Skip to content

On - Cash Forecasting Discovery - 2024-11-19

Metadata

  • Date: 2024-11-19
  • Company: On
  • External Participants: Amanda Mitt (Treasury Analyst), Lucía Galán Cáceres (Team Lead)
  • Palm Participants: Emma Sjöström
  • Type: Customer Call
  • Domain Areas: Cash Forecasting, Scenario Planning, Variance Analysis, Cash Visibility
  • Recording: https://tldv.io/app/meetings/673c9ab39f9cba001303aa07/

Summary

Context

Deep discovery call where Emma interviewed On's treasury team (Amanda and Lucía) about their cash forecasting needs, processes, pain points, and what would make a forecast trustworthy. This is foundational research for Palm's forecasting product.

Key Discussion Points

  • Purpose of forecasting: Short-term = viability (can company pay bills), long-term = performance insights and investment allocation
  • Trustworthy forecast requires: ERP data as foundation (booked AP/AR), detailed bank account level info, transparency on data sources
  • Process today: Starting balance + AP data from ERP + manual inputs (tax, salary estimates from stakeholders)
  • T+1 vs T+7: Same day is simple equation (should match exactly), longer horizon has more room for unknowns
  • Data reliability varies: Depends on industry maturity, booking practices - On is "not stable enough" to rely purely on historical/booked data
  • Visualization critical: With 150 accounts, need tool to highlight top 10 accounts to focus on each day
  • Scenario needs: Simple what-ifs (delete an inflow, add 20% to expenses) - not complex modeling
  • ML transparency: Must be able to explain model assumptions to stakeholders without being a statistician

Pain Points

  • Last-minute surprise payments - payments not in anyone's radar, discovered day-of
  • Payments made outside system - e.g., tax payment from online banking not in ERP, only see debit next day
  • Direct debits - not transparent what's set up historically, banks don't provide this info easily
  • Stakeholder education - teams making payments without informing Treasury
  • Data mining from ERP - accounting system not designed for treasury, requires combining multiple tables
  • TMS inflexibility - if tool isn't easy for quick scenarios, people go back to Excel
  • 150 accounts blindness - become "fully blind" reviewing so many accounts daily

Feature Requests & Needs

  • User-friendly manual input - stakeholders can input forecasted payments directly (not CSV upload)
  • Alerts for threshold breaches - notify when balance goes negative or below security threshold
  • Remediation suggestions - "you need IC financing" vs "you need pooling" based on entity cash
  • Forecast composition visibility - show % from ERP vs manual vs ML, broken down over time
  • Variance analysis support - explain why forecast was off (missing cash-ins? failed payments? which category?)
  • Entity-level variance insights - "your Americas entities forecast is always off"
  • Simple scenarios - delete a payment, add 20% to expenses, compare to base case
  • Pre-filled daily reports - template emails with key metrics ready to send
  • Reference entity for new accounts - "behave like this other account until you have enough data"

Jobs & Desired Outcomes

Job: Produce a short-term cash forecast I can trust and act on

Desired Outcomes: - Minimize the time spent gathering data from multiple ERP tables - Increase the percentage of forecast based on booked/actual data vs guesses - Reduce the frequency of surprise payments not captured in forecast

Job: Review 150+ bank accounts daily and identify which need attention

Desired Outcomes: - Minimize the time spent manually scanning all accounts for issues - Increase confidence that I'm looking at the right accounts first (negative balance, below threshold) - Reduce the likelihood of missing a critical funding need

Job: Explain forecast deviations to stakeholders and management

Desired Outcomes: - Minimize the time required to investigate variance root causes - Increase the ability to attribute variances to specific categories or entities - Reduce the number of follow-up questions from stakeholders

Job: Make quick operational decisions when plans change

Desired Outcomes: - Minimize the time to model a "what if this payment doesn't happen" scenario - Reduce the need to use Excel for ad-hoc calculations - Increase confidence that scenario impact is accurately calculated

