait for somebody to retire because there were just no roles in Salt Lake for Amex. So I started looking for a new role, and I reached out to a guy one day and I said, “Hey, I’ll give you a Starbucks gift card if you give me 30 minutes of your time.” He was an FP&A consultant, and he goes, “I don’t want your gift card. I want you to write an article for my website.” And that’s really where my journey started to becoming an influencer and eventually launching my own business. I wrote the article for him. I remember thinking, “You don’t want an article for me? I suck at writing, but I’ll do it.”

(03:51):

And I actually kind of enjoyed it, shared it on LinkedIn, did a few more with them. That led to being on some webinars, and really led to growing a brand. And in late 2021, everything came together in about a month period where a friend reached out to me and said, “I know a company that needs about 10, 15 hours a week of FP&A support.” Another guy reached out to me that I’d been on a webinar with and, “Hey, can you design an Excel course for me? I can sell it. I just can’t teach it. Can you teach it?” I’m like, “Sure.” He’s my partner now we’ve done training all over the world, both remote and virtual, and probably thousands of people at this point in the last couple of years. And then three vendors in one week in the FP&A space. Because I’d written a bunch about the FP&A space, that’s when I wanted a job.

(04:34):

I wrote about them because I had a following of a couple of thousand at this point, thinking, “Hey, if I say good things about them, I’m more likely to get hired during the interview process.” They didn’t hire me, but that led to them wanting to do marketing arrangements with me. Because I was one of the bigger FP&A influencers at the time because there really wasn’t any, there’s quite a few now, and so in one week, like I said, three of them offered me a job. I went to my better half, my spouse, and said, “I can start a business with this.” And she looked at me like, “What are you talking about? You haven’t made any revenue yet.” I’m like, “No, I can turn this into a business.” And we talked about it for about six weeks, and one morning she said, “I’m good with it.” Put in my notice, and that was two and a half years ago.

Randy Wootton (05:12):

That’s unbelievable. That’s great. Just you have that wonderful lived experience of luck being the intersection of preparation and opportunity, and I think gumption, the willingness to step forward and feel what the universe is telling you and saying, “I’ve done a bunch of work in this space. I have some skills and capabilities. I’m getting these different opportunities and listening for that,” and then stepping forward and embracing it. That’s awesome. So now you’ve opened up this whole new chapter. You, yourself, are a podcast host. You have three of them going. You want to talk a little bit about them for people who don’t have enough podcasts already?

Paul Barnhurst (05:49):

Sure. I’ll give the 22nd overview. One is, one of the premier podcasts out there globally, all about FP&A, called FP&A Tomorrow, talking about where FP&A is going, weekly show. Second one is called Financial Modeler’s Corner. So it’s really about financial modeling. How do you be a good financial modeler? That’s something I didn’t know early in my career. I think so many people don’t have good training, so really focusing on the fun stories in there, but also what people need to know to be a good modeler, And then the most recent one, I co-host with Glenn Hopper, who’s a global expert in AI. He has a degree from Harvard, a data science certificate, a really smart guy, CFO, and we host a podcast together about technology and AI.

Randy Wootton (06:31):

That’s great. So you’re covering the full range of things, and that’s how we got introduced, and you’ve been a great source of information and advice, and counsel in terms of what’s going on in the FP&A space and how we should think about it as Maxio really being a historical system of record, but working to partner with companies that are helping the other side, not the accountants and the controllers, but the FP&A folks do their jobs with their technology. So it’s been a fun opportunity and super informative as well. So let’s start there, because I think the FP&A tomorrow. One of the things to think about is, well, what was FP&A yesterday? And we were chatting over the last 20 years what you have seen play out, and I thought you drew this really interesting distinction between FP&A and FP&R. Would you talk a little bit about what it was 20 years ago and then how it’s evolved a little bit and what today we really think of as FP&A and why we think of it as FP&A, and then we’ll get into some other topics on that?

Paul Barnhurst (07:26):

Yeah, well, FP&A really came out of controller and accounting role, often it still is in many countries referred to as a controller and often it’s the accountant that does it. And accounting is much more of a control role, and historically, what happened and reporting. And so by nature of that, as it first came about, it was a lot more of the reporting. I can remember working at Amex and some decks that went around early in my career. They’re like 100 pages long, and I look back now, and why would anyone read that? How do you get value out of a 100-page report? I imagine, as a CEO, if someone stuck 100-page report on your desk and be like, “Okay, what’s important? Because I don’t have time to flip through this.”

