Skip to content

Levi's (Dette) - Direct/Indirect Bridging - 2025-12-11

Metadata

  • Date: 2025-12-11
  • Company: Levi's
  • External Participants: Odette "Dette" (Treasury, ~25 years experience)
  • Palm Participants: Emma
  • Type: Expert Interview
  • Domain Areas: Cash Forecasting, Bridging, Variance Analysis, Investments & Debt, Scenario Planning
  • Recording: https://tldv.io/app/meetings/693b00726567580013cbd06f/

Summary

Context

Expert interview with Odette ("Dette"), a treasury professional with ~25 years of experience who currently works at Levi's. Dette previously worked at tech companies and has deep expertise in in-house banking, cash forecasting, and FP&A-Treasury reconciliation. This conversation focused on how treasury teams bridge direct and indirect forecasts and manage the relationship with FP&A.

Key Discussion Points

  • Top 10 customer tracking - Focus on influential customers (top 10) rather than tracking every customer. These drive the biggest AR impacts and are candidates for ML-based forecasting
  • In-house banking - Detailed explanation of payment factories, receipt in behalf of, and virtual account structures. Popular in Luxembourg, Belgium, Switzerland for tax advantages
  • Bank statements as ML foundation - "The bank statement is your key for your actuals and probably the best machine learning that you can actually use"
  • FP&A vs Treasury forecasting cadence - FP&A refreshes annually/quarterly, treasury refreshes weekly. Creates divergence that needs tracking
  • Permanent vs timing differences - Track differences between FP&A and treasury forecasts, especially for items like acquisitions that treasury knows first
  • Disruptor adjustments - COVID, tariffs, etc. require adjustment layers. Suggested percentage-based adjustments with explanations for future reference
  • Payment run discipline - Mid-week payments (trigger Monday, pay Wednesday), no Friday payments. Enables predictable cash pooling

Pain Points

  • FP&A plans don't refresh frequently enough to reflect treasury's real-time view
  • Invoices "stuck in desk drawers" create unexpected cash outflows not in any forecast
  • Pre-payments to influencers don't follow standard payment terms
  • Straddling two systems during transition creates double work and delays weekly publish

Feature Requests & Needs

  • Percentage-based adjustment factors for forecasts with explanation fields
  • Calendar/timing for each forecast input (different seasonality per line item)
  • Ability to mark adjustments as "already in ML" vs "needs manual adjustment"
  • "Push button" forecast that spits out weekly view instantly

Jobs & Desired Outcomes

Job: Reconcile treasury forecast with FP&A forecast to explain variances to leadership

Desired Outcomes: - Minimize time spent explaining permanent differences between forecasts - Increase visibility into which differences are timing vs permanent - Reduce manual effort tracking acquisition-related cash impacts

Job: Track influential customers for AR forecasting accuracy

Desired Outcomes: - Minimize effort to identify high-impact customers (top 10) - Improve forecast accuracy by applying customer-specific patterns - Reduce noise from low-impact customer variability

Job: Adjust ML-based forecasts for market disruptors (COVID, tariffs, etc.)

Desired Outcomes: - Minimize manual input by using percentage-based adjustments - Increase transparency of adjustment reasons for future analysis - Reduce duplicate work when disruptors are already reflected in historical ML data

Domain Insights

In-house banking terminology: - Payment factory - centralized team that pays in behalf of affiliates - Receipt in behalf of - collecting on behalf of affiliates - Virtual accounts - layered under in-house bank, can be by affiliate, entity, or customer - Coordination centers - Luxembourg, Belgium, Switzerland, Mauritius popular for tax treatment

FP&A-Treasury relationship: - FP&A maintains "working capital hedge" or cushion - a range they can work within - Treasury provides bottom-up actuals; FP&A provides top-down plan - End of period divergence is common because FP&A hasn't refreshed

Forecasting discipline: - Each input needs its own calendar (payroll, rent, AP by vendor type all different) - Align payment runs mid-week for funding and decision time - "No one does payment runs on Fridays"

Action Items

  • [ ] Schedule follow-up session to show prototypes/designs
  • [ ] Explore percentage-based adjustment UX for forecasts
  • [ ] Research in-house banking as a domain area

Notable Quotes

"The bank statement is your key for your actuals and it's probably the best machine learning that you can actually use." - Dette

"It's when you come closer to the end of a period that you're so far apart because they haven't refreshed. And I have." - Dette

"My treasurer just wants a button... I can go in there and push a button... It spits out everything, apparently for the week." - Dette

"If you guys could actually just take that time of gathering and doing data mining and give back to a customer. Amazing." - Dette

"You can't influence when they're going to pay you, but you can certainly influence when you're going to pay out." - Dette

