What Separates an Excellent Financial Modeler From a Technically Good One with Diarmuid Early
In this episode of Financial Modeler’s Corner, Paul Barnhurst speaks with Diarmuid Early about what separates great financial modelers from technically capable ones, the rise of AI in financial modeling, and why judgment remains essential. Diarmuid shares lessons from his career, his experience as a three-time World Financial Modeling Champion, and why understanding what a model actually means matters more than simply building it correctly.
Diarmuid Early is the founder of Early Days Consulting, a data analytics and financial modeling consultancy. He previously worked at Boston Consulting Group and spent nearly a decade in Deutsche Bank’s Origination and Advisory division. He holds a PhD in computer science and is a three-time World Financial Modeling Champion, Excel MVP, and Master Financial Modeler.
Expect to Learn:
Why great modelers think beyond simply building the model
How assumptions and business understanding drive valuable financial models
Where AI currently fits into financial modeling
Why AI-generated models still require careful human review
Why continuous learning remains essential for financial modelers
Here are a few relevant quotes from the episode:
“The hard part is getting everybody aligned. What are the assumptions?” - Diarmuid Early
“There is a smart way to do it. And so if you're banging your head against it, go find the smart way.” - Diarmuid Early
AI will continue to take on more of the technical work involved in financial modeling, but speed alone does not make a model useful or reliable. Strong modelers will continue to stand out through their ability to develop credible assumptions, interpret results, exercise judgment, and understand the decisions a model is meant to support.
Follow Diarmuid:
LinkedIn: https://www.linkedin.com/in/diarmuidearly/
YouTube: https://www.youtube.com/@dimearly
Follow Financial Modeler's Corner: LinkedIn: https://www.linkedin.com/company/financial-modeler-s-corner/Disclosure: Portions of this episode (such as the introduction or promotional segments) use AI-generated voice narration produced under human editorial review.
Newsletter - Subscribe on LinkedIn: https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7079020077076905984Sign up for the Advanced Financial Modeler Accreditation today and get 15% off using the code “Podcast” at https://bit.ly/497oAqW
In Today's Episode:
[00:00] - Trailer
[03:31] - Diarmuid’s Journey Into Financial Modeling
[06:00] - Financial Modeler Stereotypes
[08:20] - Career Advice for Aspiring Modelers
[12:46] - Financial Modeling Competitions
[14:59] - Consulting, Coaching, and Career Evolution
[21:08] - AI and the Future of Financial Modeling
[25:41] - Why AI-Generated Models Still Need Review
[31:09] - Financial Modeling Standards and Best Practices
[35:19] - Diarmuid’s Most Challenging Modeling Case
[39:28] - Excel Shortcuts and Modeling Tools
[41:38] - Rapid-Fire Modeling Questions
[46:01] - Continuous Learning and Final Advice
Full Show Transcript:
Host: Paul Barnhurst (00:42):
Financial Modeler’s Corner is the world's premier modeling podcast. It is brought to you by Financial Modelling Institute, the world's leading financial modeling accreditation organisation. Welcome to Financial Modelers Corner. I'm your host, Paul Barnhurst, the FP&A guy. And in this podcast, we talk all about the art and science of financial modeling with distinguished financial modelers from around the globe. The Financial Modelers Corner podcast is brought to you by the Financial Modelling Institute. Core foundational modeling skills are more important than ever in the era of AI. It's one of the reasons that FMI offers the most respected accreditations in financial modeling because we need those core skills more than ever. And it's one of the reasons why I completed the Advanced Financial Modeler. This week, I'm thrilled to welcome Dim orDiarmuid early on the show. Welcome back. Thanks
Guest: Diarmuid Early (01:39):
For having me. It's good to be here.
Host: Paul Barnhurst (01:40):
Feels like a little bit of deja vu. I think almost three years ago, the first time we recorded, I believe was the exact same window.
Guest: Diarmuid Early (01:48):
Yep. Yep. This is my little perch up in the rafters.
Host: Paul Barnhurst (01:55):
Totally understand. I know you also, let's see, I think you've been on the show three times. You joined the great debate. Yes. You were circular references, if I remember right. You had a very strong opinion on that one, obviously. Still do. Yep. You'll get another opportunity to reaffirm that FDA. We'll see if anything's changed. But let me give our audience a little bit of background about Dim and then we'll get going. So he's the founder of Early Days Consulting, a data analytics and financial modeling consultancy. He previously worked as a management consultant at BCG London and spent nearly a decade in Deutsche Bank's origination and advisory division in London and New York. He has a PhD in computer science. He is a former CFA charter holder. He is also a three-time world financial modeling champion, an Excel MVP, master financial modeler. He currently lives in New York.
(02:51):
When this comes out, we'll see, with his wife and two children. All right. Welcome again, Dim. Ever wish you could get career advice from the best financial modelers in the world? Well, now you can. That's why we created The Career Corner. Let's go ahead and see what today's guest has to say. So I'm excited. We have a new section. We just added it. You're only the second guest I've done this with. We're starting a little career corner section. Kind of aimed toward those who are early in their career, are considering financial modeling and might listen to an episode. So how did you get into financial modeling? Was it intentional or did it just kind of happen? What's your journey?
Guest: Diarmuid Early (03:31):
I suppose more kind of happened than intentional. I used Excel a little bit in college, but purely kind of dabbling in. It was my at-home substitute for mathematica for playing around with math problems because I didn't have mathematica at home when we had it on campus. But I'd never used it for anything that you would use it for in a regular workbook.
Host: Paul Barnhurst (03:53):
Or you never built a model or really used it as a spreadsheet just for some math stuff.