Randy Wootton (08:02):

Yeah, net it out. Net it out for… Show me pictures. I’m the English major. Show me the pictures and the graphs. Many times, our controllers and CFOs give me the spreadsheet. I’m like, “Look, I can’t.” The numbers start to blur; show me the lines.

Paul Barnhurst (08:17):

As I like to call it, or finance. We have to have our walls of numbers, but put those in the back. Start with the story and the visuals. And I think what’s happened in the last 10 years, and it’s accelerated with COVID and just the constant change of our environment, we’re seeing a rapid change globally in the economy and society. With that, the role of FP&A has become much more important, much more strategic, and commercial, as has the CFO. Many ways in FP&A is a mini CFO, particularly if they’re a business unit FP&A where they’re supporting a general manager. They’re looked at as one of those right-hand people to really help guide and ensure good decisions are made. I like to think of FP&A today as what it really should be is, how do we ensure the business is best utilizing the next dollar of spend?

(09:07):

And we have to understand the historical, and we have to understand the operations and business. And that goes so far beyond the reporting and even just the analysis. And that’s really what I see as FP&A today. So I think it’s been a gradual shift. It’s really accelerated the last few years, and especially with COVID is all of a sudden every business was turning to their finance department and saying, “Okay, what’s our runway? What’s the impact if this happens?” And asking for a million different versions is we were all trying to figure out what’s this new world we live in.

Randy Wootton (09:37):

Especially, I think, in where we focus B2B, SaaS, VC, PE-backed companies where the early-stage companies are dependent on the cash because they’re in a negative burn state and trying to do the scenario planning in a world where you just don’t have assumptions that you can believe anymore; you’re just creating them. And I do think that other distinction you’re making in terms of back office to front office or policemen to partner, is this idea of tolerance for risk and uncertainty. And I think that a lot of controllers orientation is to eliminate risk in terms of the reporting and alliance with gap, whereas as you shift into the FP&A mentality, it’s you have to be able to embrace uncertainty and create scenarios, and you no longer have the answer. What you have is a way to frame the question and potential answers, and that’s a different way of thinking about the world.

(10:30):

When we were chatting, you had a couple of trends; maybe you want to touch on them. One, you mentioned COVID in particular, but this idea of thirst in access to data and, along with that, the willingness to invest to get the data, and how that was different 20 years ago versus today. Some ways compelled by the acceleration of change, the need for faster and deeper analysis. Can you talk a little bit about what you’ve seen over the last 10 years in terms of the need for data, the willing to invest in data, and then the skills to make sense of the data?

Paul Barnhurst (11:04):

Yeah, I mean, first, just data has exploded, and especially since generative AI came about now, it feels like we’re just an awash in data. And so with that, companies have wanted to get insights out of their data, and FP&A, the analysis, the A stands for analysis. So you go to your FP&A department and ask for analysis with all that data, and I think that’s also really accelerated the reliance on FP&A, and it’s accelerated so many finance departments working closely with analysis, or the CFO often owning the data analysis. Not always, but you see that a lot. One person I interviewed one time who was a startup CFO four or five times, he goes, “You can tell if you’re working with a modern CFO because they often own the data.”

(11:52):

And they’re owning it more from the metric standpoint, because you’re a third-party observer, so to speak. You don’t have a dog in the fight of how CAC is calculated or how lifetime value is calculated. You’re looking at it from an economic standpoint and saying this makes the most sense. Where marketing may say, “Well, we don’t want to include that; it makes the number look better,” or procurement may say, “Oh, no, we want to look at this thing this way.” Because we all have a tendency to do it, even finance people, right?

(12:19):

Of wanting to look good. And so, I think that also has really accelerated that front office, the analytical, the decision-making as there’s become more and more, and especially SaaS as it’s grown, and SaaS is usually cutting edge. SaaS technology companies, they’ve looked to that finance to also be cutting edge. And so they’ve been much more data involved; they’re much more into the metrics. Sometimes you see them bringing in, as they scale enough, some kind of data people into the groups, like Facebook had a data science team in their finance. That profit was part of what came out of that, which is one of the most popular algorithms out there for forecasting now.