Full Transcript

Introduction

Me: Hello.
Them: Hi. Hold on.
Me: Hi.
Them: Emma. Hi.
Me: H. I. Nice to meet you. I saw two that day. Or that. Or what?
Them: Good to meet.
Me: Was it?
Them: It's Odette. Yeah, People in the business know me as debt.
Me: Death.
Them: It's just stuck. But everywhere else, I'm Odette.
Me: Okay? Gotcha. Gotcha. What do you prefer in this context? Which one do you prefer for now?
Them: Oh. I don't care. Oh, debt is fine.
Me: Oh, dad. Okay.
Them: I am transitioning out so in a way that there's more people that knows me as Odette rather than.
Me: Oh. Wow. Gotcha.
Them: Yeah. It's weird because people. If I've met people through work, then they know me as debt. But then my Odette world was just family and really close friends.
Me: Yeah.
Them: But then that's evolved because in2018 I became a yoga teacher.
Me: Oh. Nice.
Them: It was a big following in San Francisco, so a lot more people know me as Odette. And so now, when the worlds are merging, It's more like. What did you call her?
Me: Yeah.
Them: And then when. With Covid. Right. And with Covid, I did a lot of stuff on zoom. And so they know me as Odette. I didn't bother to actually change any of that stuff because it's why I am Odette.
Me: Yeah. Fair.
Them: So Gurjeet, probably. Gurjeet knows me as dead.
Me: Very. Probably. I'd say so. He didn't write Odette. When he was talking about you, I was like, Okay?
Them: Yeah. No, I met him when he was super young. I think he was, like, 25, 26. I have a dog here, by the way, so he might bark.
Me: Yeah.
Them: But he will probably map here in a second.
Me: No worries.
Them: What can I do for you today?
Me: Well, first of all, I'm just gonna start a screen recording, if that's okay with you.
Them: Of. Course.
Me: Just internally might share, you know, with our designer and stuff like that. For him.
Them: Yeah. So I'm going to note that we don't have an NDNA for Levi's, so I can't share a lot of that stuff. So I would really think that anything that I share today would be in my past experience without talking about Company.
Me: Yeah, no, that makes perfect sense. Of course.
Them: All right.
Me: Of course. Let's keep it General.
Them: Yeah.
Me: And then if there's any high level examples that comes from Levi's or elsewhere, fine. But yeah, no, for sure. So I was keen to, well, mainly chat a little bit about bridging of the direct and indirect forecasts. But quite keen to explore some other things as well. For sure.
Them: Okay? Follow your lead.
Me: All right. I like to. Maybe. Maybe we should just do a bit of intros before, but after that I like to just kind of ask a few like open ended questions and then we'll see what that takes us.
Them: Okay?
Me: Okay? Yeah. Cool. So. I'm emma. I'm based in Stockholm, Sweden. So at least in Europe, where we did also spend a lot of his time in these place. I've been with Palm for about two years now. I joined us. I guess, founding engineer. So I have a background in business and software engineering. And now. Leading the. Yeah, leading product at Palm for a bit more than a year so far. Did a bit of a pivot there, which is fun.
Them: I like it.
Me: Nice. Yeah. I'm a stepmom. I have two stepdaughters who are 16 and 13 years old now. Been in their lives for a while. The youngest was about five when we started getting to know each other. That's been a while. Also, I used to dance a lot and done a bit of yoga and all of that. I find that super helpful. All sorts of techniques for just relaxing. And, I don't know, just going blank. I like that. That's a bit about me. I would love to hear more about you. And also for how long you've no longer did. How young was he when you first met?
Them: Yeah. No. I've been doing treasury quite a bit. Pretty close now to gosh. 30 years. Almost 25. Let's just say 25. Make it. But it started in a tech industry. Did a lot of portfolio management, investment, basically, and then cash management is my bread and butter. I really found throughout the years that I'm very good with operations. I'm not quite your traditional treasurer, if you will. Assistant treasurer, or whatever you want to call us. But I am very involved with the business, so that has helped me out for my career. So I can speak of the company and then obviously translate that to whatever tools I would have done. We've done a bazillion system administrations and system implementations. The whole gamut. Pretty much touch all areas of treasury for foreign exchange to debt management to cash management. Equity again. Portfolio management from equity bonds. Well, equity was big because then also mergers and acquisitions. Stock are purchasing. So it's pretty much everything that you have touched for Treasury. I have done so that's tech for almost 15 years and then went into apparel. Always been based in the Bay Area, Right. I lived in London for about two years. I did that through a tech company. And back then I was doing foreign exchange. I was leaving there Foreign exchange Group. And then I also was managing 3 billion dollar portfolio. And then after that, I came to Levi's, kind of nestled here for a little bit. And then had an opportunity to live in Brussels for about eight months. Building it again in FX forecasting, if you will.
Me: All right.
Them: So that was probably. It's an offshoot of cash flow forecasting. When I came to the consumer business industry, I found that the forecasting is just not built even at least they have systems. But in consumer business it's not as stable. Right? But I've been a big proponent in all the CFOs that I've worked with is big on forecasting. And so I built that in different companies, right from bottoms up, top to bottom. And we did it in Excel. So here we are. I have a CFO that loves forecasting, and it's really amazing how we impact the business, just knowing where the cash is, right? So that's that. I live here in Sebastopol, California. It's what they call the penal land, so I would compare that to the Rhone Valley.
Me: Nice.
Them: I'm right by our river, which is perfect for the Pinot. I'm 20 miles away from Pacific Ocean.
Me: Oh, wow.
Them: And it's two hours away on two hours. That's more than on a good day. You can get to the San Francisco proper about an hour and 15, but on a bad day, it's two hours. So I rarely go into the office.
Me: All right.
Them: But I pretty much rebuilt my home here. I moved here in 2022. I straddled the city in the wine country for two years, and it's just not working out. I didn't really have a home. I was going there and here and I was wasting time watching TV instead of gardening, you know. So I just decided to uproot and sold the condo in the city and then moved here permanently. So I'm an avid gardener. I'm yoga teacher on the off side. I think that's one of the things. I miss being in the city because San Francisco is a big yoga mecca, right? And it's a different. You know, my biggest class in the city was, like, 89 people, right?