Guest: Diarmuid Early (03:58):
Exactly. So I started at BCG, the Boston Consulting Group out of college. So obviously as a kind of junior person there, you're kind of hopping between companies, between industries, between practise areas. You're kind of all over the place, but you're always trying to get to grips with a new company, with a new set of data, with all this kind of thing. So there's lots of Excel, lots of exploring data. There was a little bit of financial modeling and trying to build projections and that kind of thing, but not a huge amount. And then I guess a few years in, one of my former BCG workmates told me about Modeloff, which is the financial modeling competition at the time. It's gone away since, and said, "Hey, this seems like your kind of thing." And I had sort of been the Excel guy in our office at BCG, and I think I was at Deutsche Bank by that point.
(04:52):
And I thought, "Yeah, that sounds like my kind of fun." I guess I had at least started my CFA studies at that point as well. So I had a little bit more of the technical side of things. But yeah, that kind of led me into it more from the Excel perspective. But then Modelof was quite technical in parts, which I guess got me thinking about that side of things more. And I ended up competing in that every year until eventually they said, "Okay, some of you guys have been in the final too many times and we need to make some room. So congratulations, you're all in the Hall of Fame. No longer eligible to compete." But I did learned a tonne through that and through the people I met through that. Lots of good friends that I'm still in touch with who are frankly much more pedigreed financial modelers than I am.
(05:46):
Yeah, between one thing and another, that led me down that rabbit hole. And eventually when I started my own business, a pretty large portion of that now is building financial models. So yeah, more accidental than anything else, the short version.
Host: Paul Barnhurst (06:00):
Not surprised. I think that's pretty common. A whole accidental modeler doesn't surprise me type of thing in the sense that you didn't set out to do it. I think that's pretty common for people. Next question I want to ask, and I'm really curious your take. What's one stereotype about financial modelers that drive you crazy? You're already smiling, so I know you have one.
Guest: Diarmuid Early (06:20):
Yeah. Well, it's interesting. I have one that I was actually guilty of myself because I remember. So for context, I did the Maths Olympiad for Ireland when I was in high school. And you meet some people there who are a handful of the countries that take the Maths Olympiad most seriously, select their team 18 or 24 months in advance, take them out of regular school and hot house them. I don't know if they still do. I wouldn't be surprised, but at least back in my day. So there were some extremely socially awkward people, extremely clever, but deeply awkward to have a conversation with. And so I kind of imagined that I was coming into a similar group of people the first time I went along to the Model of Final, that it would be super intelligent and super awkward people. And it could not have been more different.
(07:13):
I guess the way we kind of squared the circle over time was realising that the context in which most people become good at financial modeling is in a professional services firm where if you're going to be around long enough to get really good at modeling, you're going to have to. There's a sort of natural filter therefore you're going to have to want to spend time around people and explain your results and all of that, as well as being. It's not that the smartest people become the best financial modelers, you need something a little more balanced. But the upshot was that I was expecting these kind of mega nerds, and many of us are mega nerds still, but I was expecting this kind of very awkward group and they weren't at all just very -
Host: Paul Barnhurst (07:55):
Introverted pocket protector group that wants to stay to themselves versus, "Hey, we're all really social. We may be nerds, but we're social nerds and we can communicate." I think most people at one point have had that stereotype of modelers, so not surprised to hear you say that. What would you say to college students who are considering going into modeling or maybe finding they're really enjoying it? Any advice you'd offer them?
Guest: Diarmuid Early (08:20):
I think the biggest piece of advice I could offer would be, as the most junior person right out of college in a consulting firm or a bank as just the kind of two that I have the most exposure to, you'll get asked to do a lot of turn the crank, put in new numbers, get the output, build a model, see what it says. And the biggest thing in my experience that separates the technically sound, perfectly acceptable analyst from the one who's doing really well and going to go places fast is they'll both build the model and then the really good one will look at it and ask, "Okay, so what does this mean?" That moment of taking a step back and understanding, interpreting, processing, what is the first question that my manager is going to ask me when I show them this result? Are they going to expect it?
(09:17):
Is it going to be a surprise? If it's unexpected, is it materially off? Is it a sign that something might be wrong in the model that I should dig into? Or is it just, "Hey, this is actually looking better. This is actually looking worse than we thought. Now we need to adjust, whatever." The challenge is that those jobs often work very long hours. And so if you finish building the model at two o'clock in the morning, it's very naturally tempting to be like, "Okay, I'm done. Here. You figure out what's next." And that's also where if you can be very good technically that you get it done faster and you free up half an hour to step back and think, then there's a real value to that. But that last bit, I mean, obviously that's where the commercial value comes from is like, okay, so what are we going to do with this?
(10:07):
What does it mean? But getting attuned to that side of things at the same time as you're thinking about what are the nuts and bolts? Have I balanced my balance sheet? Have I put this in the right place? What is this telling me? Is it interesting? Is it valuable? Is it surprising? Where are we going to go with this?
Host: Paul Barnhurst (10:25):
Yeah, I think that's great advice. It's the thinking beyond just building the model. And it's something one of my college professors, he almost graded us more on our writeup than our model. And we had to do a full five forces porters on every writeup and justify all our assumptions and where the business and industry was. And I so appreciated that. He wanted to make sure we were sound in our model, but he didn't really care what the end number was. Did you make sound assumptions of where you're really thinking business? And so grateful for that approach in grad school versus the, "Hey, did you tie out to the exact NPV value I wanted?" Which, okay, yes, you have to do the math right, but there isn't an exact number in the real world. Yeah.
Guest: Diarmuid Early (11:08):
Yeah, absolutely.
Host: Paul Barnhurst (11:08):
Unless you're doing the World Cup.
