Skip to content

Personio - Weekly Sync - 2026-02-18

Metadata

  • Date: 2026-02-18
  • Company: Personio
  • External Participants: Tom Thorn (Treasury)
  • Palm Participants: Jennifer, Giannis, Emma Sjöström
  • Type: Customer Call (Weekly Sync)
  • Domain Areas: Investments & Debt, Scenario Modelling, Cash Forecasting, Categorization, IC Activity
  • Recording: None

Summary

Context

Weekly sync between Palm and Personio's treasury team (Tom Thorn). Covered investment data migration status, intercompany sweep logic, a scenario/what-if prototype demo, categorization cleanup progress, and forecast model improvements including outlier detection.

Key Discussion Points

  • Investments migration: All investment logic now in platform. Term deposits and notice accounts will migrate from bank account view to separate investment dashboards. BlackRock file ingestion pending — Tom needs to contact BlackRock to add Palm to SFTP distribution list.
  • IC sweep logic: Rodel building forecast-based funding logic for the 743 account, similar to cash pooling. Account has 500K buffer for customer refunds (not zero-balance). Tom confirmed the approach aligns with how they fund — round amounts based on forecast of collections and bank transactions minus buffer.
  • Scenario prototype walkthrough: Emma demoed early what-if prototype. Tom validated UI as consistent with rest of system. Discussed percentage-based assumptions per category/entity, layering multiple assumptions, and one-off absolute value items (acquisitions, capital markets activities). Tom emphasized need to name, save, and share scenarios with management before applying.
  • Categorization deep dive: Tom reviewing categorization account by account. Main issues: intercompany and treasury funding transfers getting misclassified as supplier payments due to inconsistent payment references. Tom taking action to standardize labeling with accounting team.
  • Forecast improvements: Moving toward transactional-level forecasting (from category-level). Tom offered to walk through transaction patterns with ML team — session planned for next week.
  • Outlier detection: Tom noticed outlier flags in the system. Discussed how it excludes anomalies from forecast projection. Currently at day-level per category, moving to transaction-level. Tom wants to proactively mark round-value treasury transfers as outliers.

Pain Points

  • Inconsistent payment references from accounting team make categorization difficult — different people use different labeling, no payment templates
  • Scenario analysis currently done entirely outside the system in spreadsheets — painful, immediately out of date once closed
  • No visibility into what assumptions the forecast model is making — hard to know if a trend is already factored in before layering additional adjustments
  • BlackRock SFTP file encryption needs workaround to add Palm to distribution

Feature Requests & Needs

  • Layered assumptions: Ability to stack multiple what-if assumptions (percentages + one-off absolute values) into a single named scenario
  • One-off scenario items: Input absolute values (not just percentages) for events like acquisitions or capital markets activities, with custom labels
  • Scenario save & share: Save scenarios, share with management for approval, then click-to-apply to forecast
  • Forecast explainability: When looking at a forecast value, see high-level explanation of why it landed there (e.g., "supplier payments trending up 10% based on last 3 months")
  • Assumption override: Ability to challenge/correct model assumptions at category level (e.g., "that trend was a one-off, revert to levels from 2 months ago")
  • Best/worst case from historicals: Generate worst case by combining maximum outflows + minimum inflows from historical data over past 6 months
  • Confidence bands as scenarios: Use forecast confidence intervals as built-in best/worst case scenarios
  • Outlier marking: Ability to manually mark transactions (e.g., round-value treasury transfers) as outliers to exclude from forecast learning
  • Account-type grouping for scenarios: Apply scenarios to grouped account types (collections, operations, tax) not just entities

Jobs & Desired Outcomes

Job: Apply percentage-based assumptions to forecasts while preserving underlying patterns (existing: sm-001)

Tom validated the prototype UI and confirmed percentage-based category-level assumptions cover most treasury scenario needs. Key additions needed: layering multiple assumptions and one-off absolute values.

Desired Outcomes (emerging): - Minimize the effort to assemble mixed assumption types (percentages + one-offs) into a single named scenario - Reduce the time between scenario analysis and applying it to the live forecast - Minimize the risk of scenario analysis becoming immediately outdated by keeping it inside the system

Job: Understand and challenge forecast model assumptions at category level (new signal)

Tom described wanting visibility into WHY a forecast value is what it is — what trends are being applied, so he can confirm or override. Currently invisible.

Desired Outcomes: - Minimize the time required to understand what assumptions the forecast model has applied to a category - Increase the ability to override model assumptions when external context contradicts the trend - Reduce the risk of double-counting a trend already captured by the model when adding manual adjustments

Job: Define worst/best case scenarios from historical extremes (new signal)

Tom described combining maximum outflows with minimum inflows from historical data to establish worst-case cash positions — the primary scenario treasury teams focus on.

Desired Outcomes: - Minimize the effort to calculate worst-case cash position from historical min/max by category group - Increase the ability to layer multiple severity levels (worst case, moderate, mild) by toggling category groups on/off

Domain Insights

  • Treasury scenario analysis focuses primarily on worst case, not best case — "you have to think of the worst worst worst case"
  • Personio's 743 account uses 500K buffer (not zero-balance) to cover customer refund risk before daily collections arrive
  • Treasury teams receive high-level strategic assumptions from FP&A/strategy teams and apply them to cash — they don't go "into the weeds" at individual revenue/cost driver level
  • Scenario analysis in spreadsheets is immediately stale once closed — only referenced later as "you said this would happen and it didn't"
  • Tom distinguishes between "nudging the forecast" (correcting model assumptions) and "scenario analysis" (layering hypothetical events on top) — two related but distinct workflows
  • Trapped cash / restricted cash in certain countries (e.g., Argentina) is a scenario planning concern as Personio grows internationally

Action Items

  • [ ] Tom to contact BlackRock team about adding Palm to SFTP distribution list
  • [ ] Tom to push accounting team on standardized payment reference labeling
  • [ ] ML team joining next week's call for detailed forecast model walkthrough with Tom
  • [ ] Palm to continue iterating on scenario prototype based on feedback