Randy Wootton (13:00):

Yeah, I think even at Maxio, we’re a relatively small company, about 2,400 customers, we have a data team.

Paul Barnhurst (13:06):

Yeah.

Randy Wootton (13:07):

That was initially sitting under rev ops, and then we moved them under finance or in the financial operations to address the exact point that you’re making in how do you aggregate and connect all the different sources of data, so that then you can drive more insight unit cost economics, you can start to look at cohort analysis at another level of detail. You can drive your go-to-market strategy. And so, I think your point around being a Cambrian explosion of data, no one would argue that, but it’s chaos. And so, I do think there’s also been this, as we were talking about this advancement of both BI and data visualization tools.

Paul Barnhurst (13:44):

Yes.

Randy Wootton (13:45):

So people being able to make sense and connect all the data. And then your point, which you made to me, was, well, Excel has improved too. And so, it’s been easier to have it be a tool of interpretation versus just spreadsheets for I-Bankers.

Paul Barnhurst (13:59):

Yeah, I 100% agree. I mean, Power Query, they can integrate Google Sheets, and there’s even new modern tools that have made it so much easier to integrate your data, whether that’s Power BI, you now have Python, you have ways you can do millions of records. You can basically do what you can do in BI and Excel if you need to. And so it’s really what I like to call it is modern Excel. And it’s based on being a more modern analytics tool. And what I’m really excited about, I think, in the next year or two, is they really start beefing up Copilot.

(14:31):

I think right now they did the typical Microsoft: you release it in the MVP state, you get all the feedback from the customers because it’s not very good, and then you keep iterating until all of a sudden it’s the best thing in the market or one of the best. That’s what they typically do. And that’s what I expect to see, especially with Python coming out in Excel. People will be on the write algorithms using the code interpreter in AI. I think it’s really going to revolutionize the citizen data science side of things in ways that we probably aren’t even thinking about yet.

Randy Wootton (15:05):

We’ll come to that in a couple of minutes, going into the rise and role of AI in finance and modeling. But before we go there, I think the other things we talked a little bit about, I just want to underline, underscore was the rise of cloud, SaaS technologies allowing it, making it easier to tie fintech systems together. So you can have a fintech stack, and you can have a stack at the CFO trust. Think that having spent my career selling to marketers and service people, and salespeople, like CMO, head of marketing, all they want to do is buy new tech, and optimize it. Which channel are we going to optimize?

(15:41):

But I think there’s been a real reluctance in CFOs because, at the end of the day, the data has to be right and they’re accountable for it. And so I think with the rise of cloud and the ability to stitch tech together more easily has led to them having access to more power. The other thing we talked about is the democratization of tech and that it’s much easier for the small and medium company to pay for a planning tool. You don’t have to buy IBM, whereas it was only the large companies. I imagine at Amex you adopted some of the technology for IBM, Oracle, whatever; you had a bunch of them.

Paul Barnhurst (16:13):

We had both Oracle and IBM. [inaudible 00:16:17]

Randy Wootton (16:17):

And huge implementations: hundreds of thousands of dollars you’re spending every year, and the early stage companies can’t spend that type of money, but now it’s much easier to find an affordable alternative, something that is if it’s not just a point solution; it’s helping you get better at your job. So all of those trends have really brought us to today, where now we have this new future of finance.

Paul Barnhurst (16:39):

Agreed, very much so.

Randy Wootton (16:42):

Well, before we go into the AI of finance, I would like to just go back to your second podcast, which I probably should be listening to myself, the Financial Modeler Corner, where you give best practices on how to do financial modeling. Do you have your top three recommendations, best practices that you tell people that are launching their career, top three things they should do, and, I don’t know, top two or three things they should absolutely avoid if they want to make good models?