Domain Insights

  • Forecast = starting balance + cash-ins - cash-outs = ending balance - fundamentally simple formula
  • Short-term forecasting benchmarks: T+7 should be ~90% reliable based on booked data
  • Buffer calculation: Treasury teams calculate buffers based on historical variance; more unforeseen events = bigger buffer
  • Conservative approach: "Be pessimistic about cash-ins" - you have to pay on time but won't receive on time
  • Cash pooling is a band-aid: Lucia: "It's like giving candy to a kid every time it cries - you're not solving the root cause"
  • Zero variance = success: "If you don't have to explain anything, it's so nice" - Amanda
  • TMS usability test: If people end up doing it in Excel, it's a bad sign for TMS usability

Action Items

  • [ ] Consider stakeholder input portal for manual forecast entries
  • [ ] Design threshold-based alerts (negative balance, security threshold breach)
  • [ ] Plan variance analysis features with category/entity attribution
  • [ ] Explore simple scenario modeling (delete payment, % adjustment)
  • [ ] Show forecast composition (% ERP vs manual vs ML) over time horizon

Notable Quotes

"Cash forecasting is to make sure the company can keep on paying what it needs to pay. So it's really about viability, the day to day... if you don't have cash, you have to close the company down." - Lucía Galán Cáceres

"For me, the difficult part is getting the right data from the ERP... the accounting system is not designed as a treasury management system." - Lucía Galán Cáceres

"At some point, you become blind, you become fully blind. You see the numbers but you don't know where to draw your attention if you have 150 accounts." - Lucía Galán Cáceres

"If the tool is not flexible or intuitive to make scenarios or adjustments, you end up doing it in Excel. And I think that's the worst that can happen to a TMS." - Lucía Galán Cáceres

"You have to be able to feel like you're contributing and understanding what the model is doing. Otherwise I wouldn't present this to my management." - Lucía Galán Cáceres

"Who would you blame in the end?" - Amanda Mitt (on ML-generated forecasts)

Full Transcript

00:00 Emma Sjöström: time. I'm really excited to just Talk a little bit of cash positioning with you guys. And I have prepared some questions and it's semi-structured. So I'm just gonna go ahead and ask a question and we'll take it from there. Okay. My first one would be like, what? What is the purpose of them? What is the purpose of doing cash really noisy out those I think, okay.

00:25 Emma Sjöström: I think it's forecasting. We'll just start like super broad and Move away. In your, in your opinion. Like why do His recording now. Okay, great. Yeah, again, thanks for taking your you do cash forecasting?

00:37 Amanda Mitt: Mmm. Many purposes. I think depend on the the short term lock long term like

00:46 Emma Sjöström: Fair. Would you like to focus on the

00:46 Amanda Mitt: Is it General?

00:52 Emma Sjöström: more short time, maybe? To stop.

00:58 Amanda Mitt: Mmm, like they want to start? No. Lucía.

01:03 Lucía Galán Cáceres: I happy to hear your thoughts first and then I can go if you want, otherwise, I'm biasing you

01:10 Amanda Mitt: um, I think it depends on the business, like it depends on your cash situation. Like now I think for on it's more like where cash forecasting I think supports with Like my thought now thinking about on, but it depends on the business, how the business is performing. is for example, how to allocate better your Cash and your investment.

01:40 Amanda Mitt: And also to get a perception on. how the business kind of, like, I think, yeah, just analyzing how the business is going and how to I know like this kind of future view and how How do you want the cash to flow and the business to be, you know, because it's kind of showing as well.

02:24 Amanda Mitt: When you see a forecast, you see things are going, I think you should, of course, you see it with the actuals, but when you kind of do this assumptions and you have this data, For the future. You kind of see things are going how are things are going to go in the future so you can act before things go bad or

02:27 Emma Sjöström: Yeah.

02:27 Amanda Mitt: if well, you can like, I don't know, do amazing things with the cash and invest really well and yeah, I think this is the purpose for me.

02:35 Emma Sjöström: Yeah.

02:37 Amanda Mitt: But it depends because we have like short-term that it's really more about kind of cash management and

02:43 Emma Sjöström: Of course, yeah.

02:44 Amanda Mitt: To move how to pay for everything. And then more long term is kind of more performance, I think Insights.