Guest: Diarmuid Early (11:12):
Yeah. But I mean, I often tell people it's the biggest difference between competition round the financial modeling World Cup or something like that and the real world is in the Financial Modelling World Cup, you start with a couple of pages that lay out all your assumptions in exact detail, everything you'll need to precisely model it. In the real world, by the time you get to that, you're practically done. The hard part is getting everybody aligned. What are the assumptions? What's the kind of data that's hidden somewhere in the background or that somebody has on a spreadsheet they haven't told anyone about or whatever? Getting to that clarity is the hard part and it's the value-adding part. I've always said that if I had to choose between someone who could get you to that couple of pages of assumptions versus someone who could execute modeling that, I would take the first person in a heartbeat.
(12:00):
Probably all the more so now because if you have that very crisply done, I don't think we're quite there yet, but you're at least getting toward a time where more of that execution can be done by AI. Whereas that is not the case for all the cat herding that leads to, okay, we're all on the same page now. These are credible assumptions, et cetera.
Host: Paul Barnhurst (12:22):
I'm 100% with you. If I can get somebody that can do all the assumptions, can tell the story really well, even if I had to have someone else build the model and they take it and review it if they weren't good at the model. You want both, but see assumptions, the storytelling, the modeling part is technical and you have to be able to do it. But if you can only do that, you're missing the biggest, most important parts of the job. Yeah,
Guest: Diarmuid Early (12:45):
Totally.
Host: Paul Barnhurst (12:46):
And the part that AI will continue to get better at. A lot has happened since last time we interviewed you. I think you were, I want to say my third interview, if I remember right. So three years ago now, roughly, that first interview. And let's start with you won the modeling World Cup last year. How'd it feel?
Guest: Diarmuid Early (13:04):
It felt pretty good. It felt pretty good. I had been locked out a couple of years in a row. Lawrence Lau had beaten me to a two years running, so it was nice to get it back. And
Host: Paul Barnhurst (13:18):
Then you also got the Excelesports Championship. And let's see, you won Europe as well. So you're on a roll right now.
Guest: Diarmuid Early (13:25):
I suppose I am. I'm definitely on a hot streak. We'll see how long it holds up, but yeah, I can't complain. Vegas is a lot more sort of thrilling to win because the financial modeling one, the one I won did actually, which it doesn't always came down to the last round. I went from third to first in the last round. It's not always that close because it's league style rather than knockout style. So it's like your best scores over the whole year. So often, even before the last round has been played, you already know who's going to win, which makes it a little less thrilling. Whereas in Vegas, it's literally 12 people walk in and a half hour later, one of them walks out as the winner and nobody knows what's going to happen until in the moment, which is definitely. And there's a whole live audience and everything.
(14:10):
So that's a lot more thrilling.
Host: Paul Barnhurst (14:12):
Yeah, having the audience, I imagine the eSports, there's a lot more adrenaline in the moment than something you do on your own and then you find out how you perform. Yes,
Guest: Diarmuid Early (14:21):
100%. 100%.
Host: Paul Barnhurst (14:23):
One's more like taking a college exam where you wait to get the answer and hope you did well. And the other one is more like a sports competition, hence the eSports side of it.
Guest: Diarmuid Early (14:32):
Yep. Yeah, exactly. Sprint for the finish. Hope you don't trip.
Host: Paul Barnhurst (14:37):
Oh yeah, I trip plenty, but that's another story. So I noticed you started offering coaching services recently, and I thought it was kind of fun. You shared that on LinkedIn with a video you had made with the CEO of Microsoft. So maybe talk a little bit why you started offering services. Then tell us about the video experience, a little bit about the video.
Guest: Diarmuid Early (14:59):
Sure. I mean, the switch to services is the official version is it's all about where I can add more value for clients, et cetera. The reality is it's like, I think I can add value in this way or in that way, but this one is a lot more flexible. If you're working on a model that someone is imminently using to raise money from investors, it's full on. And the older I get and the more time I'm trying to spend with my kids and all, the less I want to work on something that's like, "This is on fire. I need it right now." I've cut a lot of that kind of work out of my portfolio over the last few years. And so yeah, the thought is basically I can still be helpful. The way I think about it is when I worked at BCG, there was an office of 150 or so in London at the time, and everybody knew that I was the Excel go-to guy.
(16:01):
And my last six months there, because people mostly travelled Monday through Thursday. Every Friday, there'd be a bunch of people coming by my desk over the course of the day looking for help to get unstuck with something. And so the way I think about it is the value to a company of having one person go from quite good to exceptionally good.
Host: Paul Barnhurst (16:27):
I'm super excited to share this event with you. On September 24th, FMI, the Financial Modelling Institute, will be hosting a free global financial modeling conference. It'll be online running for 24 continuous hours. There'll be speakers at every hour of the day all over the globe. So wherever you are based, you'll be able to join during your working hours or in the evening if you choose. You'll be able to select the sessions most relevant to your role and learn from professionals across the global modeling community. Giles Mel, Inschnor, myself will be speaking about our Mod Squad experience of testing AI tools. Many other famous modelers will be speaking at this event. So really encourage you to take the time to attend. Where else can you learn from the best in the world? You can go ahead and secure your free spa at fminstitute.com/conference. I look forward to seeing you there.
(17:22):
In
Guest: Diarmuid Early (17:22):
Terms of the multiplier effect that they can have on everybody else's productivity, this one person can help 10 other people get unstuck and have a huge amount of leverage. And also the person who is the best to start with is probably also the person best positioned to very quickly learn a lot from targeted coaching. That's my theory. I think it's very additive. And for me, it's much less panicky.