Paul Barnhurst (17:08):

Sure. The number one thing I emphasize, and I actually have a course that I rounded, it’s about an hour and a half, is Modeling Design Principles, learning how to design a model. And I had the best analogy, Lindsay Weber shared the other day. She goes, “Think about it this way: when you pick up someone’s model, it’s like going into somebody’s house, opening their dishwasher, and starting to put away the dishes when you have no idea what cabinet they go to. It’s really frustrating; you got to spend five minutes orienting yourself.” And that’s how so many models are built, because people don’t spend time on proper design. Ian Schnoor, who’s the global executive director of the Financial Modeling Institute, number one for accreditations in modeling. They have a program similar to CPA where you can get accredited as a financial modeler. He had said, “Far and away, and there’s not even a close second. The number one problem with models is design.” It almost always comes back to design.

Randy Wootton (18:09):

Yeah.

Paul Barnhurst (18:09):

So that’s the first principle, I’d say. What does that mean? That means things like you should color code and it should be clear what your color coding is; you should have an instruction; your inputs, your outputs, your executive summary should be separated. You should never hard code unless it’s an input that you’ve clearly coded, but never put it in a formula. We’ve all been there. How many of us has opened a spreadsheet and spent hours trying to figure out where somebody stuck that stupid number, and you’re like, “Not only did they stick it way off to the side, they hid it, and they made it white on top of that.” Why? And they didn’t label it. And so it’s those type of things that are most important. The do’s are have structure, use color coding, have error checking, don’t hard code. Those are really, if I was to boil the rules down, there’s others, but those are the key ones that if people follow, they’re going to be ahead of the majority of the curve.

Randy Wootton (19:04):

That’s great. And I think you’re building for scale and building for succession when you do that.

Paul Barnhurst (19:09):

Yes.

Randy Wootton (19:09):

I think what I’ve seen, because I usually come in at a series B, series C type company and I sit on the board of a couple. The issue is it’s often the founder, who’s a technical person, who knows Excel, who builds the initial models; they don’t know how to create design, they don’t want to spend the time doing it. And so, then you bring on the fractional CFO who’s in there just for a contract. So they build the model, but they’re not getting paid to build it for persistence. They’re getting paid to get to the answer for some sort of funding round. So then the new CFO comes on board and has to basically eject the model that’s there and start over. And I would say, Paul, I don’t know if you find this to be true, but I do think it’s kind of like a rite of passage for every new CFO of a tech company come in and build their model.

(19:55):

And they spend their first month doing it, and I don’t know if it’s a badge of honor or to show how hard-core they are because they can build their model. I also think it comes from they don’t know where the gotchas are; they don’t know where the hard-coded cell is or the hidden… And it’s so hard to be the detective and ferret out that, or to impose what you described as the error testing on top of a model that you didn’t build. And I think everyone’s a little scared if they’ve got investor money. They walk into a board meeting, and they’ve got a bunch of Excel jocks sitting on the other side who are going to send them to math camp. And if they don’t feel completely comfortable in the model, they feel exposed because they’re supposed to have the answer. So we get into this really crazy cycle of not having a sustainable model.

Paul Barnhurst (20:46):

Yeah, really interesting. Fascinating story on that guy by the name; I think it’s Rami, I believe his last name is Essaid. He’s the founder of Finmark, which was bought by Bill.com.

Randy Wootton (20:55):

Yeah, I know him. I know Greg.

Paul Barnhurst (20:58):

Tells the story… He started the company, he was working for an IT tech company as a startup, and they’d raised a bunch of capital on a certain assumption in a model. All of a sudden, he realized when the CFO had, they’d built it and checked it. They had been switching from either it was annual payments to monthly I think, and they left in the annual payments. So the revenue was way higher than it was supposed to be, and he had to go in front of the company and lay off. I think it was 90 people.

Randy Wootton (21:21):

Oh, wow.

Paul Barnhurst (21:22):

Because they got way out over their skis. Now they ended up being successful, and he had a successful exit with the company: “So that was the most painful moment I’d had in my life; I’m like never again.” His next company was an FP&A tool because he wanted to solve that problem. And so it was really interesting to see it from a CEO’s perspective and hear him tell that story, and you could just tell the pain he felt. And so, I do think it’s the rite of passage, and I do think it’s so important that the CEO or CFO, when you come in, really know that model. And I think sometimes rebuilding it helps them understand the business, especially if they’re building a good driver-based model. They’re really going to start understanding what drives the business. If they’re not doing it that way, then I would question the model.

