The Analytics Lessons FP&A Leaders Need to Build Better, Repeatable Workflows with Ned Harding
In this episode of FP&A Unlocked, host Paul Barnhurst sits down with Ned Harding - a true pioneer in the data analytics space - to talk about why most data tools still don’t meet the needs of today’s collaborative teams. From building Alteryx from the ground up to launching a new venture (Enso Analytics), Ned shares his take on what’s broken in data workflows and how FP&A teams can fix it.
Ned Harding is the Chief Product Officer at Enso Analytics, a platform that helped define the self-service analytics category. He started coding at eight years old, taught himself from PDP-11 manuals, and has spent the last two decades creating tools used by hundreds of thousands of analysts worldwide. He’s a product guy through and through, and he’s on a mission to help teams work smarter with their data.
Expect to Learn:
Why most data tools fall short when it comes to real team collaboration
The importance of repeatability and testing in FP&A workflows
How to avoid common pitfalls with AI and overfitting in forecasting
Why Excel is both a lifesaver - and a landmine
How to build a true “culture of analytics” across your entire org
Here are a few relevant quotes from the episode:
“The biggest successes happen when teams collaborate - analytics shouldn’t be a solo act.”- Ned Harding
“If your team’s not getting value from your data work, you’re not doing your job.”- Ned Harding
Ned shares practical insights on how FP&A teams can become strategic partners by embracing collaborative, repeatable analytics. He emphasizes the value of integrating data across teams, avoiding AI pitfalls, and applying the scientific method to drive consistent, data-informed decision-making.
Campfire: AI-First ERP:
Campfire is the AI-first ERP that powers next-gen finance and accounting teams. With integrated solutions for the general ledger, revenue automation,
close management, and more, all in one unified platform.
Explore Campfire today: https://campfire.ai/?utm_source=fpaguy_podcast&utm_medium=podcast&utm_campaign=100225_fpaguy
Follow Ned:
LinkedIn - https://www.linkedin.com/in/ned-harding-34a57526/
Company - https://www.linkedin.com/company/enso-analytics/
Website - https://ensoanalytics.com
Earn Your CPE Credit
For CPE credit please go to earmarkcpe.com, listen to the episode, download the app, and answer a few questions and earn your CPE certification. To earn education credits for FPAC Certificate, take the quiz on earmark and contact Paul Barnhurst for further details.
In Today’s Episode
[02:43] – Meet Ned Harding & His Journey to Enso
[06:04] – Accidental Start in Analytics
[10:59] – Turning Tools into a Business
[14:49] – Why Enso Analytics?
[25:43] – Analytics Across the Org
[29:16] – Context is Everything
[35:51] – The Risk of Overusing AI
[43:56] – Are We in an AI Bubble?
[47:07] – Top Skills for FP&A Pros
[49:54] – The Excel Debate
[53:12] – Fun Final Questions
Full Show Transcript
Paul Barnhurst (00:45):
Welcome to another episode of FP&A Unlocked. Are you tired of being seen as just a spreadsheet person? Well, others get a seat at the table. Well then welcome to FP&A Unlocked where finance meets strategy. I'm your host, Paul Barnhurst, the FP&A guy. Each week we bring you conversations and practical advice from thought leaders, industry experts and practitioners. Together we'll uncover the strategies and experiences separate good FP&A professionals from great ones helping you elevate your career and drive strategic impact. Speaking of strategic impact, our title sponsor for FP&A Unlocked is Campfire, the ERP. That's helping modern finance teams close, fast and scale faster. Today's guest is someone who has spent much of his career helping FP&A professionals earned that coveted seat at the table through data analytics. I'm thrilled to welcome to the show Ned Harding. Ned, welcome to the show.
Ned Harding (01:41):
Thank you. Thrilled to be here.
Paul Barnhurst (01:43):
Excited to have you. So let me just give a little bit of Ned's background and then we'll jump into things. So Ned has held various positions in the tech industry, including his current role of chief product officer at Enso Analytics. Previously he co-founded and served as Chief Technology Officer at a Alteryx. He has shaped the self-service data analytics industry for over 20 years. His innovations have scaled to serve hundreds of thousands of users globally driving widespread transformation in data technology. Love the background. Obviously data analytics is a core to what we do in FP&A . So I'm really excited for this conversation. I'm really curious from your perspective when you see FP&A or you get involved with your FP&A professionals from either they're supporting you from a work standpoint or you're serving them as your customer. What do you think great FP&A looks like? What do you think that requires for a company to kind of upgrade FP&A ?
Ned Harding (02:44):
Well, first I just want to say I'm not an FP&A person at all.
Paul Barnhurst (02:48):
That's why I want your perspective.
Ned Harding (02:50):
I am a software person and a product person through and through. But that said, I have built products that many, many, many FP&A users have used from both a vendor and as a consumer of FP&A Unlocked being part of various companies. The thing that I think is most important is to know who your stakeholders are and serve all of them. Sometimes the FP&A person or team gets very myopic and they think that their only stakeholders are the stockholder or the CFO or they have a very myopic view. But really your customer, if you will, are everybody in the company. It's the employees, it's the customers, it's the technical support people, it's everybody in the company. And if you're not working to support everybody, you're probably not doing the whole job.