Host: Paul Barnhurst (17:50):
Makes a lot of sense. Versus the company that, "Hey, we got this deal and if we don't have the model by Thursday, the deal won't close on time." And you're like, "Yeah, let's say I'm going to be with my kids. I'm working until three in the morning."
Guest: Diarmuid Early (18:02):
Yeah. And by the way, it's midnight on Wednesday right now.
Host: Paul Barnhurst (18:06):
Versus the consultant. I mean, how often do they have an on-fire training? No, you got to be here till nine o'clock tonight. Really? Makes a lot of sense just from a family standpoint. Okay. And then the second, you recorded a video with the CEO of Microsoft. Go ahead and tell the audience first what the video was, how it came about.
Guest: Diarmuid Early (18:25):
Sure. So Microsoft had a much bigger presence at the last Xcel World Championship final in Vegas. So they had this on the way into the eSports arena, they had laid out this whole green carpet and a hall of fame, and they had people there doing interviews and all kinds of things. So they posted, Lizzie, one of their social media people interviewed me right after winning, and they posted that on social media. And Satya replied to that on Twitter saying, "Hey, this is so cool. Why don't you come to Redmond, meet me, spend the day with the Excel team?" And then someone told me through a back channel, oh, he actually means this. I'm like, "Oh, cool."
Host: Paul Barnhurst (19:03):
Yeah, because you're wondering, is that just social media talk?
Guest: Diarmuid Early (19:06):
Yeah, exactly. Exactly. I'm not in Seattle on a regular basis, but I was going to be there for the MVP summit in March anyway. So we sort of tagged it onto that. And yeah, got to meet Satya, walk him through solving one of the bonus problems from last year's final, and just chit-chat about Excel and life and that kind of thing. And then also got to spend the day with a bunch of the project managers on the Excel team, which was probably more exciting for me, frankly, no offence to Satya, to be with the people who are in the weeds building it. But both parts were fun.
Host: Paul Barnhurst (19:41):
Yeah, I've interviewed Brian Jones before, one of people that's key there to building it and about AI before, and so how they're thinking about Copilot. What I thought was fun is that you used that video when you mentioned you were going to start doing consulting on LinkedIn. I thought it was the perfect video to have the CEO of Microsoft's, "Hey, let me teach you about Excel as you're teaching him." So that was well done. From a marketing standpoint, I give that one an A+, the timing. I know you're not a big marketing person, so I know that was hard for you.
Guest: Diarmuid Early (20:10):
I have a very uncomfortable relationship with marketing of any kind. So yeah, it was very much like, "Okay, this thing is coming out. I'm going to get a lot of attention anyway. I might as well take this time to mention the new business thing as well."
Host: Paul Barnhurst (20:27):
Yeah, no, I totally get it. I do a lot more marketing now than I ever did, and there are still parts of it. I just can't do it. Not going to happen. I think I'm much more comfortable than you, but I can relate in my own small way. So love that. I thought that was a great way to start it with that video, and I thought it was a great video. So if anyone hasn't seen it, go check it out. I'm sure you can find it online, YouTube and LinkedIn and so forth. All right, so I want to get your thoughts on an area, and I think you shared a little bit of it when we talk in the Career Corner section, but just your take so far on AI, building financial models, maybe what testing you've done, where you're at, how you think about it.
(21:08):
And it'll all change next week when some new model comes out, but I would still love your thoughts today. It's
Guest: Diarmuid Early (21:13):
An interesting one for me as a topic. And I will say I have done very little Excel with AI. I explored it, I guess about this time last year was the point where AI was starting to get quite good at the eSports side of things and was still fully permissible in the competition. And there were a couple of rounds in a row that were won by people openly letting people know they were using AI. Not co-piloting with AI, but fully solved by AI. I did a little bit of testing at the time. Gemini seemed to be the one that was particularly good with writing big complex formulas. Not, I think, the kind of thing you would want to put in a financial model, but I think there's something about single cell formulas are much more akin to coding that those models have been very, very heavily trained on compared to the whole big network structure of a financial model.
(22:12):
And so it was remarkably good at writing crazy formulas to solve everything, basically a programme in a cell. Yeah,
Host: Paul Barnhurst (22:21):
Writing a big, huge Latin Lambda or whatever, that big combo where you're. Yeah, a mini programme in a cell is a good way to put it.
Guest: Diarmuid Early (22:28):
Yeah. Yeah, exactly. And I explored it just from a perspective of, from what I had heard, it sounded like a person without any particular value add on their prompting could complete around just using AI without really adding much themselves. And there's a whole debate at the time about should AI be allowed? Is it more a function of skill by the prompter? Is it other things? I think they banned it almost a year ago now, and I think the time since then has proven that they were right to do so because the fastest way to solve with AI now would basically be to point AI at it and say, go do your thing.
Host: Paul Barnhurst (23:08):
Yeah, yeah. It would just be loaded to your agent and say, read the instructions and solve it.
Guest: Diarmuid Early (23:12):
Yeah, exactly. Exactly. So I played around with that a bit. I have done very little modeling with AI, mostly because I keep tabs, including through your Mod Squad podcast and the MVP discussions and various other things. I'm in a lot of groups that are keeping close tabs on it, and I feel like I have a decent sense of where it is. And I don't think it's quite there for me yet. It's probably not too far away, but I don't think it's quite there for me yet. So
Host: Paul Barnhurst (23:43):
That made sense. Where you ended there makes sense to me. You had said it's not quite there for me yet, which that's kind of how I feel too is although I'm at the point where I'm not anywhere near as accomplished a modeler, but in some model, certain things it can build. I'm comfortable. If I have a basic deferred revenue schedule, I'm comfortable giving that to AI and then coming back. But complex stuff, definitely not there. It depends on what you're doing and how much you're willing to, from what I'm hearing, break it apart. There are certain things that if you're doing a lot and you're willing to write the skills and give it good examples and do all the instructions, it could probably save you 80% of your time for the average modeler. But this idea of turn it loose, come back five minutes later and have a model, just I cringe every time I see that.