Randy Wootton (22:05):

Right. I think you’re right. The other thing is, so I am a technologist but not an engineer, but I’ve been in technology for 25 years, so I can talk. I started as a product manager, the point I’m making is I can write business requirements and I can talk to engineers, I can’t check their code.

Paul Barnhurst (22:25):

Sure.

Randy Wootton (22:25):

I can’t tell you if the code is great or how they might want to design it differently. I have to rely on the head of engineering to be able to do that and be able to take the business requirements that we’re establishing for the market to be differentiated and win and to translate that into language, and oversee the engineers who are going to deliver. Similarly, with finance, I’m technically fluent in finance. I can run a spreadsheet; I’m not great at it, but I’m not going to be going in and double-checking every single tab when there are 50 tabs that the finance team has put together. So there’s this level of trust that you have to have with the CFO that he or she have done that work and that they know it. And then we’ll have productive conversations around the assumptions, the drivers as you describe them, especially in the scenario planning as you go into the budget season or as you’re trying to diagnose issues in specific cohorts for contraction or loss.

(23:22):

But you’re relying on the expertise of the CFO and the controller to have done it right. And I know every single company I’ve been part of, we cycle on it a couple of times, and you’re like, “Oh, well, now we discovered this, and oh, isn’t this interesting? And, oh, we forgot about that.” And so you also have to have this level of trust between the two that you’re going to explore this together. I think there’s sometimes some CFOs I’ve interacted with where they own the model, they own the answer, and they get very defensive about it. Even if you’ve been able to say, you go in and, “I’m not trying to find something wrong, but this doesn’t seem to make sense, and oops, I need to change the model.” There’s not this level of co-leadership or humility to acknowledge that there could be a mistake, and that isn’t it better if we work together to try to get a better model that gives a more accurate output? I don’t know if you have any thoughts on that dimension.

Paul Barnhurst (24:16):

Yeah, I mean, two thoughts come to mind. First, I’ll share an experience when you talked about the being defensive. I’ve been there; I’ve been defensive sometimes, but I still remember I built this model and my boss was going through it all, and I can’t remember if he was my CFO or general manager at the time. He was both at different times. And so I supported him, and I also worked for him, and we were going through the model. I looked at it, he goes, “Now this model’s good enough for government work,” which was his main way of saying, “Man, you built a terrible model here.”

(24:42):

There’s a lot of flawed assumptions. It’s like, “Thanks,” because it’s something I’d never built before; I had no idea. We just took a first round. But after I got over the initial of that kind of hurts. He had a lot of good points, and we went back and we kept working, and we got a good model. Because I was willing to be humble enough to say, “Okay, he’s right. As much as I don’t like to hear it, it could be a lot better.” And we went through it. And the second is an analogy I really like, and again, I’m going to go back to Ian Schnoor, who shared this. He tells a story, not a story, but he shares this analogy that he says, “If you have a car,” and I’ll ask you this question: do you ever look at the engine before you hit the start button?

Randy Wootton (25:21):

Never.

Paul Barnhurst (25:22):

You hit the start button, expect it to work.

Randy Wootton (25:23):

Correct.

Paul Barnhurst (25:24):

That’s what the end user wants of a model. If you’re taking them into the details, the sausage making, it doesn’t even matter if you’re right. If it’s poorly organized, you’re going go here to this sheet and that sheet, and you’re moving all over the place; they’re losing confidence with every sheet you switch.

Randy Wootton (25:39):

Right, right.

Paul Barnhurst (25:40):

And I’m sure you’ve been there where somebody’s bouncing back and forth, and even if they’re right, you’re like, “Okay, the way they designed this, I’m not sure I can trust this.”

Randy Wootton (25:48):

Yeah, there’s something to be said for that. Which brings us to maybe the third part of this conversation and tied in with your third podcast around the future of finance, AI, and tech that you do with Glenn Hopper. The role of AI in finance, one of the things that I’ve heard is finance executives are struggling a bit to trust the models. In this case, the algos that are producing things or producing answers, producing scenarios because they like to be able to look at cell ZZZ 4,023 and know they can see that number and not rely on this augmented intelligence that is taking in the data and providing answers that they can’t trace the source.