Paul Barnhurst (03:51):
That's a really good point. Like I said, that of f Penny's really to support everybody, right? It's a support function. At the end of the day, our job is to make the business's job easier, to help 'em make better decisions and give the information they need to do their job well.
Ned Harding (04:07):
Totally. And it's really important for the people in the company to know that what the FP&A team is delivering is predictable and consistent from quarter to quarter because there's nothing worse than getting a budget and then next quarter saying, oh, we got to cut it by 20% and the quarter after, oh, well we found more. We can raise it by 20%. Making sure that it's consistent and repeatable and predictable is so important.
Paul Barnhurst (04:41):
It's never a fun conversation when you're cutting numbers, but it's even worse when you change 'em again and you're going back and forth and there's just no consistency when things are moving a lot. I've seen that happen and definitely it puts everybody in a bad position. So I'm curious, how did you end up in self-service data analytics?
Ned Harding (05:04):
A hundred percent by accident.
Paul Barnhurst (05:06):
I figured that might be the answer. It often is.
Ned Harding (05:09):
I'm a software person. I started programming, I was eight years old on a deck PDP 11 and then a QRS 80 model one way back in the day. And when we started Alteryx, it was actually a company called SRC in the beginning. And data analytics was maybe the slightest part of what we did, but we're mostly in the demographic space. And we started to get customers where we were putting together websites for them to help them with their demographics, help them with sort of some data analytics. And as part of that, we got data from the customer to integrate with demographics and other data sources on a monthly basis or a weekly basis. And we just had a huge problem ingesting customer data and standardising it and cleaning it because it was always dirty. And B, it was always dirty in a different way every single time.
(06:02):
It wasn't dirty in the same way. That would be too easy. So we were using Microsoft Access and we were using handwritten code and all kinds of stuff. So we ended up writing our own internal tools for dealing with all this data. And then we got a contract for releasing the census software inside our software, which is a whole nother story on its own, but in order to load the census data into our software, it's a huge process. And it's not just summary file one and summary file two. It turns out they have, I dunno, a hundred different data sets or something. So again, we're writing all these tools. So the tools mostly evolved out of our own needs and me being lazy, I wanted to make it into a product rather than a one-off. So I drove the productization of the data analytics tools or back in the day at Alteryx, the tagline I liked the most was the self-service data prep and blend. That's always the hardest thing, just getting the data ready for the next step. So yeah, like I said, I got into it by accident, but it was a happy accident.
Paul Barnhurst (07:15):
Yeah, it looks like it turned out well for you. It's been a good journey. So that's always nice. And I like how you said the data prep and blend, I joke, how do you describe an FP&A professional's job? They spend 80% of their time cleaning data and 20% complaining about the data they have to clean.
Ned Harding (07:35):
That sounds absolutely correct.
Paul Barnhurst (07:39):
I can totally understand how you just needed the tool to make your life easier. When you mentioned Microsoft Access, I was like wondering if our audience has even used that tool. Now I live
Ned Harding (07:50):
In that theme. I hope not. I really hope
Paul Barnhurst (07:52):
Not. Oh, I hear you now. I use Power Query to do what I used to do with Microsoft Access typically or something, but it's amazing how far we've come. What was the language when you were eight that you coded it?
Ned Harding (08:04):
When I was eight, I started in basic but quickly started programming and Assembly language and Pascal and then logo, if you remember that with Turtle Graphics and then Lisp, then got into See all probably before I was 12, all of that.
Paul Barnhurst (08:21):
Wow, so you really did love it from a young age.
Ned Harding (08:24):
I really did. And there's something about being programming languages. It's like spoken languages. When you learn it as a kid, you learn it in a different way than you do when you learn it as an adult. And there's just a huge advantage to internalising all of that. And I probably learned how to read Reading the Deck PDP 11 reference manuals. I'm like 8, 9, 10 years old and there's a stack of reference manuals two feet high and I'm loving it. I read every single word. Yeah. Let's see, what
Paul Barnhurst (09:01):
Was I doing at
Ned Harding (09:01):
That age? I was reading the box scores from the
Paul Barnhurst (09:03):
Sports section.
Ned Harding (09:06):
I was doing that too, don't get me wrong.
Paul Barnhurst (09:10):
No, my daughter enjoys programming. She does a weekly class so she can relate to that. I think it's great she's learning it at a young age because like you said, there is just learning a language. There's a lot of benefits when you learn something when you're
Ned Harding (09:22):
Young, even if you don't learn it to an expert level, if you're just exposed to it as a kid, your ability to learn as an adult is enhanced. Twofold, highly recommended.
Paul Barnhurst (09:34):
So you kind of mentioned how you got into data analytics, and I would guess this is kind of similar how you ended up starting Alteryx. It started as SRC. You mentioned I think geographic data census, things like that. But how did you know when you had something with the products you productized that you now had a business that was more around the self-serve analytics, the data prep, those types of things. When did it dawn into you that, hey, this is really the use case here?