Notable Quotes

"We don't want to build you a faster horse, we want to build you a car." — ML engineer (quoted by Emma)

"This is being developed... all of this analysis, anything like this stays outside the system. And it's always like the most painful... you have to take all the data out of the system, you have to generate a whole new spreadsheet." — Tom on current scenario analysis

"As soon as you shut that spreadsheet, it's out of date. It will never be looked at or used again. It will only be used as a reference to say, oh, you said this was going to happen and it didn't happen." — Tom

"Usually worst case scenario from a Treasury side is what you focus on. And if all of those things happen together — like you have to think of the worst, worst, worst case." — Tom

"I can probably go away and say we need to apply this to our balances in six months time with some planned activities there. With this, we would be able to do it, I think." — Tom validating scenario prototype covers his needs

"I've had it in most companies I've been at. They're not necessarily using payment templates, so it's inconsistent." — Tom on categorization labeling issues

Full Transcript

Original Metadata

  • Meeting Title: Palm x Personio Weekly
  • Date: Feb 18
  • Meeting participants: Tom Thorn, Jennifer, Giannis, Emma Sjöström

Transcript

Me: Take notes. How are you doing? I'm all right, thank you. How are you?
Them: I don't have a coffee machine anymore. What happened? I don't know. It just stopped. I don't have coffee today. You have to go.
Me: That came. Up.
Them: Yeah, that's not okay. Indeed. Hey, tom.
Me: Hey. Tom.
Them: Sorry.
Me: We were just hurting with Giannis and his broken coffee machine.
Them: Very sad news. Yeah, that's rough. Get straight to the coffee shop. I will not ask you too many questions, Janis, since you are decaffeinated. Yeah, I think it's best for everyone. I'm joking. Okay? So atop, we have a few bobs to talk about this week. So I will just start. So the first thing I wanted to talk about was that all of the logic for investments is now in the platform. I can build now dashboards and things based on all the information you want to see around investments, the principles, the interest rates, average yield, all those sort of data. However, it does involve a small migration because at the moment all of your sort of term deposits, notice accounts and Emma are set up. As bank accounts. So in the account overview, they're also at various bank accounts with balances and transactions, even though the money market funds like Blackrock, which. We'll give an update in just a minute. That's not factually correct because the investments in the net, your bank account. So I just wanted to make you aware that that is going to transition. Those are not going to be feeders bank accounts anymore. They're going to be sort of your investment. They're going to be viewed in a separate way, and they're going to move into dashboards for now, and then they will move back into the app once that page is designed. But, yeah, that is going to change sort of in the next few days. Okay, great. No, looking forward to that. In terms of the data. Then behind that, I guess at some point you'll need some information from me, right? In terms of rates and maturities and that kind of thing. Exactly. Structure is in there ready to consume that. And similarly, the black graph here, there is a bit of a productivity containing the BlackRock files. But how it will essentially work is the team. I guess the short story is the team can unencrypt the files before and I guess we know that that is such a common thing in Trish. It is a bit more work, but is on your side too to add the palm address as it's something I need to confirm. Yesterday I missed a message from him. Sorry. A couple weeks ago. Please don't be angry. I know you're decaffeinated, but I need to reach out to. The BlackRock team to see how we would manage that, because usually the way that we would kind of add yourselves to the distribution list would be via a new user. But I don't know how it comes to things like this. There should be some sort of workaround generally. I see that. So essentially I just need to ask the question. So I'll do that. I'll do that after. This should be straightforward. Yes. That would be good. And then that will also be a way of starting to use the feature, I guess. So, yeah, I'll let you know as soon as those accounts are gone. And basically the information that's in there is because there is information in there already, even if it's not fully up to date. So there is that. The next point in the list is we have been talking quite a bit about how to recognize that intercompany sweep that we've discussed. It's not forecasted. And the balance on. That 5743 account is just going down when it's actually funded every month. And Rodal is really far in creating this logic, sort of how you'd ideally like it to work. So I had a look in the account and you always fund that as a round balance. Sorry account. So if there's a lot of outflows through that account, Sorry, just which account is this? Is there 575 account or the 743 account? So the 743 account is what I've got in my head. But I'll just. I guess where I was sort of coming to was that I think the logic will work similar initially to a cash reap. Where it will just look at all of the flows in that account and fund the difference. Where, even if that is not a perfect solution, whereby you would always fund, say, to a target balance, rather than funding, say, when you decide to make that transfer, and it is more of a round amount, you know, all the outflows. I'm going to fund 5 million into that account. You know, it will be more like a target balance as such. It does that sort of. Do the job. Yes. Maybe just to double check with like, with speaking the same language on that one. Just to make sure I'm understanding. So if I go into the. Sorry. In another part before this. So share this. So if. When I'm. When I'm looking at this one. I'm seeing the cash suite. Sorry. The cash sweep. Sorry. Cash sweeper. Cash sweep. Up here in the actual slide. That cash loop is always based on what is being transferred from the collections account. So if I then go that that amount in here. For, say, 26 to 1 February, 705 should match exactly what is then sweeping out of the 575 account here. For the same date, so be this one. There should be. Okay, there's a reimbursement here, so I guess net, we're pretty close. Maybe something to double check for me, but the. Oh, maybe it's to do with the customary fund here anyway. The point there is that, like, the ending balance for this account, like always, is, as you say, it's a target balance. So we have whatever we collect during the day, whether it be stripe or bank transactions above 500,000, will always get swept. So it is a target balance there. I guess. Sorry. I know we touched on it before, but I guess the point is that we are creating forecasts based on the historical transactions for the bank account. Bank transaction and for the stripe. And the point with, like, if we. Yeah, if he's already far along this. I'm hoping that that's the. Logic, but is to use those forecasts and then Maybe like the -500k as the basis for the suite.
Me: Yes, we are quite aligned. We're thinking about applying a similar logic as to that of CBA pools, essentially.
Them: Forecast creation. Okay, cool.