Me: Now. Got. Wow.
Them: And here I've got 25, you know, but that's considered full, right? Because it's. It's just. It's a. It's real country.
Me: Oh, wow. Yeah.
Them: There's. It's a small town, 7,000 people. It's mostly agricultural. It's mostly where on the west side of Sebastopol and a little bit west side, close to the Pacific Ocean. There's a lot of dairy farms, you know, a lot of. More like organic farming, all these sustainability stuff for, for farming. Yeah. And then on the east side is mostly wineries. We have over 20,000 wineries here. I can walk, I can drive, I can bike. It doesn't really matter. And then you cross a man, you know, A drive an hour. Your Napa Valley. So we're the poor cousins of Napa Valley. If you ask me. And so this is where I spend most of my time. I go to the city maybe once or twice. A week. But that's rare. I've been lucky that they let me get away with not going into the office, even if it's a hybrid. So I'm. I'm lucky that way. I work for a guy that manages international groups, so most of his groups are not in San Francisco.
Me: Yeah, exactly.
Them: And then most of my team are spread out. Got one in Oriza, and I've got one in Oregon. So we've been building quite a bit of the cash flow forecasting for this role. Right. Like, you know, in this. In this company, we have a big forecasting project and we do it, and it's, I think, compared to any other company, it's quite robust. Every time I go to a Treasury function, I find myself thinking, good job, right?
Me: Gotcha.
Them: Yeah. So it definitely my goal is to a 13 week and up to a 26 week cash flow forecasting. And then every end of a quarter, I'd like to kind of like say, all right, this is what treasury things that cash is going to end up. And this is what we call free cash flow, right? And then FP and A would come and say, well, this is what we think, and then we reconcile the two. And it's really good information for the CFO office, if you ask me. Then we can make decisions, blah, blah, blah.
Me: Yes, super interesting.
Them: So.
Me: Hey. What. What? What would you say about your current setup? Is it that makes it robust?
Them: You know?
Me: What was it set up look like?
Them: Well, we're straddling two systems right now, right? So we are. We have. We built this model in 2017 out of Excel.
Me: Okay?
Them: And we literally went into the different inputs from the accounts receivable to AP all the outputs, right? We went in there and asked them how the process works, and we kind of designed it. You know, there's a combination of. Combination of using historical averages. Depending on the input, if it works right. Or you can just say, well, this is my plan for the year. I'll upside this because this is what we're seeing. We'll adjust it upward now, depending on what we're seeing in the business. So that's one. So excel. And then now we're building something with power. Bi.
Me: Gotcha.
Them: Really leveraging information in the ERP system.
Me: Gotcha.
Them: AR Outstanding. AP but you know, that doesn't capture everything. Right, because you've got, like. For us, there's shipping, there's all kinds of stuff. Sourcing, demand planning, right? All that stuff comes into play, and they're all in different systems, so we have to be a little bit more creative.
Me: Yeah. No, but it sounds like a good idea. To scatter anything in. In one. Like, in one data lake, one database. Lap power bi on top of it and just. Yeah.
Them: Yeah, it's a hard build because we did it for Excel for a very long time. Like, we publish every week, every Monday, we meet with the inputs every week. Because we just manage. We just map. It's just easier. There's a lot of learning that happens and that we. The conversation. It's not about accuracy anymore. It's about learning. It's information. What would derive from variances and what have you. So.
Me: For sure.
Them: Yeah, it's very informative. And I can't emphasize it enough about having that connectivity of conversation. I don't care if it's five or 10 minutes, we always have something in our calendar for 30 minutes. But we might just speak for five minutes, depending on what's on our plates or what we're looking at. But yeah. That connection is real.
Me: Yeah. No, that makes sense.
Them: Yeah.
Me: We're thinking about, like, sorry, going on the tangent, but just collaborative features and the surfaces for collaboration and sharing information. I believe might be a good, like, call it feature, if you will, of any.
Them: Y. Eah. And here's another good feature that I found is there's these things. System. System fed. That's automated. Amazing, right? And then at the bottom of that, before you finalize a number, you have to have, like, human intelligence, right? I call that a manual input. Whatever. Right? We're also. You are tracking. Okay. System says 1 million, 10 million, whatever. And then. But then, you know, there's a problem here. There's an exception here. There's something that comes out that only a human would know. Right? Because it's nowhere in the system. So that that person, before they say, okay, my input is finalized, they'll put something in there to adjust the number. You never want to tinker around what's in the system, because those are going to adjust, right? But what you know, you at least know, okay, this out of 100 million is 10 million is this problem or this surprise or whatever. And then you track those so that when you look back, you know what you adjusted. For. And you know it's done already. Perhaps going forward, I need to adjust this until this problem results itself. Whatever. So that I think is very, very helpful is that adjustability by a human brain.
Me: 100%. Yeah. Now it's something we're thinking about a lot, given that we do the whole. Bottoms up approach where we. I'm sure we did. Sold you about how we use machine learning to create a base baseline, if you will, for each category and each account.
Them: Right.
Me: And obviously, like you say, there will be a lot of. Cases when a human needs to be able to influence the forecast, be it directly. Like, editing specific value or maybe even say, hey, our plan is we're gonna collect 15% more. For the next six months in this entity.
Them: Yeah.
Me: So.
Them: No, absolutely. Covid is a good example of that. Does is right. It's a disruptor. Any kind of disruptor. Right. Office in your DSO is no longer 30.
Me: Yeah.
Them: Right. All of a sudden, your payment terms, you've extended it to 90. But then you know this is due so you can produce product. Tariffs. I mean.
Me: Yeah. Yeah. Should we go there?