(24:31):
I've
Guest: Diarmuid Early (24:32):
Seen a few claims of like, "Hey, look, this thing can build a 10 tab model in three minutes." I'm like, "It could not be more clear to me when you say something like that, that you're looking at the wrong metrics. Is the model correct? Does it have anything useful to say? No, no, it has 10 tabs and it was ready in three
Host: Paul Barnhurst (24:51):
Minutes." I am so with you. The one I like to joke is build a financial model in five, get fired in 10. Yeah, exactly. If you didn't spend two hours in between reviewing it all, it's just a matter of time until you're going to get burned. There is a difference between editing a template and building a model from scratch. The advanced financial modeler tests whether you can actually build a model from scratch. What are you waiting for? Get started on your AFM journey today. Go to fminstitute.com/podcast. That is fminstitute.com/podcast and use code podcast for 15% off.
Guest: Diarmuid Early (25:41):
This does absolutely drive me crazy with the way AI gets positioned. There was one being advertised on the subway in New York recently that was like, "Compare the workflows. Here, whatever, get some data, build a pivot table, add formulas." One of the steps along the way was review results and then done. And then the other one was one prompt, done. Okay, so you're saying that you don't review the results.That's very, very dangerous. And
Host: Paul Barnhurst (26:14):
You didn't build any skills or give it any examples. You really did it all one prompt with no context. I
Guest: Diarmuid Early (26:19):
Am wary of, I think for me as a sort of professional financial modeler, I think there's a risk of being too knee-jerk to say, "Oh, well, this one person made a mistake with AI, therefore AI is not the right way or AIs have done all these silly things." And they have, but so have humans. I am trying very consciously to not end up too much in a position of like, "Well, I could have done it better." So for context, a couple of friends and I are starting a business. It's kind of low-key at the moment, but there's a bunch of legal paperwork and stuff like that with starting a business, issuing shares, et cetera. And so it's been interesting for me to have this experience going through all this kind of legal stuff where I'm sure the lawyer equivalent of DIM would say, "If you use AI for your legal work, you're exposing yourself to all kinds of risk." Would
Host: Paul Barnhurst (27:19):
That be Connie?
Guest: Diarmuid Early (27:20):
Wow, she's a consultant, not a lawyer. Oh, for some
Host: Paul Barnhurst (27:25):
Reason I thought she was a lawyer. Why did I think she was a lawyer? I don't know why she's a consultant. Okay.
Guest: Diarmuid Early (27:30):
No. Yeah, my brother's a lawyer, but he's not my - There you
Host: Paul Barnhurst (27:34):
Go. They're your brother. I
Guest: Diarmuid Early (27:34):
Don't think he wants to be my lawyer. But yeah, the way I think about it is I think that the lawyer equivalent of me would say, "You should have lawyers do all of this stuff because there are nuances, there are things you could get wrong," and so on. And I've given it a lot of thought from that perspective. And where I've ended up is I do use AI a good bit for the legal stuff, but - I do too.
Host: Paul Barnhurst (28:01):
Usually
Guest: Diarmuid Early (28:02):
I actively tier things in terms of there are some things that are purely for box ticking purposes. So for instance, the bank needs a corporate resolution saying DIM isn't authorised to open a bank account. As long as it ticks their box, nothing else matters. AI draughts, I read, I sign, I'm happy, off it goes. There's not a lot in that box, but there's some stuff there. And then there's the next level of stuff, which is it's important, but it's also quite standard. And what I tend to do there is I ask AI to find me a reputable source for a public template. And then we talk about, is there anything this is missing? Is there anything that's particularly founder unfriendly or investor unfriendly or any red flags or whatever? And then I will take its thoughts, read it myself, consider any adjustments needed and make them myself.
(28:58):
But it's mostly starting from a template rather than starting from something that feels mostly the same vibe as a template, and that feels safer. But that equivalent doesn't exist as much in the financial modeling world because I don't think there are as many. When people build good models, they keep them secret. I don't think there is quite. There are some template sellers out there, but I don't think there's the same thing of you start from this template and then whatever, but maybe. And then there's the third tier of things. They're like, "This is important and could go wrong and needs to be customised to our company, and we're just going to get a real lawyer to do it." And I think no matter how good AI gets, I don't think it would ever get to a stage where I would have AI alone do critical legal documents or do critical financial models or anything like that.
(29:56):
But I think the share that it takes up and Where you put that boundary between them will keep moving. But I think we're a long way from, "Hey, we're going to make this $15 billion acquisition because the AI says the price is right."
Host: Paul Barnhurst (30:13):
Yeah. I think you make a lot of good points in there. And I think the lawyer example and the legal is a really good example. And just that reminder, it's probabilities. How many of us are going to invest billions of dollars strictly on probability without human judgement before AI? Nobody? So why would we do it with AI? Yes, it's very good and it's narrowed down those probabilities and it's been trained on so much that 99% of the time it might get the right next paragraph or the next formula or whatever. But it's not deterministic. One plus one does not always equal two. It's a matter of time till it equals three or five or seven or whatever. So I'm with you. I think that's a good way to look at it. So I'm curious, what is a kind of a best practise standard you hear out there that maybe you think is overrated or that you're like, why do we always hear that?
(31:09):
Again, I see you smiling, which means I know you have one.
Guest: Diarmuid Early (31:11):
Well, I think I may have told you my dirty secret before. I don't know if it was on the podcast or just in person.