Me: So exactly that. But then I wasn't aware of the 500k limit, so that's good limit, good information to have. A buffet. It's not a sterile balance thing that we.
Them: Not, not, not quite zero balance. And the reason behind that is probably a bit too high at this point. But we do, we do give customer refunds. So we want to make so if we depending on we may receive a lot of cash during the day and then when we get so we get to I guess the example is again, the example is that we receive cash throughout the day. We receive a strike payment usually like mid morning, maybe early afternoon. We see bank transactions throughout the day, but we're starting position is always 500K. If we have a really significant customer refund and that goes first thing in the morning, we need to have some funding in that account in order to make that transfer, so that. That's why you have that.
Me: 're buffer. Is that based on your largest historical.
Them: Yeah, I. I think it is, but it's probably worth me re checking this to see, like, what. What that the maximum value that gets. But some. Some could be quite sizable for sure. So I think we're always conservative when we come to this kind of thing.
Me: Thank you.
Them: Sorry, Jenn. Does that match? Kind of what you were thinking, or. I don't know. From apples and apples, yes. And no, I think I was. So if you go back into the other account, I was also talking about the treasury funding transfer, the topping up of the account. Yes. All is clear. So. Yeah, rounded amounts for sure. Yeah, the rounded amounts. Okay. Yeah, that is all super clear. And then. Yeah, I wonder if an intraday overdraft, and then you could invest that 500k, that would be. That would be great. The intraday, though, might be a challenge. Depending on how unicredit I have actually have a question like, we have a workshop with them in a week or so's time, but that that's the issue is whether they're comfortable to take that kind of opendraft risk on us at the moment. Sorry, that's completely off topic. No, it's a fair point, though. No, those are the kind of areas that I'm starting to look at to get a bit more efficient and a bit more. A bit more from the bank side, I guess, to help us out with how we manage our balances. Yeah, how? Smooth it could all be. Okay, that's great. Yeah, the. The next one before I hand over to Emma was talked about the Blackrock file ingestion. And also on the timing improvements. We should already see. That the forecast is improving on this. It's not 100% complete yet, but where we are moving towards, we're talking with the guys yesterday, is into more transactional level forecasting. So rather than at the moment like D category level. So we should see. Yes. The pattern is improving, but they're constantly iterating on that at the minute. I think as soon as I can see the clear pattern, In the forecast. I'll let you know. But it's already getting. Yeah, no, that's good. So I only. I only switch to this account because this is one that I'm funding today. And so I wanted to double check how it looked versus what I was expecting. And yeah, this would be one I. Guess as an example of so, yeah, I won't go through it again. We know the story a little bit. But I get what I would also add to that is if it's ever any help for me to to run through that with whoever's working on it, like what these transactions look like, I'M also happy to do that, if that would be useful at all. I don't know. Maybe they already have what they need, but if not. Yeah, of course, the guys worth potentially, you're going to join the Call today, but we've pushed the next week as one of the day guys is is off today, but they're going to join the Call next week to. Yeah. To give a much more detailed walkthrough of all the work that they're doing, and then you can show them, and they can show you as well, a lot of the data behind the scenes of how things have already improved and what their plans are for sort of the next iteration of the forecast models. Yeah, that. That makes sense. I mean, from my side. Sorry, Emma, just before we maybe jump into to your piece, I know I promised last week to do the deep dive, and I've been doing that. Like yesterday afternoon and today, essentially going account by account on transactional basis, really going through in quite a level of detail to check on the categorization. Categorization piece. That's taking a little bit longer than I was hoping, but I'm essentially trying to correct as much as I can see, you know. Not, probably not 100%, but it gets to a really good level in terms of that recategorization and then I guess probably now next week, unfortunately. But when that forecast refreshes, I'm interested to see what that looks like. I'm sure you know, there's other pieces going on that will improve that forecast. But for some of the stuff we were looking at previously, I have a suspicion that is to do with some rogue categorizations that just need. Need reviewing and correcting. So, yeah, yeah, that's what I said. Which categories are you seeing? Yeah. I think that the bigger ones tend to be around intercompany and around, around treasury funding transfers. Investments now look to be a lot, lot more straightforward. I'm not seeing so many issues there, which is great. But there's, yeah, intercompany and treasury funding transfer sometimes getting mixed up with, like, supplier transfers. And I can see. But the issue is, and this is a takeaway, I've noticed as I go through it, there's a takeaway from me to the accounting team to be far more consistent in the way that they're labeling these transactions, because sometimes I'm struggling to understand what they are. And I have to go through my statements to say, oh, that was a transfer. Just they didn't put any reference in there. The only thing that would. The only thing that I can see that stands out with those types of transfers is a big amount, and it's a round number because that's the. Way we fund it. That's the way we fund our entities. So it's actually a takeaway for them that they need to be. You know, it's almost impossible to. If it's just a blanket value. Like, it's almost impossible. I can't. I need to spend 5, 10 minutes checking. What that is, so it's. It's a bit of an update for us as well. Yeah, that makes a lot of sense, especially because it's just a person. Yeah, yeah, yeah. I've had it in, like, most companies I've been at. They're not necessarily. Well, one of the issues is they're not using payment templates, so it's inconsistent. Right. But that, that would be like, a really big improvement already. But the other one is that, yeah, you might have different people jumping in and out. You also have people over time. So we talk about that. I guess that six months of forecast, historic transfer work, you could have had multiple people making these transfers, and maybe they're consistent for a couple of months, but then it switches over to someone else and they start using different language. So that that's something I need to catch up with them on, probably after they get past that year end activities. But something I have in mind to, to push a bit harder because it makes everyone's lives easier and it's actually just good. Yeah, it's efficient all the way around. Like it helps. Everyone else. Yeah. And that'll help there reconciliation, too. You would think so, right? You would think so. Let's see. Let's see how open they are to that, I guess. I know we're going a little bit off topic, but that, that's for, like, intercompany