In this interview, host Paul Barnhurst talks with FP&A technology veteran Lee Lazrow (Revelwood) about building effective what‑if analysis. Lazrow contrasts "good enough" siloed analyses with truly useful ones that model the entire flow‑through impact of a change—linking units, revenue, variable costs, headcount, cash flow, taxes, etc. He stresses selecting drivers that directly affect the target outcome and using allocation methods that reflect real cost drivers. The conversation then shifts to the pitfalls of scaling Excel‑based models: larger workbooks increase formula‑error risk (citing studies that >90% of decision‑making spreadsheets contain errors and high‑profile $6 bn and $2.5 bn losses). Modern planning platforms are presented as a remedy, offering built‑in error checking, consistent calculations, and auditability. When asked about AI, Lazrow notes limited adoption due to fear and recommends phased, parallel pilots to build confidence. Finally, he offers three practical steps for FP&A teams: (1) identify the top 20% of drivers that keep the CFO up at night, (2) model multiple scenarios (high/medium/low, best/worst, confidence intervals), and (3) ensure data is specific, measurable, accurate, reliable, and especially timely (SMART). The interview wraps with a call to action for listeners to explore Revelwood’s solutions.
Chapters
Today, I'm joined by Lee Lazrow from Revelwood, who has over 30 years experience in FP&A technology. Lee, would love to know what separates good enough what-if analysis from what-if analysis that actually drives decisions.
First off, thanks for having me. I would say the difference between good versus good enough is thinking about the part versus the whole. So good enough is when we look at the part. Good enough is when we do our analysis in a silo and we're only thinking of that one piece. Good is where we also analyze the whole flow-through. What else is impacted by that one change? If I'm doing an analysis, maybe I'm looking at units, maybe I'm thinking about how are my units gonna grow. If you think about it, the units don't just impact revenue. The units are also gonna impact your variable costs. It's also gonna impact, if you're a manufacturing organization, headcount overtime. It's gonna impact your cash flow because you're gonna need the cash to support the excess material cost. Some of the best systems that I've worked with, some of the best systems that we've designed, are models that link all these pieces together. One example of a model that we, we created was for an oil and a gas company that we work with, and we wanted to determine the cost impact of adding another pump. In addition to the cost of the pump, We then took into consideration the people that we needed to run it, which led to staff costs. Well, that flowed into the fringe benefits, making sure that they have phone and benefits and things like that. That flowed into the taxes associated with all of these. The difference between the good enough and the good analysis is making sure you're thinking about the big picture.
As we talk about that big picture, I love that answer. And we talk about the importance of going beyond just, hey, good enough to thinking about the whole thing. When teams are building what-if analysis, how do they make sure they're doing something that's good enough and actually focus on those drivers that matter so they get something that's good, not just good enough?
That's a great question. And when you think about the drivers, th— this is similar to what I, I reference when I talk to people about allocations. The answer is think about your target, think about your goal. If you select a driver doesn't really have an impact on the target, then this is going to shift your focus away from what you're actually trying to achieve. Let's think about another example. Maybe you have a small sales department. Maybe they're not too technically savvy. If you're trying to think about IT costs and you're creating some kind of an allocation, if you look at a driver such as headcount, well, you have a small sales department. You're not going to allocate much of your IT costs, and that's going to make that department look very profitable because the costs are low. But if you allocate those same IT costs based on the number of support calls, they're not too tech savvy, they might be taking up half of IT's time. This could have a major impact on the department's profitability. So thinking about that driver is thinking about the whole concept of what's my focus? What's my end goal?
Great reminder of the whole idea of the end goal. So important to keep in mind is what am I trying to accomplish and what are the key things that influence that? With what-if analysis, obviously, is You know, many companies do their what-if analysis in Excel, right? It's where we all start in a spreadsheet. As many companies scale, they do this all in Excel. They do their what-if analysis. So what are the challenges? Kind of where does it break down as organizations scale?
I think the bigger question here isn't just about the what-if analysis as the organization scale, but how Excel comes into play as the organization scale. As companies get bigger, they obviously, they're gonna have larger spreadsheets. They're gonna have more rows, they're gonna have more tabs, they're gonna have more cells. And these cells reference other cells. Put this all together and you probably heard this before, but put this all together and the probability of some kind of a formula error gets larger as the spreadsheet gets larger, as the organization gets larger. I once read a study that found that over 90% of all spreadsheets that are used to make some kind of a business decision have an error. 90%. That's huge. And this leads to dollar impacts. You know, some of the massive stories that you've heard, you may have heard the story about a simple formula error at JP Morgan based off an Excel spreadsheet led to a $6 billion, billion with a B, $6 billion loss. Fidelity had a similar situation where they had a formula, they forgot to include a minus sign, and that led to a $2.5 billion error. I think the concept of scaling obviously comes into play, but I think the bigger use is the linking of those spreadsheets and trying to eliminate that.