Host: Paul Barnhurst (31:16):
I think you did. You mentioned you don't follow a formal standard.
Guest: Diarmuid Early (31:20):
Yeah. I mean, it's not even that I don't follow that. I don't think I actually know any of them. I know the kind of general rules that people start for all of, you separate your inputs and your outputs and your calculations and that kind of thing. But I know that there's fast and I know that there's, oh God, I'm even forgetting the name. Smart and smart come to mind. I was going to say smart and then I was like, no, hang on. You totally just made that up. I don't know. Anyway, I know that they exist and I think that I know the general vibe of them, but I've never actually studied the details of them. So I have no idea if they're overrated. What I've heard -
Host: Paul Barnhurst (31:58):
Maybe we'll go with guides or things you hear from people that say you should always do this with the model. And you kind of go, really? Is there one of those that you see a lot of people imply just in the general rule? So ignore the detailed standards.
Guest: Diarmuid Early (32:11):
So I guess for me, the positive that I've heard about the modeling standards is that it's good for team development. As in if you're going to have a team work on different pieces of a model and put it together, or one person builds this part and hands it over to another to do this part or so on. I can see the appeal there of saying we don't just want people to be doing things approximately the same way. We want in a sort of strict sense, a standard. And I guess in many ways that has never felt hugely applicable to me because back in my employee days, I was basically always the exclusive owner of the files I was working on. People would take data from them and reports from them and extracts from them and I would show people things in them, but nobody else was touching them until I handed it over and left.
(33:06):
And obviously that just means I don't need to think about collaboration features in the way that some people do. I don't need to think about standards and what could get lost at gaps in handovers and that kind of thing in the way that some people do. So I'm not against the idea. My sense is that I would find them excessively rigid, but there is benefit to having rigidity if it means that everyone is going to be moving in lockstep.
Host: Paul Barnhurst (33:36):
FP&A guy here. One of the certifications I'm most proud of is the Advanced Financial Modeler or the AFM from FMI Institute. Why am I so proud of it? Because it taught me how to be a better modeler. I didn't just take a test. I had to demonstrate that I could build a fully integrated three-statement model, a skill that is invaluable for financial modelers. It's one of the certifications I'm most proud of, and I encourage you to earn it as well. So go ahead and join me. Get started on your journey to becoming an AFM charter holder. Go to fminstitute.com/podcast and use code podcast for 15% off. Yeah, it's one of the reasons we have so many different modeling tools. They all give up some flexibility, whether it's planning tools for FP&A or other modeling tools that are more structured in the name of rigidity to try to prevent certain errors or make certain things easier.
(34:41):
Everybody has to decide whether it's worth it. But when you need ultimate flexibility, you always come back to a spreadsheet, whether that's Excel or Google Sheets or some other alternative that's out there. At least that's what I've always found. As soon as you need that ultimate flexibility and you'd have to make a change to get the system to work, everybody just goes back, where's my spreadsheet? There's the reason why it is what it is. All right, so this is kind of fun one. You've done hundreds of cases, I'm guessing at this point, if not more. So let's talk e-sports. Least favourite case. Do you have one that still gives you nightmares or that you kind of hate? You're already laughing. That's a yes.
Guest: Diarmuid Early (35:19):
I mean, it still gives me nightmares. The one that leaps to mind is Eve Online, which was a case based on the computer game, Eve Online, which was the finals, the first live final in Vegas in, I guess, 2023. A very kind of rich world-building case. They kind of worked with Andrew, the founder, worked with this guy, Neils Terfer, who was a top Eve player and YouTuber, I think, to build this case that was very rich in detail and kind of true to the game and so on. But the result was we didn't see anything until they said go, the 30 minutes was counting down. And there were tonnes of instructions to understand. And there were, I think they have upped their quality assurance game since then. There were mistakes and things in it as well. But so the result was here is a tonne of information to digest.
(36:19):
Also, whoever's in last place in five minutes is getting eliminated. Off you go. It's just incredibly stressful. And yeah, trying to manage the different units like, okay, this is the name of a type of asteroid and this is the name of a type of mineral and this is this. It was just an information overload in a short space of time, and that still kind of traumatises me.
(36:46):
They did post-match interviews with the people who got eliminated. And I think Michael Clark was maybe the second one out after 10 minutes. And Katrina said to him, "So what went wrong?" And he was like, "I still have no idea what I was supposed to be doing up
Host: Paul Barnhurst (37:03):
There." A lot of people, I think the first time they ever do a case, I still remember Carl Seideman telling me, he's like, "I still don't think I understand the instructions. The first case he hit." And you know what I'm talking about. There are just some of those cases, especially your first start that you're just like, "What do I hit?"
Guest: Diarmuid Early (37:22):
Yep. Yep. I
Host: Paul Barnhurst (37:23):
Hear you. I'm glad that you're human like the rest of us. You have those cases. And obviously you powered through and I know you placed somewhere in there, third or fourth, I think that year, if I remember right. Or somewhere in the middle. Yeah, fifth
Guest: Diarmuid Early (37:34):
Maybe. Yeah. I
Host: Paul Barnhurst (37:35):
Think it was somewhere in that. I remember you got past a few rounds, but I couldn't remember how far.
Guest: Diarmuid Early (37:40):
Yeah. I mean, it's interesting. Since then they have, I think possibly based on my strongly worded feedback from that case, they've taken to giving us a partial writeup of kind of like how you see the case animation video before you start in most of the rounds now. In person in Vegas, backstage, they will give the people who are about to go out a page or two, or in one case, seven densely typed pages of information.
Host: Paul Barnhurst (38:09):
Was that world of warcraft?