Them: Don't ruin my day.
Me: No. Exactly. Oh, man. Yeah, well, hopefully the madness will end at some point, I'm. I'm sure.
Them: Okay? Y. Eah.
Me: But. Okay, let's. I'm super keen to chat to you about a lot of things, obviously, but if we do, the main focus being the bridging with, like, your plan and. Yeah. What's actually happening and where your cash forecasting you're going versus your. Your plan. You did mention it a little bit, but my first question would be, how are you currently managing this? And like, how does the process of bridging this looks like for you today? If you don't want to do the Levi's example,
Them: Yeah.
Me: Like, fine, but I. If you keep it high level, I don't need super.
Them: Yeah. Yeah. So how do we do this? So this is an fpna, right? There's a forecast. There's a forecast. You're sitting right here, and there's this treasury forecast, right? This forecast for the fpna. Doesn't get refreshed as often with actuals, right? So as you come towards an a point that the treasury forecast is much more accurate because of the actuals. How do we bridge that? Let's just say at the end of a quarter, we are trying to see whether we're going to borrow or not. Let's just say that's the problem, right? Or that's the issue at hand. Or are we going to do a stock repurchase? Whatever. An outflow of some sort. This one. The FPA is not going to have that most recent data, but it has all the other assumptions. Increasing revenue, possibly increase in dividend. These are the things. So it depends on what you're looking for. Right. For most common conversation that we have is are we going to borrow or not? Right. But the Treasury. The flaw of the treasury model does not reflect. The most immediate. Like sales. There's a flash sale here. We're doing this for sales that's not going to be in the forecast. There's a number of for that that was forecasted in the beginning. But as you come closer, there's a push sometimes on sales. Right. And that's not captured there. That's not captured in treasury as well, then that's an overlay. That's one of the things that we talk about. We talk about what's in common. That is already stipulated in the FP and A model. And in treasury model, we talk about things that are not in the FP and A model, but it's in the treasury model. So we identify those differences. And then we come up with, what do we say to the street? Do we say this is how we're going to end up in cash? Let's just say that that is the common denominator that we want to talk about. So then we'll able. Then we're able to give them a range. This is how we're going to end the year. This is how we're going to end the quarter. I like the 13 weeks because. Because we do it like 13 weeks and 26 weeks. The 13 weeks is much more. There's more actuals in it, you know, and more real time. You, you get. Better data. Right? Because the forecast maybe gets refreshed every month, depending on the company. Every month or every quarter. So that's reflected. Where am I getting with this? God, I lost my train of thought.
Me: There was a range, maybe.
Them: Yes, we are able to get a range. So the 13 weeks is closer. Right. But the 26 weeks, when you think about it, in the middle of the year, you're ready to have a window to your year end. So because of that horizon, It's easier to focus on what the storytelling would be. This quarter we can monitor this. But for year end, we can aim for this.
Me: Yeah, that's. That's a really interesting distinction as well. Because.
Them: Yeah.
Me: And this might seem like a very need greatly question, but the range you mentioned, like what is an acceptable range, Right? And how do you work out that number? Like, I think that.
Them: Y. Eah. Y. Eah. And it depends. Like, if you're talking about the 13 weeks. And let's say your DSO is 90 days. I don't know. Your. Your life is set right. Because whatever is in the books already is the revenue. Whatever is the payment is already in the books. There's not much. That you can manage. But what you can do is say, let's push for a collaboration. That's for. For a promotion. Let's do. Let's push for shipping. You know, those are the most things that you can do, you can't do anymore. Maybe in the top line or on the. I mean, depending on what you're trying to do, but for the 26 weeks. Then you can say to the business. You're not meeting this for the 13 weeks. For this quarter, next quarter, this is what you need to make up for. This is what we're holding you to. These are the workforce actions that we're going to have to do to meet year end. So if it helps on both.
Me: Gotcha.
Them: Ways. And how do you do that? Again. It's conversation. It's having that line that makes you flexible.
Me: Yeah.
Them: You as in the entity. Yeah.
Me: For sure. So about dsl, how do you get that number today? Like from the cash poker point of view, is it that you have everything by customer in your erp, for example, and then you track like the payment behaviors or how does it.
Them: Yeah. That's a very interesting thing because I found, you know, there's different ways that we've done this, that we've done. Okay, machine, this is your top 10 customers, right? Let's just say there is such a thing as top 10. You have a region and it has a top 10. One region does it, and then you have another region. That is, there's no such thing as top 10, because each customer is like 5 grand or like 5 million, and all of a sudden you have top 25. Right. How does. That influence. Sorry. How does that influence how you view dso? Right, because. Do you track that? On a customer level. Have a DSO or do you do it in a total ar? This is my AR looks like. Right. We found depending on the company that I've been working on is sometimes going down to the very meant for the to the minutiae of of each customer might be helpful. For ar because then what influences depending again on the product. If you have AR, they have top 10 and that those top 10 customers can sway.
Me: Yeah.
Them: Payment behavior, right? That's one. Another is. Do you have a logistics problem? Does this top 10 say to you? You didn't deliver the products I wanted.
Me: Gotcha. Yeah.
Them: Right, so you've got top line revenue and then you've got dilution. Right. Those are handled differently. In any company. So you might have influences on those. So what? I found that if you have an influential group, I. E. 10. Then you go down to that level.
Me: Yeah.
Them: Not only are you able to. Adjust. These different factors, but then you can actually apply machine learning for the forecast. This is how they pay. This is how they do it. Like. But then your dilution, it is what it is. They're going to cancel this because you, you know, you screw it up on your shipping. Right. Or we have a logistics problem. Man. All these can. All these things are going to get canceled. So all of a sudden you have more control on the input. Does that help?