chances, treasury funding transfers. But I discussed it a few times and I would really love for that to push them to be standard across the board. So if it's a supply payment, they tend to just input the invoice reference. Whereas on a second line, if they could, like, you have the invoice, fine, that's first off. On the second payment. On the second line, you could just have some sort of unique identifier, unique code to say this is an employee expense or this is a supplier payment or XYZ. It's something I tried to implement at Uber when I was there, and, you know, that was pretty difficult. We had. We had transactional volume of, like, in the millions weekly. So that was kind of a difficult one. But I think it's here that we could get into a good practice of implementing something. Like that. So, yeah, off the back of literally yesterday and this afternoon, I have that as a takeaway to engage with them a bit closer on this. Yeah, for sure. That makes a lot of sense. So the cleanup after that one reference I get. That's such a good idea. Okay? I think Loads of echo. Now, Emma, T should jump over to you for this scenario. Emma.
Me: Thank you. There is a bit of a delay in the official prototype, but I have one that I just built super quickly. I would still love to show it to you. And before we dive into it, I'd like to say that we are currently focusing quite heavily on the what if use case, because scenarios is the more we dig into it, it's a very, very complex feature. And I'd like two types of feedback from you. One just on the ui. Like, is this even usable for you? And second. We would like to challenge a little bit. So we're seeing a lot of development with technology. We want to think big, like internally, in terms of what type of tooling we can provide. You'll see that. Our machine learning team are working uneven better forecasts using even better underlying models that are. Might be able to do things such as covariances, like, so if I ship this category around a little bit, what might be the ripple effects across other categories? So there's stuff that we can do. And I think one of our machine learning engineers put it very nicely. He said, we don't want to build you a faster horse, we want to build you a car.
Them: That's really this good one.
Me: Exactly right.
Them: Yep. I might. I might use that one myself.
Me: And we are very curious to learn truly about why you're trying to achieve when you do these what ifs? And scenarios rather than getting too stuck in your current process. And, like, right now, if I can just do this percentage flexes, like, what are you really trying to achieve? And we would love to deliver on that, if that makes sense.
Them: Y. Eah.
Me: But again. Fast. Look. We've used. The forecast overview page without the investments for now. I am aware investments would add on here to check. For example, if you liquidate it, are you within policy? Yada yada. I am aware that's a use case.
Them: Y. Eah. Yep.
Me: But for now, just a quick what if. Let's say you are working on just you're worried about this entity, for example. Something is going on there. You want to see? Then what happens? We're in this assumption view now. It's very basic. We're going to have a date picker as well, so you can select for how long this assumption applies. But you can say that we're a bit restricted in categories here. But let's say your costs are going to increase a lot. Like the date. This is the problem with this prototype, that the data is crazy.
Them: Yep.
Me: But imagine this is not 550%. Imagine it's like a 20% increase or 10% increase in cost or whatever. And you can see a preview here, the blue line indicating where your balance likely will end up. Having this assumption applied.
Them: Yep.
Me: And then an assumption on our side was that you might want to have one scope for the assumption itself. Or the what if, like, world. Work on the naming and then you'll have another scope for like, how you view this. Maybe you want to see the impact that a global scale. Then you can. Okay, that's fine. Or you want to see the impact as specific region or so on. So the data, it's probably going to break, but yeah, it's not working. But you can imagine that you'll be able to preview purely balances at this point. Right. And then.
Them: Yep. Y.
Me: In this table as well. We're suggesting to it's not in here, but to be able to see the before and after end of week balance per entity or account or by currency even.
Them: Eah, yeah. Yeah, I think UI looks good. I would say it's kind of in line with the rest of the system when you're looking at forecasts and so forth. So I think it makes sense to kind of maintain that. So I guess that would be the initial feedback when it comes to assumptions. I think the difficulty is is the complexity that you can get into because, yeah, you might have. Supplier costs are going to increase. But then actually. And you see that. Okay, but maybe in six months time. Our forecast for collections also increases. So is there an offset? So maybe at some point you want to layer in several different assumptions, you know, multiple. And that I'm sure from a data perspective, that's an absolute nightmare. But that would probably be what would be helpful to see for sure.
Me: So let's say in this view, I could still apply assumptions to the same scope. I could keep going here, but then also.
Them: Okay. Oh, sorry. Okay. I see. Perfect. Okay.
Me: And then our plan is, let's say now it says done. But if you want to, like, actually save this assumption and look at, go in here and see like, what your forecast would look like would be supplied, you could save. We would want to have, like, when you save it, we want to pop up that forces you to actually add a reason behind each assumption as well. So, like, your colleagues or yourself would remember why you added them. But yes, for now, that would just be in this same scope. It's a bit. It's actually a bit challenging, but. Ok, you can see you have these assumptions, okay? But then maybe you want to add a new one. And then you want to apply that to something completely different, like just the UK entity, for example. And just.
Them: I guess. Sorry. Just quickly, I guess on the entities, you can select multiple entities at the same time as well. Yeah, okay, perfect. I think that that that bit is important as well. Just you were talking about kind of the use case is that we have some entities that are kind of grouped in a way from like a tax perspective or from an intercompany perspective, for example. You want to see those, those two together. So almost like a dual balance, but also. Yeah, for sure, you will probably always come back to the. The group. But one scenario here. Yeah. You have currency. I think that's an interesting one as well. So you say. Like for me, the parameters work here because, yeah, if I say my payroll expenses for my US and UK entities are going to increase by 15%. And then I put that into currency. That's quite done a good scenario test for my liquidity, like my FX liquidity planning. So even with this it's like the layout, I think it kind of works. It ticks those boxes right. You can probably with entity and then the categorization that works those are selections and then flexing it like that. Is. Is. Is. Yeah. Is helpful. I guess the one other thing is, like, is there a way to. Input. Not necessarily a percentage, but a different type of impactful activity. So that that could be just a one off value. Say we do some capital markets activities and we say, oh, we're going to have a loan. Or are we going to have a, you know, we're going to acquire an entity in six months time. And the value of that, that shop, I mean, that would be a forecast item, but maybe you don't have to include it as a forecast item because it's not finalized yet.