(26:30):

It’s like you can’t go to the transaction level of detail. And again, if we think about controllers or people who come up through accounting, part of being able to do gap reporting is you can go find the receipt for that expense that then rolled up into sales and marketing that then rolled up into your CAC. So maybe we start there in terms of this shift. I hate to use the word paradigm, but I do think it’s a paradigm shift in terms of how finance executives need to think about modeling and the potential power of leveraging AI versus, in that balance, the power of leveraging the AI versus the risk or uncertainty of using AI.

Paul Barnhurst (27:13):

It’s a great question, and I don’t know that we, obviously we don’t have it all worked out. We’re all struggling with it, but I’ll share a couple thoughts on that. Like you said, finance, they hate the black box. We want to be able to see everything and tick and tie it, and make sure it all flows and that we’re confident and can sign off on it. Well, it is tough when somebody comes and says, “I used AI, and I use this really complicated fiscal model that you wouldn’t understand, and it’s more accurate than what you do understand.” Right? That’s kind of the feeling underneath, but it’s been done again and again.

(27:47):

And I think some good examples of how it’s done is if you look at Microsoft, I had on my previous podcast called FP&A Today, I had on a guy who was part of what’s called FIN, which is their open source algorithm they use for forecasting; they use it for all their revenue and expense. And what they did is they built the model, and they used it in partner with the human. Humans still did their forecast for their first year or two. They compared them, and what they started finding is the revenue on the automated, using AI was more accurate. And so what they did is they still give the person who’s over that part of the business the option to override it, but they have to explain why.

(28:28):

So I really like that. Let’s make sure we have human judgment with the technology and go slow, and test it. There’s ways we can look at those numbers and go, “Okay, how did that compare to what I have?” And then, what are the differences? And if there’s somebody who can help explain that, knowing the inputs that went into it, so I think with AI, particularly, I’m thinking more algorithm, statistical for predictive modeling. It’s a co-partnership with somebody who has that experience to help understand some of that machine learning so that you can get the best answer possible. I don’t think it’s just take the AI, trust it, and move forward. It’s going to be that journey and getting people comfortable. And that will be true with generative AI, and all of this is having ways to back-test it, to get comfortable. I think any company that’s just saying, “We’re going to start using AI and trust the numbers,” and not get comfortable is being foolish, and they’re going to get burned sooner or later. Because AI, just like human, is not infallible, it’s going to make mistakes.

Randy Wootton (29:29):

I think in some ways, as a populace, we have developed a relationship with computers where computers aren’t wrong, but we haven’t had that intelligence layer that goes on top that we’re trusting to make sense of what the computer output is, right? The computer [inaudible 00:29:46]. I think we were drawing a distinction in our pre-brief between machine learning and generative AI, and I do think machine learning has been around a lot longer. In fact, I was a CEO of a company, Rocket Fuel, that was doing logistic regression analysis back I mean, they started way before I joined, but that was in 2015 to 2017. So it wasn’t gen AI; it was more of the algos, the predictive analytics type model, statistical models to perform complex tasks. In this case, bidding algos for marketplace for advertising. But can you talk a little bit about, through your podcast, how you’ve maybe drawn that distinction between AI as the superset and then machine learning and how developed that is and how risky it is versus the gen AI that’s really taken off over the last 18 months, at least in a commercial sense?

Paul Barnhurst (30:41):

Yeah, I mean, so a couple ways I think of it. I think the machine learning is still the biggest area that’s used, particularly in planning, which is the key part of FP&A, right? Predictive modeling. I think that stuff is pretty refined, and if you have good data, you’re going to get a pretty good output. Now, what you have to remember with any of that is it’s using historical assumptions and then any variables that you’ve put in there. So it’s only going to be as good as the inputs. After COVID, many of those statistical models were no good anymore because, “Well, we had this and then this and then this,” and it’s like, okay, they’re all outliers.

Randy Wootton (31:19):

The black swan events, right?

Paul Barnhurst (31:23):

Exactly.

Randy Wootton (31:23):

How do you account for that, right.