Ned Harding (09:59):
Well, that took a while. I mean the first step was finding the right partners. Like I said, I didn't plan on getting into this business, but I knew I wanted to be a product guy. I mean, that's who I am. And by product guy, I mean designing it, specking it, but building it also. But you have to have the right partners. I don't want to do everything. So when we started Alteryx, Dean and Libby had probably had some customers already, but they were amazing salespeople and got a lot of customers. And so the products really got built organically out of having customers first, product second, which was really a unique way of doing it and I think worked really, really well because we didn't have to design the product at all until we knew who the customer was. Whereas so many people try to design a product then figure out who the customer is, and that sometimes works, but it's hard. We had good traction with the early products, with Allocate, with Solocast, with all that, but obviously the product named Alteryx took over the company. I mean eventually we're like, okay, this is the product and let's rename the company after the product. So I think in 2010, that's when we renamed the company. That's clearly when we knew that was the product that was going to take us somewhere.
Paul Barnhurst (11:30):
And how old was the company at fan? I think you said 97 was SRC, is that roughly when that started?
Ned Harding (11:35):
We started in 97. I started writing Alteryx January 1st, 2005. I literally started writing on January 1st because I was so antsy to get going. I had said I'm not going to start until the new year. And then that morning I'm like, I'm starting and released it by the end of January, 2006. And like I said, four years later, we renamed the company after the product.
Paul Barnhurst (12:00):
Got it. So it was several years in before really the kind of product of what it became, the name changes over a decade. It sounds like there's quite a journey with different iterations and experiences there.
Ned Harding (12:14):
It was quite a journey. I mean when we had the IPO, somebody said to us, oh, you guys are such an overnight success because data analytics was so hot and nobody had heard of us before and we're all like, yeah, it was a 20 year overnight success, but you have to do that, right? Maybe there's a few companies that are true overnight successes, but most companies or people in their careers, you only find success from years of hard work.
Paul Barnhurst (12:46):
No, I a hundred percent agree. There's rarely ever such a thing as an overnight success, especially at any kind of sustainable or high level.
Ned Harding (12:55):
100% agree.
Paul Barnhurst (12:57):
There's one-offs. There's someone who just hits the jackpot like the pet rock. It wasn't a sustainable business, right? I mean there wasn't probably picking on the poor rock. But yeah, there are examples
Ned Harding (13:11):
You say that I have no idea what they did after Pet Rock. They might've had 10 other successes for all I know
Paul Barnhurst (13:17):
They could have, but I just mean this sometimes it seemed like overnight where you're like, all right, how much time could it have taken to colour a rock? But anyway, I'm probably going to get myself in trouble and get an email from the founder of Pet Rocks.
Ned Harding (13:29):
I hope you do.
Paul Barnhurst (13:30):
I hope I do too. They could be an interesting conversation. So fast forward to today, you're the chief product officer, and so analytics, you've moved on from Alteryx. Tell our audience why you thought another data tool was needed. How did this come about that you're back in a startup again, building another tool?
Ned Harding (13:49):
There's different answers for different parts of that question. Why am I back in a startup? Well, I love working with smart people and so analytics has some of the smartest people I know, and I love that why is there need for another data tool? Because the data tools that are out there don't do what I think they should do. Alteryx was great Power Query, all these are great, but they really solve two different use cases. They either solve the lone Wolf individual trying to get a task done, or in some of the other cases they solve it controlling what can be done. But what they don't solve is a work group working together collaboratively. And that's where the biggest successes happen is when you have a team of people that you get the best of everybody. And so I want people in a self-service way, not it driven.
(14:55):
Obviously it can be involved but not it driven to be able to work together to have as many strengths as they have as a team. So if one person knows financial modelling and another person understands reporting, but then the third person understands actually what to do with the data, and then you have a fourth person who is good at operationalizing it and making it happen every month. If we can get all them working together, it's so much better than each person being in their own silo. And so is designed around teams from the start. It's not designed for individuals, it's not designed for it. It's designed for teams of users. And then the other really unique thing is it's designed hybrid from the start. Alteryx historically was primarily desktop, which is really important because that allows you to keep all your data behind the firewall, allows you to use really fast hardware that is much cheaper as local hardware than it is in the cloud, which I know surprises people and that's great, but then you don't have any collaboration because it's all on the desktop.
(16:15):
And then you look at other stuff like Snowflake or Databricks, that's all about putting your data in the cloud. A lot of times you don't want your financial model to be in the cloud at all because this stuff gets leaked. How often has Snowflake leaked all the data from at and t leaked all their customers via Snowflake? And I don't care how good a cloud provider is, if the data's in the cloud, it has a high chance of getting leak. So Enso is designed around the concept of hybrid. You can keep everything in the cloud if you want. You can have all the execution in the cloud if you want, but you can also keep some of your data in the cloud, some of the data on your desktop, keep your workflows in the cloud, versioned available to your whole team, auditable by your whole team, auditable by it, but at the same time reading data that's purely behind the firewall or even on your local machine and never exposed to the cloud at all. So that hybrid access is really important going forward and hopefully eventually the cloud will get secure enough that people can trust it. But I feel like every week I'm reading about a leak somewhere or another. So I don't trust it
Paul Barnhurst (17:38):
Mean you don't like having your data spread everywhere.
Ned Harding (17:41):
I mean, I guess personally I'm just resigned to the fact that it is spread everywhere. I don't know.
Paul Barnhurst (17:48):
No, sadly enough on the personal front, I'm there with you. I think it was almost 12 plus years ago now. I finally just locked my credit after, I think I had one year, like five different places lost my information. I'd given him and one of 'em, I was with you guys for one weekend. I provided you one email. Somehow you managed it was my information. So I get it. It feels like I said, if you want to live in this world, you just have to accept most of your data's out there and protect it as best you can.