Me: So basically a one off. But I say more in the context of what if or scenarios or something or whatever we are calling this.
Them: Yeah. Yeah. Yeah. Exactly so. And it may not even be like at that point, it may not even be a category. You know, you may not even want to create a category for that. You may just want to say like a one off issue, maybe just label that one item individually to say, 50 million out of the door in six months time. And that's obviously important. When we you mentioned the investment side as well. Right. That's when. So you have your.
Me: So when you. Sorry. Just super niti. Would it be okay?
Them: Yeah, yeah.
Me: Would it be okay to have just a general inflow or a general outflow type of okay?
Them: Yeah. Yeah, yeah, yeah, I think so. I mean, it's your own scenario analysis, so you should know what that that's for. I mean, if you what would be. I guess, most ideal, if you're talking kind of blue skies, being able to label that specific scenario, because you could say maybe you do have an acquisition, but then you also need to layer on top multiple different things before that. So maybe there's 50 million out of the door, but actually, before we get to that, position. We're planning to do some capital market strains. So you have several one off items that aren't forecasted. We're not flexing them by a percentage, but they're just several, their outflows that are going to impact our overall balance. But it's something that is almost like a sequence of events that we, that we wanted to include there. So if there was, you know, item one, 50 million in six months time, but then item two, five months time, we take a capital markets transfer and we receive that 50 million, or we see 40 million, then you have the investments laid on top of that. That's when that would also be helpful.
Me: A little bit like assumptions could be percentages or one offs. It could be assembled a little bit like Legos into different, highly customized, basically scenarios. How interesting is it to then have the second step being okay?
Them: Y. Es. Yeah.
Me: The scenario you mention right now. Hey, I want to operate under this assumption in this scenario in mind now, like, would you like to ingest, like, add it on top of your actual forecast, if that makes sense.
Them: Yeah. I. Yeah, I think, I mean, if that was possible to say, okay, click, that looks good, that would, that would be great. But I also understand from, like, if you, if you have those very, very specific one offs, that may not be the most straightforward. It would be great if you could have it like that, but maybe you would. Want to save it, I guess, but, like, the process for something like that would probably be to save it, share it with the management team. You know that that's why you're doing the scenario analysis in the first place. And then once you have that, if there's a transition, like click to click to apply. Or something like that. That would be helpful. But at that point, you probably then. There's probably another sequence of questions, right? Like, which account are we talking about that the focus is going to get applied to, you know, specific timing of the transaction, this kind of thing that probably needs to consider as well.
Me: Yeah, I wanted to ask about it because right now we're quite high level doing like the very, very first, like naive little. Okay, assume for all accounts that has the tax fables category in this. Let's assume it. It gets more expensive, like so. But is that useful as well to start, like for anything in your mind like there's quite high level or would.
Them: Yeah. Yeah. Yeah. Yeah. No, 100%. I mean, to me, that's. You don't. Obviously, you can get right down into the weed of things, like you can, but.
Me: It.
Them: Yeah, this covers a lot. I mean, based on the. And the important thing I think here is the categorization of transfers. Like if you say bank fees or use, like, you can go, you, you can say, okay, bank VC are going to increase by 10% this year. Or you say due to interest rate issue, like deductions for the central bank is going to reduce by 5% this year. The assumptions are all based from our side. And even with these kind of parameters and what you can set out, that, that, that kind of covers it, to be honest, I don't think it needs to go into so much. I can't. Think of examples where I go so into so much detail. I think that really sits with, like, other teams, like the, like our FP and A team when they're looking at revenue, they're going right into the weeds with their business partners and stakeholders to understand all the various different cost impactors and all the various different revenue generators. They go. They go into immense detail. Whereas the treasury team isn't, isn't necessarily doing that at that level. We're not so involved there.
Me: Be. Could it be the case that you have, like, an alignment meeting on the budget versus cash, and you realize, like, oh, maybe we want to plan for that in the cash book.
Them: Yeah. Yeah.
Me: So you would get your assumption from them. So you'd be aware if there's anything significant you'd like to add, but you don't necessarily.
Them: Yeah. We would add that. But yeah, we would add like a higher level if we get it from like a strategy team, as I say, or FPNA team if they give us some indication of. You should probably consider this. With, with this, I can probably go away and say we need to apply this to our balances in six months time with some, some planned activities there. With this, we would be able to do it, I think.
Me: Okay. I know where I'm probably out of time. Ray.
Them: I can go over if needed.
Me: Yeah.
Them: Yeah, yeah, yeah.
Me: No, because I. What I'm also, like, curious. Then is. Is to understand. Because this is a first. Fairly naive stuff. Like, I'm a fan of keeping it simple too. Like, sometimes simple is the way to go.
Them: Yeah.
Me: But if you could, like, go a little bit, like crazy. Just imagining. Like, why. Why are you doing these scenarios? Like you're managing risk, uncertainty. What would be something that that would be. Making your life even easier here. If you could just.
Them: Yeah, I guess I always have to go to a business model that we're not in, in the moment because that, like, at the moment, we're, we're relatively simple in terms of our corporate structure. But the, the stuff within here, if you think about, like, regions or you think about what one, I guess, one. Scenario I can think of is, say, for example, you start to collect across multiple different regions and you start to have issues with certain countries. Because of Trap cash. Something like that, where you have a scenario to say if we continue to operate at this level. And we can't repatriate. What is that situation going to look like in that entity? In X amount of time. Right. You can do that, I guess through the forecast in a way. But when you would want to layer on a dividend or those kind of things, but pick out really specific, maybe not at an entity level, but actually pick out almost at an account level to see what the bounce is going to look like over a very long period of time. Those are kind of scenarios. You can probably go like one, one level beyond, because some point you're looking at scenario for sure, mostly at like an entity, sorry, mostly at either a boot level or maybe in an empty level. But you can also get to a point where you say scenario analysis within a particular account for an entity. So you have an. An entity that's across a number of different countries. Maybe you actually want to look at Basonio se and cokg. And that entity's bank account in Argentina, where there's a cache is complete. Cash is completely restricted. So from a revenue perspective, from a cost, everything looks great for the accounting team. But from a Treasury perspective, it's a real issue using this kind of analysis. But at that level, at an account level, Could. Yeah, for sure. That could be helpful.