Guest: Diarmuid Early (38:10):
It was. How did you know? It's like, here is the manual for the case. Please memorise it and. You go f minutes. Good
Host: Paul Barnhurst (38:20):
Luck.
Guest: Diarmuid Early (38:21):
Yeah. Yeah. I mean, it's one of the things that I think is the real art of making a good challenge. You can always make things challenging by adding in seven extra rules. I think last year's finals case was a great example of a sort of natural challenge. The instructions for that were two sides of one piece of paper and more than 50% of that real estate was a couple of pictures to illustrate the idea. It was like a paragraph and a half of text. It was just really, really simple to express and just quite complex to execute. And it was very nicely done.
Host: Paul Barnhurst (39:02):
As I always like to say, complex is easy, simple is hard. Getting those instructions to be simple when it's a very complex task is really, really hard.
Guest: Diarmuid Early (39:11):
Yes. Yeah, yeah. And I mean, to me, the best case writers are very much artists in my mind. They're incredibly creative. Yeah, I'm perpetually amazed at what they come up with.
Host: Paul Barnhurst (39:28):
Yeah, no, I believe it. All right, so I got two more questions, then we're going to do rapid fire again because most people won't remember your answers from first time. And there'll be somebody that will check them to make sure whether they've changed or not. So we'll get to that in a minute. But what's your favourite Excel shortcut if you had to pick
Guest: Diarmuid Early (39:45):
One? Oh my God, pick one. Just one. Okay. Well, I suppose now it's probably Control Shift O, which is the launch code for OA Robot, which it's like wishing for more wishes. If I have that keyboard shortcut, then I don't need any others.
Host: Paul Barnhurst (40:01):
All right, so we got that one. Now let's go with the normal one that's not a custom one.
Guest: Diarmuid Early (40:05):
I guess maybe old ESV. I paste special a lot. I made a video years ago, not long after I started my channel. I had a bunch of comments requesting like, "Hey, what keyboard shortcuts are you using?" And I made a video where I built one of the simple models and consciously tried to make a note every time I used a keyboard shortcut. And in a half hour, I used at least 60 different keyboard shortcuts, many of them many times. I used tonnes, so don't know.
Host: Paul Barnhurst (40:35):
I'm surprised to hear you say all the ESV, which means Control C has to be up there too.
Guest: Diarmuid Early (40:41):
I use Control C a tonne. Because
Host: Paul Barnhurst (40:43):
It's the other side of that one. Exactly. All right. So what's the most unique or most fun thing you've built, personal life, work, whatever, using a spreadsheet?
Guest: Diarmuid Early (40:53):
So a couple of candidates for it. One is a TV company approached me years ago when I lived in the UK to build a financial model to figure out the expected payout for a game show that they were designing and preparing. It wasn't the most complicated model, but it was quite an interesting one. A whole big Monte Carlo thing ran a million simulations and did probability distributions and so on. But it was quite cool because they didn't end up commissioning it, but it got far enough that they took over a studio and recorded a pilot with Carl Vordeman as the host. And I got to come along and be backstage. And that was very cool. That was very cool.
Host: Paul Barnhurst (41:38):
RapidFire incoming. 20 questions. Yes or no answers. No hedging aloud. Circular references, VBA, the future of Excel, and more. Let's find out what our guest really thinks. All right, so we're going to go to rapid fire. As you know, it depends. You got to pick a side on these. And I've added a few more since last time we did it. So there'll be a couple new ones, but there'll definitely be some similar ones. Circular references.
Guest: Diarmuid Early (42:07):
Do you even have to ask? No.
Host: Paul Barnhurst (42:09):
The answer was no for anyone wondering. I don't have to ask. VBA, yes or no?
Guest: Diarmuid Early (42:15):
Yes.
Host: Paul Barnhurst (42:17):
Lambdas in financial models, yes or no?
Guest: Diarmuid Early (42:19):
Yes.
Host: Paul Barnhurst (42:21):
External workbook links?
Guest: Diarmuid Early (42:23):
No.
Host: Paul Barnhurst (42:24):
I knew you'd say that. Should modelers primarily use keyboard shortcuts? Yes or no?
Guest: Diarmuid Early (42:30):
Yes.
Host: Paul Barnhurst (42:30):
Okay. Should models always be print ready?
Guest: Diarmuid Early (42:34):
Yes.
Host: Paul Barnhurst (42:35):
Is there ever a situation where merge cells are acceptable?
Guest: Diarmuid Early (42:38):
Yes.
Host: Paul Barnhurst (42:40):
All right. I think you might be the first one to have said yes yet. Most people are like, no. One was, use merge cells, go to hell.That was his motto. So should financial modelers learn Python in Excel?
Guest: Diarmuid Early (42:54):
No.
Host: Paul Barnhurst (42:54):
Power Query?
Guest: Diarmuid Early (42:56):
Yes.
Host: Paul Barnhurst (42:57):
Power BI?
Guest: Diarmuid Early (42:59):
No.
Host: Paul Barnhurst (43:01):
Okay. Should every financial modeler, does every financial modeler need to be able to build a fully integrated three-statement model?
Guest: Diarmuid Early (43:09):
No.
Host: Paul Barnhurst (43:11):
You though about that one for a second. Will Excel ever die?
Guest: Diarmuid Early (43:15):
Yes.
Host: Paul Barnhurst (43:17):
How long?
Guest: Diarmuid Early (43:19):
Well, the way I think about it is it'll probably die, but probably not before I do.
Host: Paul Barnhurst (43:24):
I had someone say, yes, it will die. I'm just praying it's not before I do. Basically. So same idea. All right. What financial statement is most important for modelers? Income, balance, or cashflow?
Guest: Diarmuid Early (43:39):
I'm team cash is king.