Ned Harding (18:21):
But I don't think companies are ready for their data to be out there.
Paul Barnhurst (18:25):
And I agree with you, it's very different. I worked at American Express and they did things like search. They would scan everybody's hard, drive the network every night, and if they saw anything with that credit card date, you got a note that you had 24 hours to delete it or it would be deleted
Ned Harding (18:43):
Your whole hard drive, I hope at that point.
Paul Barnhurst (18:46):
Yeah, because the last thing they want, and you can imagine if you have a credit card breach, what's that's going to cost? We've seen it not cheap, and it's not the monetary cost, it's the reputational cost that's often higher.
Ned Harding (18:58):
Yeah. So anyway, those are the core features of Enso is hybrid between local and remote and team collaboration regardless of whether the team's in one location or multiple locations.
Paul Barnhurst (19:13):
That makes a lot of sense and I can definitely see where that valuable having the hybrid approach and especially the team side. I am curious, as you've been in data analytics for quite a while now, how did you find that one of the core customers for your products would be FP&A professionals? And so I've had some conversations there. They've mentioned FP&A , I know Alteryx, lots of people in FP&A . So how did you guys end up kind of finding that as kind of a core customer base?
Ned Harding (19:40):
Well, FP&A is kind of at the core of every company. Even if it's a one person company, some portion of your time is spent balancing the books and doing planning and FP&A generates a lot of data. Again, small retail shop or multinational company, you have a lot of data and FP&A has always migrated to Excel because I mean go back in history, iCal and doing spreadsheets by hand and all that kind of stuff. But Excel is so flexible and whatnot, but it's not repeatable. So FP&A quickly outgrows that as they get bigger and need to do monthly reporting and monthly budgeting and whatnot where they want to run the same workflow every month and have it be auditable so that they know they're running the same workflow. Excel is so hard to even audit a single run because a single calculation in Excel can spend hundreds and hundreds of cells and hundreds and hundreds of formulas and sometimes even have circular dependencies. And it's really hard to know that, hey, if I change this one value, what else am I breaking? So tools like Enso, obviously tools like Alteryx tools like that allow you to repeat in a way that you just can't do otherwise. And so those customers have that problem and then find us,
Paul Barnhurst (21:19):
Is that how it happened? The FPA professional more found you, then you guys found it as you're building a product?
Ned Harding (21:24):
Yeah, I would say that's true. You get one sale and you look at why did they buy this product? And you're like, oh, well there's a thousand other people who have that same problem. So it's circular, right? I mean, you help one person that and then that person helps you identify people like them that you want to target, and then you add more functionality to specifically target that group of people. So it's not that they found us or we found them or anything, it's more that it became a conversation where we can both help each other.
Paul Barnhurst (22:04):
No, I know what you're saying. You get someone that comes and you have a conversation, you find you can help 'em and you're like, oh, every company has FP&A . This is a use case we could focus on what are the things we need? There's some scale here as far as the market.
Ned Harding (22:17):
Exactly. And is that the only business we target? No, but it's a really good one and it's a core user of data in almost every company. And so yeah, it's a very core vertical that we're looking for
Paul Barnhurst (22:35):
For sure. And there's definitely others. I mean, you see more and more analytics often rolling up and under FP&A in a lot of small companies nowadays. And so that makes it even tighter of that connection there. I remember the last company I worked at for I started my own business. The first thing the new CFO did was bring in, it was put analytics under FP&A and well finance not directly fp a, but kind of finance as a whole. He brought it in and that was interesting to watch.
Ned Harding (23:02):
People always want to put analytics under something. And the true answer is it should be under everything. Sometimes it's under finance, sometimes it's under marketing, sometimes it's on the production side of things. It can be all over. But the reality is analytics is both smaller and bigger than people think it should be. It should be a small thing integrated everywhere rather than a big thing in one place.
Paul Barnhurst (23:34):
So I'm curious on that. How do you integrate it everywhere? What have you seen work for that? When you say that, I'm just curious your perspective
And now, a brief message from our sponsor, Campfire. Build a best-in-class AI-native ERP designed for modern finance and accounting teams. With Campfire, you can automate revenue, streamline your close, and centralize accounting, all in one unified platform. And now they've hosted Finance Forward, a first-of-its-kind AI summit in San Francisco. They brought together the sharpest minds in finance and operations, all focused on how to actually use AI to move the needle. You don't even have to travel to learn the tactical skills to put your team ahead. You can stream the whole thing on demand, totally free, anytime after the event. So if you're ready to get tactical about AI and finance, head to campfire.ai and register for the on-demand content for free. That's Campfire, where finance teams go to think forward.
Ned Harding (24:43):
I've seen examples of people using analytics in manufacturing for looking at all the data associated with manufacturing and then looking at batch quality. And you can find out, and I'm making this up, although I do have specific examples, I don't want to name any companies, but you might find that manufacturing after 4:00 PM has a higher failure rate, or you might find that you're putting too much fuel on an aeroplane and actually shortening your range. Stuff like that where it's not always obvious what the right solution is until you really look at all the data, you can find issues and you can find ways that you can improve your business from manufacturing, from production, from prediction. I mean in FP&A , making better predictions, but often just in data quality, if your data is good, then what other people are looking at is better. So if you have data, you should be looking at it and it shouldn't be a data analyst looking at it, it should be somebody who knows that data.