Me: Yes, very helpful. Because that's something I've just been noticing a bit of a. It seems like everyone's trying to build a feature that does track at each individual customer, like, you know, their payment behaviors.
Them: Okay? Yeah.
Me: It's also quite complex to do it for every single customer. But if it is possible, like you say, if you have in fact, like a cluster, like a top 10 of really influential customers that do in fact have a measurable, like high impact in your dino or whatever, right? Today, then maybe that's maybe where we need to spend our energy as well, to help identify or, I mean, the business would know, right? Like these big customers and then maybe focus on them instead of every single customer. Like, where does good and what does good enough look like in such a feature?
Them: Yeah. Yeah. And we haven't talked about this, Emma, but there's also an influence on the bank structure of the customer, right?
Me: Okay?
Them: You've got a customer that does not have. They are quite decentralized. Again, that's much harder because then your cash management is decentralized.
Me: Yeah.
Them: Right. But if you have a customer that's centralized, Like a company that we know has, that I've worked with, has an in house bank. So they actually have a parent company in an in house bank. Right. All of a sudden you've got concentrated data. On these things. Now, an in house bank is quite sophisticated and not a lot of people have it. There's two companies that I've been with that has it, and I built it.
Me: I'm super happy to have like explain it. Like I'm five on in House banks. I've just seen it like a little bit here and there, but it's not a topic I've been able to dive deep into yet. So if you.
Them: Yeah. So pretty much an in house bank. When you think about it. It is a bank that's in house.
Me: Want to. Yep.
Them: Crazy.
Me: I know.
Them: Name. But. That's what it does, right? It is a particular entity of a company, and that's all they do is financially coordination. So all of a sudden. You can set up your in house bank to pay in behalf of your 10 million affiliates. Right. And every payment and every goes in there. And how do you do that without transferring or disrupting tax things and legal stuff? Right, so this is where a bank structure comes in. You have the in house bank and then underneath that is supported by a virtual account.
Me: Okay?
Them: Structure. And there's different ways to skin a cat on virtual accounts. You do do it by affiliate. Do you do it by entity? And then is there another layer? You do it by customer. Right. So all of a sudden, it really depends on what the company's trying to solve. So if they have a hard time applying cash, a lot of companies would want to do it. And the customer level. Right? And all of a sudden, that's information that you guys could extract, right?
Me: Yeah.
Them: And there's companies out there already doing that, AI learning for you. But do you want to take that? Do you want to just leave that for those companies? But if a company already has a bank structure, you can do that. And across the board. I'm sorry. I don't care what structure you have, but the bank statement. Is your key for your actuals and it's probably the best learn machine learning. That you can actually use.
Me: Yeah.
Them: It's a heavy lift. But if you have a system that could designate your. All your outflows and your inflows, Half of your problem. Or challenges could be answered by that. I jumped in different places. So you want to go back to the in house bank? Okay. In house bank. There's such products in the in house bank level where you have a payment factory. So you've got a team underneath the In House bank, supported by a payment factory. It's centralized. And they pay in behalf of these are real terms payment and behalf of and receipts in behalf of. You collect and you pay in behalf of your different affiliates. And that could be handled in House bank and you can layer that with the in house bank has a set of bank accounts. Each affiliate has a central bank accounts.
Me: Okay?
Them: Then the concentration just flows. I'll stop there.
Me: No, but that's super useful. Thank you. I think. I like. I get. I get.
Them: Yeah.
Me: The. The. The big picture, at least. But then.
Them: It's common. This Luxembourg is a country that has a lot of these coordination because there's some tax things to it. Brussels is another. Switzerland, right? Belgium, Switzerland, all these. I don't even know. Maybe, I don't know. Mauritius, I don't know. Some of those offshore accounts. These are very popular places. For an in house bank because of the tax treatment.
Me: All right. I think I'm gonna need to dive into in house banking as my next next adventure. But, but that being said, like on the topic of the like maybe working capital and the connection to the FP&A in cash forecast and how that all ties together, I know we've already spoken about it, but I like to just ask and be quite specific. Like in your current role, maybe, what are the specific outcomes you're looking for when you're making this exercise? You've mentioned, for example, do we need to borrow money and stuff, but also you mentioned the how Might we improve the business, right? Or help the business see where they need, maybe collect faster or just what are the typical. Outcomes there.
Them: We usually because we have information. Offhand, we because we have the information. We are usually very successful. In achieving the goal because of that information. Right. And then let me add this. Then you're able to put your money's where. Another, not just from a working capitalist standpoint, but also from a tax perspective. Tax might say to you it's best. If we pull our cash. Internationally. That's another right. And so you plan. To know that this is what you're going to do in the US because it's stacked advantages. To do that. And you just do this everything international because there's tax advantage to do that. And when you cross the two, sometimes there is a tax liability. That you can. Defer control, whatever it is. That could lend you in jail. Whatever.
Me: Nice.
Them: Those are the outcomes. It's always easy. When you have the information, then you're able to act. As an example. Free cash flow. You tell the street or you tell your board of directors. We're going to hit free cash flow this quarter. You know what your inflows are? You know where your outflows are. Now, there's no such thing as legally saying you can't pay these vendors. That will show this cash balance. But what you can do is push your sales team. To make up free cash flow number. Does that make sense?