Me: Gotcha. So am I hearing correctly that some way to layer in, like, the nature of cash, like, if it's restricted, invested, is it, like, available and to see. Ok.
Them: Yeah, I would say so. And then nature of cash. Sorry. Nature of the cash and then maybe nature of the accounts. As well. That we're obviously have the operations accounts, obviously investments now separately. But maybe you have operations, you have tax and customer collections accounts. You have use those as the kind of criteria at the top to lay those in. So maybe instead of the entity, you actually just want to look at your cash generation and it's not at the moment, as I say, we're quite simple. We only have one customer collections account, but maybe actually when we expand, grow, hopefully we have several of those. And we just want to see the makeup of, of all of our collections and apply that 10% increase to a number of different accounts.
Me: Ay. Gotcha. Would it be fair to say that for now you could proxy it? With using the entity and the category filter. Like if you just have one account with that category.
Them: Yeah, we could. Yeah. Now, at the moment, we could for sure, but.
Me: Be a bit more precise. Perhaps if you have more. As.
Them: Yeah, if, if we opened up a new collection to Count, say our corporate structure at the moment, as I say, is very straightforward, but if we went to more of a subsidiary structure and we started collecting in accounts in the Netherlands and in Belgium and in Luxembourg and all the various different countries we're actually operating in, they're not necessarily all going to our German operational collections account, but they're within the country. Then you maybe have instead of one collections account, maybe quite quickly have four or five, and you want to see those groups. Good bit of accounts.
Me: And we haven't even spoken of, like, traditional scenarios yet, right? Your best case, your worst case, like.
Them: Yeah, yeah, yeah, yeah. Sorry. Yeah, you're completely right.
Me: Yeah. And I think this is so interesting because on the other hand, you can define a scenario or maybe it's something else as, like, the acquisition of something like chore. But what would be, what would the use case in your mind to. To truly, like, paint a picture of the world's best west and best. Cases or like.
Them: Yeah, well, I mean, that's. Sorry. That's a good question. What would probably be like, helpful in terms of just a really quick access is, say, for example, you said you had. So you're doing some analysis on your bank accounts and buffers that you should be holding in quite quickly. What would be helpful is to say over the last six months, what are the most supply payments to be paid out in a month? Sorry the most.
Me: The most. So, like, the. The biggest drop in, like, outflow.
Them: Yeah. Yeah. So. So what you like, best case? Worst case would be a combination for me. Well, obviously it is. Like, your maximum outflows within a month and your minimum info is within a month. And that's usually based on historical data. Right. So you. At the moment, we're looking at percentages, but if you combine those two like was over the last six months, if both of those things happened, say we had just a massive amount of supplier payments because even it can be because they rolled over from the previous month. They were all kind of issues in our accounting team and things just didn't get paid on time. So they all happen within one month, and then you can bind that with our lowest seasonality of customer inflows. And then what is the, like, lowest amount of cash we can have? Because obviously, within our transactional side, even on the budget side, it's always based on assumptions, right? We're paying everything on time that everyone's paying us on time. But actually, if we combine those two, it's like a worst case month. What is the lowest and what. What is what? What? You know what. What kind of things should we be be aware of tracking to say? Actually, you know, we're not going to do the usual this month, actually. We're 10 million lower because both of those things came to be. And maybe, and maybe then that aligns with like bonus time or something like that. So you have all of these various different things that those, those kind of instances, like bringing all of that together is best case scenario is nice. But usually worst case scenario from a Treasury side is what you focus on. And if all of those things happen together. Like you have to think of the worst. Worst, worst case. What would that look like from a Cash perspective?
Me: 100. How would you feel about having that? So I'm totally with you. Like, I also like that approach. Or I understand it. And it feels like it's quite easy to explain to someone as well. Right. Why that is. Worst case, because you can point to your historicals and it's also quite easy to understand conceptually. That makes sense. Like we're minimizing risk.
Them: Yeah.
Me: How comfortable would you be with something? I know we're way over time. Just final question. We want to explore this. We do also have the capability to, let's say self as confidence bonds. In our forecast here.
Them: Y.
Me: Where one could argue that if this is your balance. You'd have your worst case being following the lower bond, and your best case, the top one, and then your likely case somewhere in between, right?
Them: Ep. That would. That would be a preference to see that? Yeah. I mean, if you can layer those in. Often when, I mean, for my, my own experience, I tend to do like, if I'm looking at doing some scenario analysis, I tend to do like four or five levels, like, something like that. So you best case, worst case, maybe then halfway between that and what, what's expected, Something like that. So you start to see. Like a paint a bit of a picture all the way down. Yeah. I mean, best and worst case is great.
Me: Yes.
Them: For sure.
Me: But what would you need? What type of information is required? Because if it's just like data points on a card, for some reason, this is your worst case. For some reason, this is your best case. What level of explainability or data or like, what do you need to understand or trust Even the different.