(26:05):
So if your data on production and quality shouldn't be looked at by the finance department, it should be looked at by the production or quality people. The data on marketing, again, shouldn't be looked at by the finance people or the production people. It should be looked at by the marketing people. So that need for data analytics is everywhere. I think it's generally a smaller need than people think it is. I think people want to make it into this big thing, and I think it's really just a small thing done often is the best scenario rather than so often companies are like, oh, we haven't done enough with our data, so we're going to do a big push and we're going to do six months of nothing but data and then they forget about it and they don't do anything again. So yeah, you have to integrate it into your processes.
Paul Barnhurst (27:00):
Yeah, it sounds like kind of what I'm hearing is a little bit of really you have to create that culture of data where everybody's looking at the data, they have ownership and using it as part of their job. It's just something that's kind of expected versus trying to do those. Yes, we've all seen them. The big huge push where we're going to focus on data really hard and you find a few insights, you implement them and you forget. You have data and then some time rolls by and you do it again.
Ned Harding (27:24):
And rarely is a data analyst department helpful because the people who are the most helpful are the people closest to the data. So you want to train the people who are close to the data to look at the data rather than to try to centralise it because then they don't understand the context as well.
Paul Barnhurst (27:44):
Definitely is a little more work. I've seen both, but I know what you're saying. You definitely don't have the context the expert does. And so if you're going to do that, you have to work closely with for sure. So I'm curious, and this is everywhere, but in FP&A , we're constantly being asked for data insights, feel like we're always being asked for more and more. I'd love to get your thoughts as far as anything you found that helps deliver insights from data framework you might use processes, how you think about analysing data.
Ned Harding (28:16):
The first thing to do is to think about where you can bring your data together with other data sources. Rarely does your own data paint a whole picture. A good example, I was on the board of Colorado Public Radio NPR station here in Colorado, and they wanted to look at their donor data and you can look at where you have the most donors, but it's not enough to look at that. You have to look at how many people are in each location. So if you can take your donor data and join it up with at least census data to come up with a penetration of in these areas, I'm getting 20% in these areas, I'm getting 1%. And then you start getting to insights that you don't get when you have your data by itself. So the first thing to do when you're analysing your data is to look at what other data sources are available that can augment it.
(29:30):
And often there's a lot of good data out there that's free in the us, the census data, I love the census data. There's so much good data there, but there's all kinds of other data sources, our world and data and whatnot that are free. And then of course you can pay for all kinds of data from Dun and Bradstreet and Experian and all those kinds of providers. But the insights you get when you combine your data with other data are so much more impressive than anything you can do on its own. And oftentimes your data by itself is misleading. I'm trying to think of good examples, but just about any data you have in the us if you're looking at raw numbers, you're finding, oh, the highest is California, the second highest is Texas. Well, great, that told you nothing because that's where the most people are. So at a minimum you need to divide by a population if you do nothing else. But yeah, figure out how you can work collaboratively both inside your company and outside your company. You'll find much, much more insight than you can do by itself. Rarely is magic in a single dataset by itself.
Paul Barnhurst (30:52):
It's funny when you gave the example of population, you need to at least realise, okay, California, Texas have the biggest population, so why is it not surprising that they have the most of X or Y or whatever it might be. A lot of people have been complaining a lot about how high salt lake housing prices are out there. Some highs in the country. I'm like, every house has a basement look at it on a per square foot basis and all of a sudden it goes to the 30th in the country or whatever it is, it drops way down the list from top 10 to middle of the pack.
Ned Harding (31:24):
And yeah, you always have to find context and you always, another point there about looking at your data or anybody's data is you always have to be wary of correlation. Correlation is not causation. And so often people look through data and find some magic correlation and think that they've discovered something really interesting. And oftentimes it just is happenstance. Correlation is not causation. You got to be really careful about that.
Paul Barnhurst (31:58):
And I'm curious, any advice, so if you find correlation, what should you do to try to determine if there's causation there or not? Any advice? Obviously it's easy to find correlation and sometimes there really isn't any there other than, oh hey, the numbers correlated. It was just randomness that caused it.
Ned Harding (32:16):
I mean, this is where you have to learn to think in terms of a scientific method. You have to think about creating a testable hypothesis. You see that correlation, you're wondering if it's causation. Say your correlation is you ran an ad on Super Bowl Sunday and you got more business in February than other months. Maybe not the best example because you can only test that once a year. But ideally one year you would run no ad and the next year you'd double your ad spending and you'd see if that changed the outcome. And if it did change the outcome, you can be confident that that ad spending was causing the outcome. But if reducing or increasing doesn't change the outcome, then you're just looking at correlation. And I picked an awful example with the Super Bowl, like I said, once a year, but a classic other example would be, I spent a lot of money in a Christmas flyer for advertising a retail store and my business before Christmas was the best it was all year. Is that correlation? Almost certainly because people are buying presents for Christmas. So you just have to think in terms of testable hypothesis. If you find a correlation and it's something really interesting, you have to figure out a way to test it and then move forward with that test and spend a little bit of money testing it before you spend a lot of money acting on it.