Paul Barnhurst (31:24):

And so that’s still a challenge you’re going to face, and that’s true of any of it, but I think the most mature and where we’re pricing the most benefit and planning is with machine learning, predictive modeling. Now, when we talk generative AI, I think there’s a few areas that’s going to be really helpful, where we’re seeing the most use right now. I think the number one area is helping with emails, content, brainstorming, writing. I’ve seen people who’ve used it to write collection letters, and it was a lot better than the letters they were writing. A lot of those type of use cases. Also, analyzing external data, helping you with an Excel formula, all those type of things. Most people are still not able to run their company data through AI. If they are, great; if they’re doing it on a public tool, shame on them.

(32:14):

But that’s… Right, you got the whole security. And so that’s going to take time is how many data centers and security until everybody will be able to do that. I think that’s the third step, and we’ll get a lot of benefit from that because it can analyze that data, help with scenarios, start to bring in some of that predictive modeling, and I think that’s where it’s going to be a little slower moving, but we can definitely do it today. Even if you don’t want to give it your data, you can randomize some data, use code interpreter to write some of the Python, and then run your data against that Python and different things. So there’s definitely opportunities to be on your data, but I like to think of it, predictive modeling. I almost think of it more as an analyst or a little more mature, a little more seasoned. Generative AI is that intern that you would never just take the interns data and stick it in front of your CFO.

Randy Wootton (33:03):

Right.

Paul Barnhurst (33:04):

And you shouldn’t be doing the same. You’re still going to check the predictive modeling, but you may not go as deep. You may have more comfort because you’ve been doing it for a while. Generative AI is that new employee, the intern, or just out of college. You need to be more careful in checking it. And I think content is the big area. Data is continuing to come more and more data uses, feels like almost every day.

Randy Wootton (33:26):

And just two points in that. One is, I think, spot on; the broader point you’re making is everybody needs to develop some sort of familiarity with AI and generative AI, like a new language.

Paul Barnhurst (33:40):

Yep.

Randy Wootton (33:40):

It is going to be part of the future. And to your point, it’s an early stage, but if you don’t start now, you’re going to get left behind. And I think that’s true, especially for any company that’s VC, PE-backed; they’re going to be asking you about how are you using AI both within your system. And how are your people using it so they become more efficient and effective? Just throwing more people at a finance team is not going to work. You’re going to need to be able to use tools to scale your team and your effectiveness and efficiency, and I think there will be an AI tax, but efficiency assumption and go-forward operations.

Paul Barnhurst (34:15):

100% agree; it’s a productivity tool. And one other thing I’ll say: I’ve had multiple people say this on my different podcasts, and at first I kind of pushed back, but the more I’ve talked, the more I think it’s true. What we’re going to see is AI is going to make us more strategic and more technical in FP&A at the same time. What I mean, not so much Excel, but more a little bit of the data science, maybe needing to do a little bit of coding, maybe needing to understand the algorithms, not to the point of a degree, but I do think it’s going to make us more technical in an area that, frankly, FP&A has been left behind in a little bit, and a lot of times you’ve had to bring in others to fill that void.

(34:53):

So I think it will be a good thing long term, but I think the technical is going to shift a little bit from, “Hey, technical is just you’re great at Excel,” to now you need to understand a little bit of AI, maybe some Python, some other things to really get even more out of the AI tool.

Randy Wootton (35:10):

Well, that was what I was going to go back to because I think your opening, where you talked about that you are interested, you’re an entrepreneur, so I think that is the future of CFOs. They need to have a little bit more risk tolerance. Number two is you had this background in tech, being able to write SQL, so similar to me being technologist, but not being an engineer. Is you having some of that capability allows you then to think about understanding the data, how the data is organized, how it gets accessed, how it gets aggregated, how it gets shared, which is more than just making sense of the data.

(35:41):

So I think as we think about the future, CFOs of tomorrow is having this tech dimension as well as having a comfort with the underlying data infrastructure, not just taking the data. It was great. Well, hey, Paul, we’re running up against time. I always love to do the speed round. So the speed round is meant to be quick. Three questions. One: What’s your favorite metric? What’s your favorite book? The book could be personal; it could be business. And then who’s the influencer? You’ve given us some great names for people to go track right now. I’m going to go sign up for Glenn Hopper, but if there’s anybody else that you really think people should be following, but all right. So, what’s your favorite metric?

Paul Barnhurst (36:17):

Favorite metric? I’m a big fan of… I’ll give two.

Randy Wootton (36:21):

Okay.