Them: I think in, in that sense, it would usually be like, as a reference to those historical transactions. Right. I mean, in terms of explainability, if, if the question is, like, what does our cash look like? How confident are we that we're going to be, we're going to hit that line? And you say, okay, if the worst thing happens. That if the worst case cash scenario from the last six months happens in perpetuity, This is what it looks like. So this is like a lower level. This is the worst we could see if it's below that level. Something drastically has gone wrong in various different operational things. There's caches like cash ins are as low as they can be and supplier payments and payroll is just as high as. It can be in an extended basis. Like you can explain it that way, maybe. Then on that not necessarily. Worst case scenario, you turn off one of those levers, right? You say, okay. Cash ins are really bad. Supplier payments are really bad, but actually payroll is normal. So you can actually turn them off and generate various different levels of scenario underneath the line that you have here. So that's the level of explainability almost that I would rely on is you have these various different, like, grouped categories. I don't think you can go down to individual category levels, but you want to. Go down to grouped categories, maybe, and you play around with those a little bit, so you get worst case, everything's wrong. Like some. Almost. Worst case, only two things are wrong. Just not. Not so bad. Only maybe supply payments are high, something like that. Can I ask a question? Connecting like the scenarios to the forecast because you isolated like a few categories there, save supply. Your payments were going up and that was partly worst case scenario. But also in your forecast, you are trending to that scenario because they are going up. And it's like, how do you marry those things together? If you have these scenarios, what is actually playing out when you do your report? Where? How does this page link to? Because these are all hypothetical, but one of these is going to happen, especially if you run in five or six. Where are Again, good question. I. I guess what would be helpful to know is like where, if, if that, if that assumption is already laid in to supply payments, for example. So if there's a trend of the last three months where weekly or even monthly, whatever, we're paying out more than we have been over the previous six months, so that's hopefully then layered into the forecast. I guess when we're looking at this view or in the forecast view, I don't know where they would be, but would be to give me or whoever some indication of what's being applied is the assumption there. So you're not just taking 1 million from last month, you're. You're taking that 1 million. But also, it was only 900k the previous month and 800k the previous month. So the assumption would be that, okay, we're growing by, like, say, 10% in terms of our supplier payments. Each month. If that's the value that's being laid on top in the forecast, it would be helpful to see that and say, okay, maybe it doesn't grow by 10% next. Next month, we actually layer in an additional 10%. So you got a 20% flex or something like that. Yeah, It's a good point, though, if that's really it kind of intuitively in there.
Me: But if the forecast models manages to capture all of this already, Themselves. I'm just trying to figure because we want to again, give you a car, like, if you want to play around. So let's say. Now I'm getting a bit meta, but let's say the forecast model made an assumption.
Them: Yeah.
Me: Right. It looked at the historical trends it's seeing growing by 10% in factors that in like, ideally, that's what it should do.
Them: Y.
Me: When would you want to see that assumption? And then when would you like to challenge those assumptions? Because we are aware there will be events. There will be, like, tariffs, Covid, whatever. There will be things going on that won't be visible in historicals.
Them: Eah. Y. Ep. Yeah, I think. Okay. Yes, I think there's a good. If you, you know, when you kind of click into the forecast,
Me: Yeah.
Them: You see a forecast item? At the moment, it has, you know, the forecast, what's been created. If there was like some, like, you know, when you, when you're recategorizing as transaction, the AI is telling you why it's categorized as such. If there was the opportunity. When you're looking at a forecast value, say you see your previous month, and then you say, okay, my forecast value for next month looks pretty high. So actual versus forecast, when you click into that forecast item, if there was like some sort of. High level explanation of why it's higher or why it's landed on that value for that month. That would be quite a helpful. That would be quite helpful visibility. To have. I don't know. Yeah. To have it within here when you've got all that multiple multitude of data, I think is maybe difficult. But to see that detail in there, because that then allows me if I see that. Okay. We're flexing our supplier payments. Up this month based on the previous few months, but then I know from outside, actually, that was based on us settling a bunch of really overdue invoices. Let's say that's just built up over a long period of time. There's been a project on the accounting team, so that's why. You've seen that increase. Next month we're going to revert back to normal. I can maybe then see that, that rationale, which would be correct, right? From a learning perspective. I can, I can see that rationale in the forecast side, but then say, actually, you know, not, not, not the forecast fault or the AI fault, but I can actually input, like, a offsetting value to bring that. To bring that down, something like that.
Me: Or if we're even crazier, you could input like, I'm just going crazy right now, because I like that. But how valuable would it be to do something like, yeah, you're right. But, hey, from next month, it's probably going to go back to the levels it was two months ago.
Them: Yeah, if you could just. Yeah, yeah.
Me: Now. Like, we want to think ahead. Like, what. What if you could. Don't quote me, but like, if you could reason with.
Them: I will. I'm going to remember this.
Me: If you could reason with your forecast as you could a human.
Them: Yeah.
Me: How would you like? So that's something I think we could say for a separate session as well. But I think that's really interesting because our goal is to capture as much of these patterns as possible. Finals. Final. Final. Nitty gritty. I would assume that when you're saying you'd like to. See you'd like to see at a cash flow, like by category level, maybe not just like the balance level, then.
Them: Yeah. Yeah. Yeah. I think that that's like, to Jenn's point, I think where I would see this would be in the forecast view. If we're trying to do it in a scenario view, I think that's, that's like. Yeah, you're at a too high a level at that point. I think I see that. That, that has to be part of, like, the process from my side. Right. Is that when things are grouped there and we're. Comfortable, all the categorization, everything's running as smoothly as it can. And then I see for the following month. Oh, supply payments are looking pretty high next month. Why is that? Oh, yeah, makes sense. It's learning off that. Then at that point, it's my. My size. Excuse me? On my side. To take some action and then and correct things.
Me: So a little bit of. Yeah, I think I get what you mean. A way to understand and follow the trends and be aware the assumptions essentially that forecast is making when.
Them: Yeah. Exactly.
Me: It's.