Paul Barnhurst (33:58):
I really like how you said hypothesis. It's funny how often it goes back to stuff we learned in school. I think it's statistics and hypothesis testing, scientific method and all those things because it's easy to take something and run with it with nothing more than, well, the data showed a pattern, but why did the data show a pattern?
Ned Harding (34:19):
And it might be something real. It might be random noise
Paul Barnhurst (34:25):
Without testing it. You might get lucky, it might be real. You might also lose a bunch of money depending on how much you decide to spend. I appreciate that. I think that's a great advice there. Alright, so now I want to kind of look forward as to where do you think analytics is heading? What do you think it's going to look like in five years? Right? We have all this automation and AI and technology and it feels like things just keep changing pretty rapidly these days. I'd love to get kind of your thoughts.
Ned Harding (34:51):
Well, I think it's really clear that everybody is heading down the AI route, and I think that's really unfortunate because I don't think people are learning how to do the hard work themselves. And we talk about testing a hypothesis, creating hypothesis and testing it and then repeating it. AI is really good at just reinventing everything from scratch every month and there's no guarantee it won't reinvent it in a different way next month. And it just doesn't lend itself to repeatability and it doesn't lend itself to the scientific method at all. And it's really hard to review, but it's clearly where everybody wants to go. The siren song of efficiency is way too strong. And it's hard because AI can make people more efficient. It really can, but it can also lead them astray. When you look at AI these days, people who have some of the best AI in the market are bragging about 90, 95, 90 8% accuracy. Can you imagine if you had a planes autopilot at 98% accuracy? How about 99.9% accuracy? So only one out of a thousand flights you get on is going to crash. That sounds great
Paul Barnhurst (36:16):
Until you realise how many flights that is a day.
Ned Harding (36:18):
Yeah. Oh, it's horrible. It's not like I think AI is all bad, but I do think it is being misused and I think it is being overused. I think using AI to help you start to answer a question is great, but you need to make sure that you answer the same question in the same way every month. Especially in the FP&A world. You don't want to have a new methodology every month. That is the worst thing you can do. So using AI to help you build your methodology for reporting and planning and budgeting, great. Use it once, build a model, understand the model, audit the model, review the model, and then deploy the same model every month for a year, refine it in a year. But please don't just ask the AI to do your reporting every month and let it come up with a new methodology every time. But unfortunately, I think that's the world we're heading towards.
Paul Barnhurst (37:26):
And so obviously when we talk generative ai, it's probabilistic by nature and that's what you're getting at as you can get a different answer every time versus machine learning, which is deterministic, right? If I put the same formula in, I'm getting the same answer. It's math.
Ned Harding (37:43):
Well, yes and no. If you put the same formula in with the same data,
Paul Barnhurst (37:48):
Same variable, same data,
Ned Harding (37:50):
Even machine learning, it's so easy to overfit and people have this tendency to overtrust machine learning and overfit their data. You're much, much, much better off with a simple model that is less accurate for any given month, but more consistent than an overfit model that might be perfectly accurate for 11 months out of 12. And then that 12th month, there's just one or two values in that causes that overfit model to just go wildly out of control. So you do have to be careful even with machine learning, old AI if you will, you have to be really careful of the noise.
Paul Barnhurst (38:36):
And now, a brief message from our sponsor. Is your AI thinking like a generalist or like a finance leader? Here's the thing: most large language models weren't built for finance. They're trained on everything from Reddit threads to recipe blogs. So when you plug them into your accounting workflows, the results can be, let's just say, not balance sheet accurate. That's why Campfire, the company pioneering AI-native ERP, just made a huge move. They've raised $65 million in Series B funding, co-led by Excel and Ribbit, bringing their total to $100 million raised in just 12 weeks. And they're putting that momentum to work with LAM, the large accounting model. It's a proprietary AI trained specifically for finance and accounting, already hitting 95%+ accuracy on accounting tasks. That's redefining what ERP can do for modern finance teams. Unicorn companies like PostHog and Ripple are already using Campfire to automate their finance operations and seeing real results. You can see it for yourself, sign up for a live demo of LAM at campfire.ai. That's campfire.ai.
That's a really good point about the overfit, right? Because get more variables you add, the better fit you're going to get. But that doesn't mean it's really a better predictive model.
Ned Harding (40:08):
It doesn't mean it's a better predictive model. And it takes me back to the few questions ago with correlation is not causation just because your data for the past year, the weather has been correlated highly to sales. Therefore I'm going to put that highly in my model. That doesn't mean there won't be a hurricane this year. You have to be really careful what variables you fit because they're not all predictive and it's not obvious. You can't just look at the data and know what is a predictive variable and what is a dependent variable and what is a randomly correlated variable. It's so hard to know. Weather being such a great example because retail sales are highly correlated to certain kinds of weather. Should you use that in your planning? Probably not unless you're really good at predicting the weather for the next year.
Paul Barnhurst (41:05):
Yeah. Unless you have some kind of algorithms, something that's good at predicting the weather that you can trust and keyword there is trust. So I hear a couple things here and just kind of to recap. I think one, you want to keep it simple, simple is generally better than complex when you're building your models, predictions and things like that, you need to be careful of that. You need to be really careful of how you're thinking about and using AI because you're not always going to get the same answer. And you got some challenges. And then you mentioned hypothesis testing as well, is just really being able to test and make sure you're not relying on just correlation, but you're trying to figure out if it's causation.