Me: No.
Them: That's all right. So when your vendors, I mean, your invoiced already, right? So those are legally. You're legally liable for it.
Me: Yes.
Them: Most companies would say, let's hold. Not paying.
Me: I got you now. Yeah.
Them: That's not legal.
Me: Yeah.
Them: Right. That's not legal. But what you could do is all these invoices, all these products are in. Collection is slow. What do you do? You target. Your top 10. Your past dues. So that you can release your cash flow.
Me: Gotcha.
Them: Outflow. And still meet free cash flow.
Me: Gotcha.
Them: Yeah. Those are all positive outcomes.
Me: Yeah.
Them: To be honest. An acquisition. Is another.
Me: Okay?
Them: Timing of when you close your acquisition. Could be also very influential if you have this data.
Me: In what way?
Them: It's mostly most acquisitions. You know this. You're working in an acquisition, maybe six or eight months. Ahead of time. Then you can say, Let's just say also the acquisition is in foreign currency. Then you're able to then plan ahead. Because you have a cash flow forecast by currency. Work the region. And then now you also know your outflows from a US dollar standpoint. And then you can then trade. How much you're going to hold so that you can pay for an acquisition in US Dollars or in currency. You know where your money's at.
Me: Yeah.
Them: Does that make sense?
Me: No, that makes a lot of sense.
Them: Okay?
Me: So I'm. I'm assuming, like, for the fpa. They build it based off both, like. I guess things we know as a business will happen. But then I guess the other part is a bunch of assumptions. Or like drivers that they use to model out what they think is the most likely plan, right? Is there ever a way or a need for in your cash forecast to model in some of those assumptions as well?
Them: Yes. Yes. Yes.
Me: Like, is there an exchange in that sense?
Them: We. We use their assumptions quite a bit.
Me: Yeah.
Them: But what happens in the forecast world? What I've seen in FP&A is they do this in the beginning of the year or a beginner of the quarter.
Me: Gotcha. And then it's not. Yeah.
Them: It doesn't get refreshed until. I mean, you have a. You have an annual financial plan, and that usually is what people keep their eye on. Right. But what happens in between there? Sometimes. There's a slowdown, tariffs, whatever, all that stuff that's not in the plan. And so this is where the treasury then comes. Comes up. This is not in your plan, this is in mine.
Me: Yeah.
Them: This would be a permanent difference between the two numbers, and we track those.
Me: So you can.
Them: So knowing those differences between the two numbers helps you manage the business.
Me: Yeah.
Them: For sure.
Me: So in a way, then you adjust your cash forecast using a set of different assumptions that are more like timely or relevant or new, I guess.
Them: I mean, what I know is what I know, right? I can't influence. Because it's what's already. It's the bottoms up.
Me: Yeah.
Them: It's what's real.
Me: Yeah.
Them: For them. They can then say, all right for next quarter. Let's refresh revenue to look like this. So that the it matches reality.
Me: Yeah, that's fair.
Them: Or because they. We can do the forecast here, knowing. But also, you know, we. We take their forecasts and we refresh ours, right? But again we have act holes they don't we. There's probably a portion in, in a year that we're pre aligned because they're actuals and you know, they're, they're forecast in. My forecast has the same data.
Me: Yeah.
Them: It's when you come closer to the end of a period that you're so far apart because they haven't refreshed. And I have.
Me: And then you need a way to identify and also explain, like the, if you will, variances between the two.
Them: Yeah. Like acquisition is a big is a huge example. You don't come into a year and say, you know what? We know we're going to have an acquisition. We're going to say this level. Because this are the parameters of my acquisitions, right? So they're going to put a number in the. The annual plan.
Me: Gotcha.
Them: We're not.
Me: Yeah. Now understand?
Them: Right. But then once that engagement negotiations are on the table, I have a better number of what that what that acquisition looks like. They need to follow my number. So there's that give and take.
Me: Very nice.
Them: I hope.
Me: Yeah. Now it makes perfect sense.
Them: That. Sorry. I mean, I'm not even paying attention. I'm enjoying talking to you so much. But. Oh, we have a few more minutes. Okay. I do have. I do have a 1030. Okay, good.
Me: Oh, like. Yes. Okay, okay. Hard stop.
Them: Yeah.
Me: Yeah. I'm just gonna skim through some of my. What if we just do this instead? Then? So. If you would bear with me, indulge me in just a little bit of a.
Them: No worries. No worries.
Me: Imagine that. If you, if you'd like that. Your forecasts are, in fact, machine learning based. So you could have your AP and AR in there as well, or you could have only APA of a certain places. But let's imagine a big chunk of it is machine learning based.
Them: Yeah.
Me: Just staying on the topic of assumptions a little bit. Would there be value in being able to influence those, like machine learning generated forecasts with the stuff that, you know. Like, hey, we know we're dealing with some tariffs now, so probably, you know, let's not be so optimistic about our sales. Or stuff like that. And how would that Would that also be a larger conversation then across. Your stakeholders in the FP&A team and I guess, other people in the wider finance organization as well. And how would you arrive at which assumptions, if you know, if that makes sense?
Them: Yeah. I mean.
Me: Yeah. Exactly. What would need to. How would a cash forecast need to be adjusted accordingly?
Them: Yeah. This is why I like that line item, right? That we talked about earlier, where you. You can influence, right? So let's just say your inputs are all machine learning. It's all machine learning, as in looking at the past.
Me: Yes.
Them: Right. But you have disruptors like a Covid, tariffs, all these things. How do you adjust that? Great machine learning is this. So you look back. Let's just say your machine learning is looking at 10 years. Obviously, the bigger it is, the harder. But let's just say that Covid those machine learning still has Covid features to it. Now. What's the conversation that we're having? In Covid. How much did revenue decline? 20%. I guess that's what I'm trying to get. You look at the differences caused by the disruptor that is embedded in your forecast for machine learning. And there's really just knowing that number. And maybe there is a metric that you guys build on that machine learning. And say this is caused by. I'm thinking out loud here. Now 20% decliner revenue by Covid. Is it the same what we know? Because it's in the future. 20% decrease of. Because of blah, blah, blah, because of tariffs. Then, literally, if it's the same number, you don't really need to adjust anymore, right?