Paul Barnhurst (36:22):

CAC and lifetime value in the SaaS area.

Randy Wootton (36:24):

CAC and lifetime value. So lifetime value, so you have a really good understanding of the customers you’re signing and you’re spending all that money, all that CAC to acquire, how long are they actually going to be there? I think there are a lot of companies when they do it, formulaically though, their early stage, they’re like, “Yeah, we’re going to have customers for 10 years,” but they’ve never had a customer for 10 years. How do you help people understand lifetime value and make sure that it’s reasonable when they start to build their models?

Paul Barnhurst (36:48):

Yeah, I think some great things is, I think it’s called Benchmarkit now. Ray Rike’s done great work there. There’s others of go look at the benchmarking for the industry you’re going into, or if there isn’t one, something similar. And don’t assume you’re going to be way better than the average; maybe a little bit on both sides, but at least for starts be somewhere similar to the benchmark because that at least keeps you relatively conservative versus making an aggressive decision that could completely blow up your cash runway if you’re wrong.

Randy Wootton (37:17):

Awesome. And then, similarly, on CAC, you can use benchmarks to help understand, by size of company, by size of deal, how much is it costing. And I think what we would see more broadly, we do a growth rate analysis across our customers and produce it in the Maxio Institute growth report. And what you’re seeing is growth rates have dropped, but if you line that up against cost per keywords across Google, the price has gone up. So CAC is going up, but growth rates are going down, it leads to really interesting conversations with the board.

Paul Barnhurst (37:53):

And then money’s no longer cheap on top of it.

Randy Wootton (37:55):

Then money is no longer cheap. So you just can’t; as everyone has talked about, you just can’t grow at all costs. You need to embrace this broader idea of efficient growth, which is inclusive of CAC, and understanding what that is. As well as LTV, how much are you going to invest in your current customers and try to get them to buy more stuff? So two great metrics. Favorite book?

Paul Barnhurst (38:15):

Oh, that is a tough one for me. I’m going to, How to Win Friends and Influence People, is just a classic.

Randy Wootton (38:22):

Okay, there you go.

Paul Barnhurst (38:24):

And I have to go with that one. I love that book.

Randy Wootton (38:26):

All right. We all could be nicer.

Paul Barnhurst (38:28):

Very true. And remember people’s names. That’s one I need to get better at.

Randy Wootton (38:31):

Remember people’s names and be gracious is one of the things I remember. Be kind. We need that in the world writ large. That’s great. Great book; I haven’t read it in a long time. I’m going to stick it on my stack. Influencers, other than the ones you mentioned today, who’ll include links to in the show notes.

Paul Barnhurst (38:47):

Yeah, obviously Glenn Hopper that I mentioned. Another one there; I’ll mention two. One on the AI side. Actually, I’m going to do three one on all three of my podcast areas. So in Future Finance, AI side, Christian Martinez is fabulous. I recommend following him. He really knows this stuff very technically sound, and he’s a veteran in finance.

Randy Wootton (39:06):

Sorry, which company?

Paul Barnhurst (39:07):

Christian Martinez. He’s out of Europe. I think he’s out of Amsterdam, if I remember right.

Randy Wootton (39:11):

Oh, okay.

Paul Barnhurst (39:12):

He works for Kraft Heinz at the moment and teaches some courses in his spare time around AI and how finance can use it in technology.

Randy Wootton (39:20):

Awesome.

Paul Barnhurst (39:22):

He’s the first one. In the financial modeling area, a great one to follow is Chris Reilly, especially if you work anywhere in the middle market. He’s just an amazing teacher, has a lot of great courses out there that will help you be a better modeler. And then, when it comes to FP&A, one of my favorite people to follow is Carl Seidman. His content is just some of the best out there. He really knows how to relate to people and continually give valuable added content, versus what I’ll call fast food content. And anyone who follows LinkedIn or any other social media will know the difference between the two.

Randy Wootton (39:58):

Wow, that is great. And thank you for what you’re doing in the space, Paul. I think you are helping to carve out this broader understanding of what needs to happen with FP&A and helping folks get better at it. And so both from the articles that you’re delivering, the teaching that you’re doing, and the podcasts that you’re hosting. Thank you.

Paul Barnhurst (40:19):

Well, thank you.