Host: Paul Barnhurst (43:42):
Alrighty. Favourite LLM? Do you have a favourite you like to use? Cloud, Copilot, ChatGPT, other?
Guest: Diarmuid Early (43:49):
I have a soft spot for Gemini.
Host: Paul Barnhurst (43:50):
Okay, that works. If you could pick only one for all of your models, which one are you picking? Sensitivity analysis or scenario?
Guest: Diarmuid Early (44:00):
It's got to be scenario.
Host: Paul Barnhurst (44:01):
Do you believe financial models are the number one corporate decision-making tool?
Guest: Diarmuid Early (44:06):
I think I took the no on this one in the great debate. All right. Well, there's no it depends, so I'm going to say no.
Host: Paul Barnhurst (44:12):
Yeah. My favourite answer I've got all the time is no, it's politics.
Guest: Diarmuid Early (44:16):
That's pretty much my feeling. I think that in a perfect world, they would be. I think in reality, the vast majority of times that financial models are used to make a decision or to justify a decision, the decision is made before the model is built and the model is instructed to match the decision.
Host: Paul Barnhurst (44:35):
Yeah. I still remember one where I asked my boss, the deal fell apart, but I go, "So what was the deal like?" He goes, "Oh yeah, the model was sprinkled with rainbows and unicorns to make the math work." And they ended up doing the deal a couple years later when that CEO left. The private equity wanted to do it. They just couldn't get him to agree at the time and they still rushed it through a few years later. And that company's really close where I expect a restructuring bankruptcy in the next year or two. So we could see how well that deal worked.
Guest: Diarmuid Early (45:04):
The world would be a better place if a strong, independent financial modeling individual had more authority than they typically do in making these kinds of decisions.
Host: Paul Barnhurst (45:15):
I 100% agree with you. What's your lookup function of choice? I'm not going to give you options. I think last time you said to me, "Well, I question if those are the only options."
Guest: Diarmuid Early (45:23):
I mean, if I'm allowed to wish for more wishes again, I actually have a custom Lambda that rolls my favourite bits of many of it. It does kind of lookups, it does filters, it does Python-style slicing, does tonnes of things. So that's my favourite if I get to wish for more wishes. Otherwise, I guess I'll say XLOOKUP.
Host: Paul Barnhurst (45:42):
I'm going to have to join your consulting to learn more one of these days. You never know. I might just be like, "Okay, so how much? What is it?" And then I'll be like, "Yeah, maybe not." I'll go back to content creation. I'm better there. So before we finish, any of those you want to elaborate on?
Guest: Diarmuid Early (46:01):
I mean, I guess the only one I would elaborate on is all the learning ones. Should financial modelers learn Python? Sure. Most people should learn Python. Is it necessary to be a financial modeler? Probably not. But is it a good thing to explore? Should you learn Python? Should you learn Power BI? Yes, always keep learning more things. And similar with the three-statement model, would ideally all financial modelers be able to build a three-statement model? Absolutely. It's just the world is sufficiently niche and siloed that I'm sure lots of people do perfectly good financial modeling jobs without ever touching a three-statement model.
Host: Paul Barnhurst (46:37):
Makes a lot of sense. And I'm with you. I didn't touch my first three-statement model till I started my own business four years ago from a work perspective because it wasn't the type of models I built. I'm very much with you on that. I love your approach. Yeah, we should all be learning. Is it necessary? No. Is it good? Yes. Which are two different questions. So any last piece of advice or any parting words before we let you go?
Guest: Diarmuid Early (47:01):
My Excel advice always has been and remains there is a smart way to do it. And so if you're banging your head against it, go find the smart way.
Host: Paul Barnhurst (47:12):
All right. So go ask AI.
Guest: Diarmuid Early (47:15):
I mean, honestly, it's not a bad way to find new things.
Host: Paul Barnhurst (47:19):
In some situations, it really is. I 100% agree. I've had to solve several things. I'm banging my head on what did I do wrong in this formula? And AI will pick it up in 10 seconds. It's really good at where did you make a mistake in code?
Guest: Diarmuid Early (47:32):
Yeah. And I think it's also, I worry a little about coming to rely too much on it. But what I think is quite good for is if you think about. There's a word for it that I can't think of now. It's like the most reasoning that we're trained to do as adults is deductive. You reduce down to this is the answer, as opposed to what are a thousand things you can do with a paperclip? Think of all the different possible. I find AI is quite helpful sometimes when I'm like, okay, I've got two or three ideas here, but there must be 10 plausible ideas for this. I do worry a little that I'm coming to rely on it too much and maybe therefore weakening my own just sit and think about it for 10 minutes and come up with more ideas, goddammit.
Host: Paul Barnhurst (48:21):
Thank you for saying that because I feel the exact same way where there are times I'm like two minutes in and I'm going to AI and I'm like, am I just depending on this thing too much? Stop and think for a minute.
Guest: Diarmuid Early (48:33):
Yeah. I mean, I feel like it's something that is going to be increasingly important to remind ourselves of as it keeps getting better and more ubiquitous. 100%
Host: Paul Barnhurst (48:41):
Agree. Well, thank you so much for joining me. It's always a pleasure to chat with you. I hope to see you here in a few months in Vegas and best of luck with the competition.
Guest: Diarmuid Early (48:51):
Thank you. Thanks for having me and I'll see you soon.
Host: Paul Barnhurst (48:54):
Financial Modeler's Corner was brought to you by the Financial Modeling Institute. This year I completed the Advanced Financial Modeler Certification and it made me a better financial modeler. What are you waiting for? Visit FMI at www.fminstitute.com/podcast and use code podcast to save 15% when you enrol in one of the accreditations today.