Me: Yeah, no, that's true also. Yeah.
Them: And that can only happen in conversations.
Me: Yes. Yeah, 100%. And I think.
Them: And that's the value of those conversations that we have with fpna, because it's not necessarily line by line item. Most FP and A groups would have what they call working capital hedge.
Me: Working capital hedge. That's a new term for me to Google as well. Nice.
Them: Question. Right. They'll say they'll have a cushion. They always have a range.
Me: Got.
Them: In every line item that they give to the CFO as a metric, they always have a range, so they can influence those range. All I do is feed that information. At least they know.
Me: Yeah.
Them: Because it's all working capital they usually have. They. They'll identify. Just say, I have a hedge of $50 million. Well, I'm off 50 million. There we go. We're offset, right? It's an easier conversation. But I guarantee you, I don't know if FP&A people actually say I have a hedge or a cushion. I don't know. But we're quite transparent here internally, so we know what that is sometimes.
Me: That's good.
Them: Yeah.
Me: Yeah.
Them: But it is. Like you said, it's outside machine learning.
Me: Yes. 100%.
Them: But if you have that on the line, Like, if you have that line of adjustments. And then you can say, this is what I adjusted for. Maybe in the future years when the machine learning looks back. You kind of know your disruptors and you can just say, okay. And maybe that adjustment line would already be 20%, 10%, 15%. All I have to do is check the box. And then you can have a description so that the storytelling is clear. I just had a Tiffany here. Maybe that's what I want built into the system. I don't really need manual input. Maybe I just need percentages of how I want to adjust the number.
Me: And an explanation. For. For the next poor soul who goes into the system and tries to understand what went on there.
Them: Oh, we did an adjustment. Downward adjustment of 20%. What was that about? And you look at. Oh, maybe it's the same factor. Oh, so you just check. You just check. It's already in there.
Me: Exactly.
Them: Or you said, I don't need this because it's already part of the machine learning. Looking back.
Me: Yeah. And you can see it may be in your forward looking forecast, you can see a dip coming in or some. Yeah.
Them: Yeah.
Me: But I guess also sometimes if, let's say, Covid, that's like you say, it's a bit of a disruptor. And I would hope that you've recovered somewhat since, in terms of sales. You know, you have to share, but it's. I guess that's also interesting.
Them: Yeah.
Me: To separate it from, like, I guess, seasonality or like this is a baseline now or no, you know, because there can be so many reasons why that 20% drop occurred.
Them: Yeah.
Me: The business can be a real crisis. Like, coming from within or like there's something wrong with the product, Right? Or. Yeah.
Them: So. Pre Covid. I'm going to tell you this Pre Covid. My forecast was right on the dot. Pre Covid.
Me: Yep.
Them: That's how good that Excel model that we built was.
Me: That's also interesting.
Them: Yeah. And what we did there is all this is why I think machine learning has a lot to do with it. We looked at historicals, and that's all we did. And then we say, okay, historically. The last three years or maybe last year. It's easier to just do prior year. Last year. This is the sales. And now we're saying we're going to increase the sales by 20%. And all I had to really do is look at that, add 20%. Decide on my timing.
Me: Yeah.
Them: Right. Oh, that's one thing. That's one thing. Each input has a calendar.
Me: Yeah.
Them: So I don't think I've ever mentioned that. Because I was just like, what was that? Because seasonality. When you say to seasonality. It's different for each one. For apparel. Right now is big. We're flushed with cash. Anything with a consumer, right? Holidays, back to school, summer, blah, blah, blah. Beyonce has this and Beyonce has that. Wherever.
Me: You need a way to model Beyonce into your capital.
Them: But. But those are. Again, you can control those. I love it that we have a calendar.
Me: Yeah.
Them: For each input, because for outflows, right, there's a certain thing, like, you know, you have payroll, you know, you have rent.
Me: Yep.
Them: They have their own calendar. But the ap. Accounts payable, let's just say. You manufacture, right? The contract manufacturers is a different schedule as you're indirect your SGA. And so it's really nice. What we've done here is synced up payment runs so that. We group all our payments in the middle of the week. So there is time to fund and there's time to make a decision to postpone for next week. Does that make sense?
Me: Yeah, that's very clever.
Them: Yeah. So. And if you have big, if you have a structure that lends to that, even better, because then your pools of cash would be very, very predictable.
Me: Yep. That's super interesting.
Them: Because you can't influence when it's going to pay you, but you can certainly influence when you're going to pay out.
Me: Yes, of course. Yeah, that must be tricky. Like, how do you pull that off internally? If you're in a company as big as yours and you don't have that structure in place, it must be a massive project.
Them: Yeah.
Me: Just initiate.
Them: Yeah. It's, it's really a lot of discipline. So having an in house bank, you control a lot of that. No one does. Payment runs on Fridays.
Me: No.
Them: No one.
Me: Yeah. Nice.
Them: You. You trigger your payment. Rents on a Monday, it gets paid on a Wednesday.
Me: Gotcha.
Them: That makes a lot of things easier.
Me: Yeah.
Them: And the goal of a build, really, is to have a very dynamic cash flow. Right. My. My treasurer just wants a button.
Me: What does the button. Do.
Them: On and he goes, I can go in there and push a button.
Me: And then what happens?
Them: It spits out everything, apparently for the week.
Me: Yeah.
Them: And it's really nice. The Excel one we were able to publish every Monday night. And right now, because we're straddling two systems, we publish in the middle of the week because there's so much work. Because you're basically doing it in two systems, right? And so we'd like to go back on Monday, publish.
Me: Yeah. I can.
Them: Where we at the end of Monday, we say to the CFO team or the CFO office, and we just say, it's ready.