Ned Harding (41:46):
Correct. It's so important to understand the scientific method and how to make a testable hypothesis and then to accept it when the data proves it wrong. So often you make your hypothesis, you run a bunch of data, it proves it wrong, and your boss really wanted to build a store in that location. So I'm going to look at other data to make that work. I can't tell you how often I've seen that. I know the answer. I'm going to find some data to justify it.
Paul Barnhurst (42:22):
If we searched long enough, we could find something to justify just about anything. Always. Always.
(42:29):
Yeah. I think we've probably all been guilty of that before where we let our biases creep in and we keep pushing to get an answer we want. And that's a dangerous thing. It's that we all have to be aware of. Love your thought. I know ai, you have some real concerns. Do you think right now we're in a bubble, kind of an over hype cycle and it's going to come back down? Or how do you see the future? You mentioned, hey, it's where we're heading, but what's your thoughts over these next few years with AI just in general?
Ned Harding (42:56):
Well, I mean, is there a bubble? I think that comes down to how you define a bubble, but almost certainly, yes, all of the AI companies are circularly investing in each other and raising their valuations to ungodly amounts. And is there revenue behind it to justify it? No, there's not. So yes, there's definitely some sort of bubble. This is not to say the AI is going away. I mean, it's very similar to the.com bubble back in the late nineties in that everybody's investing in every.com under the sun. Does that mean that computers and the internet went away? No, it just meant that we weren't there on the maturity curve yet. We didn't really know what the right questions and answers were yet. And so it took a number of years after that crash in order to figure out what the internet was good for. But we did.
(43:59):
And it will be the same thing with ai. And the other thing I want to say about ai, this is totally unrelated, but I just feel like it needs said, there's a lot of doom and gloom about AI and I have some of it myself about killing jobs and all this kind of stuff. We have to remember that the technology that destroyed the most jobs ever is the tractor. The tractor destroyed about 90% of the jobs, literally 90% of the jobs AI is not going to come close to that. We survived the tractor killing all those jobs. We came up with new industries. We enjoy the efficiency that we have 2% of the population growing our food versus 92% of the population growing our food. It's awesome. And it'll be the same thing with ai. It'll make people more efficient in the long run and it will enable new things that we can't even imagine now. So there's going to be some short-term doom and gloom, but long-term, it's not doom and gloom. It's just one more of many technologies that have disrupted humanity and will adapt to.
Paul Barnhurst (45:12):
I think that's a great point. I was in college during the.com and I remember that bubble burst and I have similar views to you, right? It's going to transform things. It's here, it's here to stay. But at the same time, it's hard to look at it and say, okay, you have this much revenue and you're going to be a trillion dollar company when you go public or whatever the math is, right? Yeah. You just look at it. Alright. Especially when we see company after company.
Ned Harding (45:37):
Yeah, the math doesn't math, it just doesn't add up.
Paul Barnhurst (45:42):
As someone says, the math isn't mapping as a friend of mine puts. So I'm there with you. I'm very excited. I'm bullish on ai, but I also look at it and there's a bubble in my opinion. Alright, so we're going to move into, this is where I ask a couple fp a questions, ask some similar ones of every guest. So what do you think is the top technical skill that people who work in FP&A need to master?
Ned Harding (46:07):
I'm going to be a little generous in answering what I think of technical skills. I'm going to say number one is the scientific method. I think that is a technical skill. I don't think people learn that enough. It's great to learn your tools, but if you don't learn the scientific method first, I don't think knowing the tools helps you. And then I'm going to have a close second on that and say test driven design. I don't think data people in general really think about the repeatability of their outcomes and being able to test their outcome and their results. So if you're producing a product and FP&A , when you're working with data, you're producing a product. It might be for the cfo, it might be for production, it might be for somebody else, but you're producing a product. If you haven't designed a way to test that product, I promise you it's broken. I promise you there's bad data. So I think FP&A people should study test driven design from the programming world and think about what they're doing on reporting in that context,
Paul Barnhurst (47:22):
Right? So we got a hypothesis testing and test driven design. What about soft skill?
Ned Harding (47:28):
Soft skill? I would say that it's really important to remember that this is a service job, not a technical job. You are providing a service for your company. And that doesn't mean the customer's always right, and you should do only exactly what the customer asked for, whoever the customer is, because the customer's often not, right? And if you look at that quote, the customer is always right. It continues in matters of taste, which everybody seems to conveniently drop off. But you do have to remember that it's a service job. And you do have to remember that if your customers, your CFO, your production department, your marketing department, whoever your customers are, if they're not getting value out of what you're giving them, then you're not doing your job. And so you need to make sure that you're talking to them about the value of what you're delivering and you need to adapt to work with them, not against them.
Paul Barnhurst (48:27):
Thank you. I appreciate that. I really enjoy these answers different than the typical ones we get from FP a professionals. So always nice to have different takes and views. Alright, so I have to ask this one just because I'm curious to see what you'll say. Obviously Microsoft Excel is the biggest analytics tool in the world despite all the other ones we've built. At the end of the day, people use it for just about everything. What's your favourite thing feature about Microsoft Excel?