Them: If we're talking going all the way there, like, if there was. What would help me to do that maybe would be if there was, like, a report. Of the back of that. So some sort of forecast to show like, this was your actuals from the previous month. This is now like per category. Maybe this is how we've adjusted like, or maybe not the previous one for maybe the previous settlement. The actuals from the previous period, whatever it is that are generating your forecasts for the next 13 weeks. This is now the like a summary of these are the assumptions that we've taken on those forecast items based on the trends that we've seen. And so maybe you can even input it as at that level. Right. You can see it already to say six months, you've. We've seen an upwards trend and so we're continuing that trend. Actually, you know what? That's not going to happen. I don't think that's going to happen. So maybe you're not doing it in an account level at that point. You're doing it like a report category level, I don't know.
Me: Yeah. No, mate, that would be really cool, and it would give you a chance to easily sort of just influence.
Them: Yeah. I wouldn't be doing it by transaction. By transaction, I'd be doing it more.
Me: Yeah, just give a high level inspection, essentially.
Them: At. Indeed.
Me: I think that's really interesting as well. But then we're not really in scenarios anymore, right? And.
Them: No, no. We moved away from scenarios, but I guess.
Me: We're not nudging the forecast in the right direction.
Them: It's nudging the forecast, but then it does feed into the scenario, and I guess it's going back to what Jenn was mentioning. If. If there's already some sort of scenario built into the forecast based on what's being trending. Like, how do we account for that? Plus what I would want to build in as additional scenarios like one offs or whatever. So I guess the process there would be if it was say this, this magical report or whatever dashboard to say to get that visibility on those categories and what's being applied, like in terms of assumptions. For the forecast would be to look at those, make sure I'm comfortable with all of those, that they look about right. And then I would go into the scenario because that means that my cash, cash balances and the forecast, I'm comfortable with what's being generated there. And then I would start to flex. The categories a bit more.
Me: Yeah, I think that makes so much sense. Super cool.
Them: Cool. Then. But this is like, this is really interesting to see. Like this, this is being developed, like, all of this analysis. Anything like this stays outside the system. And it's always like the most painful. It can be very, very painful activity. You have to take all the data out of the system. You have to generate a whole. New spreadsheet, and you have to work on that really, really specific piece of information. But if we can get to something that does that, and when I say, like, building in balance, like, what's the worst case, best case from the previous six months, all that kind of stuff essentially replicates, I don't know. 90% of the scenario analysis I've done in the past probably. That's a huge tick. That's so good, because I do feel like as soon as you shut that spreadsheet as well, it's out of date. Yeah, yeah. No, it will never be looked at or used again. It will only be used as a reference to say, oh, you said this was going to. Happen and it didn't happen. That's the only time that comes up again. So excited also to play about with this. I think. Yeah, it'd be great. Do you have any more follow ups, Emma?
Me: Not really like our ambition is to get some version of this out that you can play around with. It wouldn't like edit any of your actual data right now is the ambition, but just like some version of it for you to just.
Them: Awesome. Sounds good.
Me: Very. Cool.
Them: So I have. Sorry, I have one question on the transaction side. I've started seeing, like, outlier, and that's popped up a few times. I was just wondering what that was. Oh, that's cool. Yes. So this is part of the work that the guys are doing on the forecast. And it will also help with the timing of the flows too. So the outlier detection is it looks at all of the transactions on each account on each category. Each day. And it will identify if something is an outlier and therefore it won't project it forward. So the example he gave before about the supplier, if you have this big run on suppliers at year end, it then pop that in for the next month because it would class it as an outlier. So it's being refined. We have some really good. I don't know if I'm supposed to. I know I always say this. I do have really nice dashboard in the back end that I can track all of your outlier days per cat. Interesting. Was unsure if we should share one off, but that's great. No, no, I assumed it was something like that. And it might be good to at some point to spend a bit of time on that. Like, what I would see is you mentioned rounded high value transfers, for example, and any time I see that from an account that's probably going to be from a forecast side, that's probably some sort of intercompany. Or treasury transfer, maybe just tick all of those all outliers or something like that, because we don't want that forecast generation to put back them. Yeah. Okay. Interesting. Good to know that. That's kind of what I was thinking. Yeah. It's going and it's only going to get more specific as well, so at the moment, it's at the day level per category. So if there's a big spike on that day. But pretty soon it's going to be a transaction level. So it'll be even more specific on those, like, say, those transfers. Cool. Perfect. Okay, awesome.
Me: And you'll be able to at some point, just mark yourself transactions, fill out layers.
Them: Yeah. Yeah. The other. Great. I was thinking. Because I was maybe thinking slightly different. Slightly different way or just. Just how I assumed it was outlined for like a day. Makes sense. Like even at a transactional. Makes sense level. Makes sense. But going back. Sorry, where we're still going. But we were going back to that point. On when references are consistent. Right. So maybe there would also be a use case for when a payment's gone through and it is a Treasury transfer. But what we say, for example, you have that normal fund, I would say treasury funds transfer always. Right. And that's where what's being picked up in terms of applying a category always. But maybe for whatever reason, someone else processed that transfer. I want to. What I want, I just want to convert that to a Treasury funds transfer. But I don't want to the. The system to use any of the logic from what I'm inputting there, because it shouldn't be like that. And it, and it may be in the future, actually someone maybe makes that same mistake, but it's actually a supply payment or something like that. So you can. And that's where, obviously, I want to get more consistent with the accounting. But some, some of those issues that we see when it's like, it's inconsistencies. And the labeling of transfers. It might also be helpful to to have that outlier piece as well.
Me: Yeah, for sure.
Them: All right. Awesome. Yeah. Thank you for that jamming session. It was really. Yeah. No, that was good. No, I've got. No, I'm happy to do it is. I'm always. Yeah, every week. I feel like we're touching some sort of new development. Yeah, it's really good to see. Yeah, this is AI tom. This is how it is. But yeah, we won't take any more time. Send the actions out. Yes. See you next week.
Me: Thank you so much.
Them: Yeah. Cool. Thank you, everyone. Bye. Bye.