Me: Yeah. But just like. Like a final topic. I think I found it interesting that you said, like, what you did was, hey, you looked at your historical, your actuals. And then you added your assumptions on top of it. And that was, of course, it's also given your internal structure, and there's, you know, that comes with a lot of benefits, but. But at its core, that's something you trusted.
Them: Yeah. Yeah, you really have to look at your. Your processes and then build up to it. And then kind of like, okay, what is not feeding something? Because the more noise you have in the process use than your the more you miss your forecast, the more machine learning. Doesn't because of the variables. But if machine learning, if you already have your processes, you have your payments, blah, blah, blah, all the outgoing, the machine learning, you can actually lean on it because dingy say here are the outliner. This is across the board. Any industry, there's always somebody who's going to get an invoice and stick it in the drawer of their desk. Right. There's always that one person. How do you do that? And this is why I like that line of like, oh. We're remodeling our building. But. Oh, I forgot. I didn't pass this invoice. You have 90 day terms. But you know what? I just found this invoice on my desk. But it's due tomorrow. That's not in anywhere. Anyone's forecast in fpa, that's in capex.
Me: No.
Them: It's been accrued for. And here I am, sitting in cash. All of a sudden I owe somebody $20 million because it's sitting in someone's desk. Those are the things that are not like, they're so common. Right. Here's another. Here's another example. Influencers. They get paid a front. So when you're looking at future influencers, all these things that. They get paid up front. They don't care if you have a 90 day term.
Me: Gotcha, gotcha, gotcha. You stood Errol. This always exceptions to rules.
Them: Yeah.
Me: Yeah. There's so much complexity. That. But, like, what would it take? Like, as a baseline or as, like, how. How would it be useful for you? Like looking. I know you have pretty good processes in place. So sure, but. But. If you were to suddenly, like, have a system that did, in fact, use machine learning to create all of your forecasts. And it's quite abstract. But what. What would make you trust? Because, I mean, you use it to drive decisions that are quite high stakes.
Them: Yeah. How would I I definitely some reasonableness test there, right? Like, you'll have to do trends. You'll have to say, You'll have to do a year over year just to get the numbers. And you say that doesn't look right. Right? Like you do. You need a person to actually do this, and I don't know. I use AI I use machine learning, but at the end of the day, I look at it, I'm like, You know, it's. That's what it's critical is, is having conversations, knowing the business, applying those to those numbers. It can't be just AI right? It cannot be. They'll spit out the numbers for the stuff that you don't want to be like.
Me: No.
Them: You know, nonsense in Excel. That alone. We spend most of our time. So if you guys could actually just take that time of gathering and doing data mining. And give back to a customer. Amazing. This is what we found. And then you. You know, for. Let's just say, Ari, you have different kinds of ways to do machine learning. You're just not trending. What else you got your dilution? You know what else? What else? I'm blanking out now. Because you know the future. There was one point in time where I was straddling. Cash, credit cards. Payment terms, right? You could use that as a company matures, Right? Like you can say, Yes, we have this as revenue, but then the way we collect is mostly cash. Or maybe mostly credit cards. Or on account. You know? Yeah, if you have those kinds of, like, flexibility. Depending on what your client profile is, it might be very, very helpful. Because it's just a click of a button.
Me: Yeah, no, definitely.
Them: What else? Flexibility is key.
Me: Yeah.
Them: That's it. That's all I have. I hope that was helpful.
Me: It's super helpful. Very, very helpful. I've been given a lot. A lot of good things to think about, for sure.
Them: Yeah. And if you want to. You want to chat again and just concentrate on something, I'm happy to do that.
Me: Yeah. Would love to.
Them: Yeah. I'm happy to do that.
Me: Yeah, me. I'm thinking if. If you're up for it, like, maybe even look at some bits of the product or even just some features we're exploring, like designs or whatever, that would be super helpful.
Them: Y. Eah. Kevin and I. Kevin is my right hand person. He does most of the work. I don't do anything. But we're meeting again. Regurgit. I think they put something on my calendar. So we're seeing the product.
Me: Yeah.
Them: We can't. We can't put you guys in yet. Because I don't think it's worth doing in the US but at least I get to have a feel of your product. I missed it yesterday. I blanked out, but Kevin was able to look at it. But, yeah, the more I see the product, I'm. You know, if you want to give me, I. Mean, I don't know the cadence that you want to talk as you want to talk.
Me: It doesn't have to be a specific cadence. I'm just super, like, grateful for whenever you have.
Them: With. Just give me a hanger.
Me: Yeah. It's, it was more because it's. It's sometimes interesting to talk about something very concrete. We might have a prototype or a design on. Hey, you know, when we get to the point of we're looking at stuff like modeling assumptions into your forecast, right now, we're looking at we haven't gotten to the design yet, but the raging how would that look like in a, in a modern product that makes that work so easy and easily? You can spot all of these differences and also perhaps even if there are underlying drivers to those differences and all of those things.
Them: I'd love to see it. I'm not very technical, but I do like I do do concepts. I mean, that's in. In any of these system implementation that I've done before. And I always have a technical person next to me.
Me: Sure.
Them: And I'll just explain what I'm doing and. You know, I don't even know half of the things that could be possible.
Me: Yeah.
Them: And that's. I love that collaboration, to be honest. So. Yeah, I. I'm happy to. I don't know how much I could. Help, but yeah.
Me: No, but it would be like. It could be even things like we're thinking like, okay, forecast. There's a different dimensions to it. One building trust. Having it been explainable, making sure it's transparent, making sure it's clear. What are the kind of building blocks of this forecast? And the other. Sorry. I know you showed on time just being how to use it as a surface positioning, but. Yeah, I'll reach out. I don't want to. From your next next segment.
Them: Yeah. All right. They're. They're bugging me now. They're paying me. Thank you.
Me: Thank you so much. Bye.
Them: All right. By. E.