Ned Harding (48:54):
Well, I mean it's the same as everybody's ultimately, whether they say it or not, the flexibility, the fact that you can do so much in it. I've used Excel for generating source code. You have a table of, I need to do a whole bunch of stuff based on all this table of information. I've written Excel formulas that generated c plus plus code and copy and pasted it into c plus plus. And was it designed for that? No, but it's awesome. But please remember, Excel's not a database. It was never designed as a database. And Excel should never be used when you need to repeat something every month and you have any manual steps because it's really easy. It's so flexible. It's so easy to break. Yeah, its strength and its weakness is the same thing though. It's flexibility.
Paul Barnhurst (49:46):
No, I agree with people. What's its greatest strength? Flexibility. What's its greatest weakness. Flexibility.
Ned Harding (49:51):
Exactly.
Paul Barnhurst (49:52):
It's not going to go anywhere because of its flexibility.
Ned Harding (49:54):
I mean, you never know. I mean, honestly, there was a time when we said VisiCalc was never going to go anywhere, and then it got replaced by Lotus 1, 2, 3. We said Lotus 1 2 3 was never going to go anywhere. It got replaced by Excel.
Paul Barnhurst (50:07):
I guess I should say the spreadsheet form factor is not going anywhere.
Ned Harding (50:10):
Correct? Yes, I will. Whether a different
Paul Barnhurst (50:13):
Application becomes the king one day, we all die in the long run.
Ned Harding (50:17):
Exactly.
Paul Barnhurst (50:19):
So I agree with you. I do think Excel itself eventually will be replaced by something. I don't see the spreadsheet form factor going away anytime soon.
Ned Harding (50:28):
And funny point that your users might not know whether you want to include this, but Excel had to code in bugs that VisiCalc had back in 1978 that then Lotus 1, 2, 3 had VisiCalc counted February, 1900 as a leap year. It wasn't. But if you look at the date coding still in Excel, there's a missing day in the date coding so that it can have the same compatible dates as VisiCalc from, what are we talking 47-year-old software at this point?
Paul Barnhurst (51:05):
Well, if you convert number zero to a date, it's January zero, 1900. That's how it reads.
Ned Harding (51:13):
But I think it's, what is it, 60? You convert to a date and it's invalid.
Paul Barnhurst (51:17):
I hadn't tried that. I didn't know that one
Ned Harding (51:18):
Because that's the one that in VisiCalc was the leap day.
Paul Barnhurst (51:24):
That makes sense now. Yeah, I did know. Yeah, there are definitely some little quirks like that that are kind of fun.
Ned Harding (51:30):
Yeah,
Paul Barnhurst (51:31):
Not fun when they mess you up. But fun in the sense of reading about 'em. It's just the history of our unique
Ned Harding (51:37):
And as software people, people still want to read and write XLS, the binary format that was deprecated by Microsoft, what, 15 years ago or something. And why do people still use that? I don't get it.
Paul Barnhurst (51:54):
If you figure that one out, lemme know.
Ned Harding (51:55):
Okay.
Paul Barnhurst (51:56):
All right. Well we probably should wrap up here. We've got a little over, but I have just a couple personal questions for you. We'll just cover, maybe we'll pick two of these then we'll wrap up. If you could have any one job in the world for a week, what job would you pick and why?
Ned Harding (52:12):
This one caused me a lot of thought because there's jobs the world needs, but realistically, could you fix what's broken in a week? And the answer is no. So would I want to redesign the power grid? Yeah, but you can't do that in a week. So I had to go to what I want to do, which would be fun. Like I said, I'm a product guy. What I want to do is bicycle design. I love designing products, I love bicycles. If I could spend a week designing bicycles, I would just be happy as a clam.
Paul Barnhurst (52:45):
Alright, we're going to ask one more. Which fictional character do you think would make a great data analyst and why?
Ned Harding (52:53):
Oh God, I'm going to just totally cheat on this one. I was a big Isaac Asimov fan as a kid, and Harry Selden, who invented psycho history in the foundation series, absolutely was a data analyst. He figured out how to predict all of future history. So he's a really easy pick from books of my youth.
Paul Barnhurst (53:16):
Love it. Alright, well wrapping up, if someone wants to contact you or learn more about Enso Analytics, what's the best way for them to do that?
Ned Harding (53:25):
You can either go to our website. We have I think weekly demos where you can sign up for a demo or feel free to reach out to myself or Sylvia, the CEO on LinkedIn. We love to chat and we will make ourselves available. I think we're both easy to find on LinkedIn, so yeah.
Paul Barnhurst (53:42):
Alrighty. Well thank you. Thank you so much for joining me today, Ned. I've enjoyed the chat. Always fun to get the perspective of someone coming from the product and the analytics side versus the FP&A side, so thank you for sharing some of your knowledge with us.
Ned Harding (53:55):
My pleasure. Thanks for having me.
Paul Barnhurst (53:57):
Thanks, everyone. That's it for today's episode of FP&A Unlocked. If you enjoy FP&A unlocked, please take a moment to leave a five-star rating and review. It's the best way to support the FP&A guy and help more FP&A professionals discover the show. Remember, you can earn CPE credit for this episode by visiting earmarkcpe.com. Downloading the app and completing the quiz. If you need continuing education credits for the FPAC certification, complete the quiz and reach out to me directly. Thanks for listening. I'm Paul Barnhurst, the FP&A guy, and I'll see you next time.