Great point about errors. I know that's a common challenge with spreadsheets. You know, as I think about it, there's obviously some of the other things I see is the whole scalability, audibility, security, and then the one that I see floating around LinkedIn all the time where people joke, you know, Excel is not a database.
Absolutely. Excel only has so many rows and so many passwords you could put in. So you're 100% right. Yeah.
And so as you scale, that's where we see a lot of companies say, hey, I need a modern planning platform.
Yeah.
It's where they move. So what role do you see that platform playing specifically in what-if analysis?
So coming back to, to the concept that I had about errors, I think that one of the best opportunities with a planning platform is the ability to help prevent those accidents. And the best way to prevent those accidents Really comes down to helping your users. Modern platform includes some concepts that'll help make it easier, make it more user-friendly. One of those concepts is the ability to check your numbers. Yeah, humans will still have the ability— they're still going to be humans. They're still going to fat finger a number. They're still going to put in an invalid number. But modern systems will now have much more error checking. It'll make sure that if you're doing compensation planning, that you're not going to put in a negative salary, or you're not going to accidentally enter in a global 300% bonus instead of a 3% bonus. And it's also going to be set up so that you'll be able to implement and utilize consistent calculations. You want to make sure that if you're calculating some kind of a ratio, Department A is calculating it the same way Department B is. That JPMorgan issue we talked about before, that was a result of a manual copy-paste within a formula. Had that formula been automated, they might have been able to avoid that and the system would've properly performed the calculation.
All right, so artificial intelligence, it feels like the word of the century. What are you seeing there? How are companies using artificial intelligence to improve their scenario analysis?
I'm gonna say at this time, I'm not seeing a lot of companies adopting it yet. And the biggest reason why, the biggest thing we're ever seeing is just fear. It's a fear from a lack of understanding. And I think the biggest piece that comes into play is getting over that fear. Now, what I often do with systems, whether we're implementing a planning system, an AI system, even a spreadsheet, if you're changing your process, I'm a big fan of saying, do it in phases. Do a piece. And if you have fear, start with a small piece. Maybe you're gonna take your direct expenses. And when you do it using an AI platform, run it in parallel. Use your existing process, use an AI process, and then look at the differences. As you start understanding the differences, as you start getting more comfortable about, here's what I'm thinking and here's what the AI system is thinking, you'll be able to put the two pieces together getting comfortable. And as that comfort improves, the use of AI will improve.
Great point about how AI improves as you get more comfortable, right? Change management is such a huge part of any new technology, and it's gonna be a big part of AI. It will take longer for us as humans to catch up to what's capable than it will for the technology to catch up with what's possible. It always stays ahead of us humans. All right, so last question I want to ask you here. If an FP&A team wants to improve their what-if analysis starting tomorrow, what are the first 2 to 3 things they should be doing?
A couple of things. The first thing is what we talked about before, making sure you're identifying your top drivers, your key drivers. Think of it like the 80/20 rule. Don't try to model an entire P&L when you're starting. Think about what, what are the components that keep your CFO up at night? Model that, focus on the drivers for that, keep it limited. The second piece though, which is almost in contradiction to what I just said, is don't keep it limited to one what-if analysis. Don't model a single what-if number. Instead, model a set of what-if numbers. Some people like to think about high, medium, low analysis. Some people like to think best case, worst case. Statistical people like to reference it with confidence boundaries. But whatever you're thinking about, think about the fact that there's different possibilities and as a result, plan for multiple what-ifs. And the last piece will come into play, which is possibly the most important piece. In business school, we all learn the acronym SMART: specific, measurable, accurate, reliable, timely. I think that last one, timely, is really, really important. If your source data is out of date, if your data's not timely, then your end results will be inaccurate. That's going to eventually lead to bad decision-making.
Yeah, nobody wants bad decision-making. I would agree with that. I appreciate the SMART. I like that reference. Also the whole idea, keep it limited, but remember you need to think about those key situations, the key things you need to be optimizing. It's not just keep it super limited. You have to be realistic in that. So thank you for sharing some of your insights on, you know, what-if analysis. It's something that's so important to FP&A professionals. And so Lee, I really appreciate, you know, carving out a few minutes, sharing a little bit of your perspective. I feel like this was packed with practical insights that FP&A teams can use as they work to improve their what-if analysis. The other thing I'll say, it's great to see how the work you're doing at Revelwood is helping teams make that shift. You're doing a lot of great work there. Speaking of Revelwood, they provide solutions for the Office of Finance. So if there's anyone out there who wants to learn more, you can learn more at revelwood.com and you can contact them at info@revelwood.com. Lee, thanks again for coming on the show.