02:54 Emma Sjöström: Yeah.

02:55 Lucía Galán Cáceres: The original question was, Why did

02:56 Emma Sjöström: Makes a lot of sense. Lucía did you want to add anything or

03:04 Lucía Galán Cáceres: cash forecasting, right? What do we need?

03:07 Emma Sjöström: to warm up a little bit, like what's

03:08 Lucía Galán Cáceres: To meet our ensuring that the company is viable. That he can run.

03:21 Emma Sjöström: You want to expand a little bit?

03:22 Lucía Galán Cáceres: Sure. I mean, it's I think there are different types of financial planning, but when I think of cash planning or cash forecasting is to make sure the company can keep on paying what it needs to pay. So it's really about the viability the day today. Mmm. That's why I think it's a very Good.

03:43 Lucía Galán Cáceres: What's the opposite of abstract here? So it's not about profitability and numbers that investors analysts are going to look like, but it's actual Viability. So if you have enough.

03:54 Emma Sjöström: Yeah.

03:55 Lucía Galán Cáceres: cash, you have to close the company down at some point. You know, it goes be your

04:00 Emma Sjöström: So, it Yeah, we just

04:02 Lucía Galán Cáceres: Is beyond investors analysis.

04:04 Emma Sjöström: Yeah. So, would you say it's like almost like if you compare to like pyramid of needs, you know, mass flows, you know human pyramids of needs say like the short term for Slime and like you need to pay your bills and then like you can work your way up to those more.

00:00 :

04:21 Lucía Galán Cáceres: Yeah. It's very, really when you compare it to how complex financial planning can be, it's really cushions cashouts and what you expect the balance to be and

04:31 Emma Sjöström: Yeah.

04:31 Lucía Galán Cáceres: and this is it's fundamentally. If you don't, you can spend so much money in financial planning. But if you don't have a cash in the bank account, Stop operating as a business. That's why I mean it's about viability.

04:44 Emma Sjöström: Yeah, makes sense. Thank you. Make sense. What bother you said? That's super nice. So like yeah if if we focus more on the short term one now actually that they viability like keeping cash in the correct places. What what like what do you think is? Needed to make such a forecast usable or trustworthy.

05:06 Emma Sjöström: You know what I mean? Like, what would you need from it to feel that? Hey, yeah, I trust this forecast, I will plan my cash accordingly, you know,

05:16 Lucía Galán Cáceres: Yeah.

05:17 Emma Sjöström: it's a bit of a broad question by design, but just to keep

05:20 Lucía Galán Cáceres: I think one thing that is super important to me. Is that it contains that it uses the information from the ERP so that it is based on booked items?

05:34 Emma Sjöström: Yep.

05:34 Lucía Galán Cáceres: But not necessarily all including that. So, I think if you already have invoices booked to be paid, reinvoices book to be received from customers, then that should be the starting point of a short-term forecasting. But it shouldn't doesn't need to be limited to that. And I think there is a lot of value in, not limiting.

05:52 Lucía Galán Cáceres: It only to that. But it needs to be there for it to be trustworthy for me to connect it to tangible data. Whatever Amanda.

06:04 Amanda Mitt: Yeah, I agree and I think then you need more detail like for example, magnesium or long term, you can be have something more high level and then when you're doing the short term you really need detailed information. So for example, this payment from the ERP will be devoted from these account so the exact bank account

06:17 Lucía Galán Cáceres: Yeah.

06:19 Amanda Mitt: account number for example,

06:21 Lucía Galán Cáceres: Yeah, I think on the short term, it's Accuracy and reliability is super key for me to trust the forecasting versus in the long term. You can allow for many more abstract elements to come in because it is. Anyways, a bit of a guesstimate But when it comes to the short-term, yeah, it's the details and it's the being able to bridge it with with ERP data for me.

00:00 :

06:49 Emma Sjöström: Yeah, that makes makes a lot of sense. So what if you are to do like a short-term forecast today? How would you do? What steps would you take? Given your current like systems and tools and if you really needed to produce some sort of short-term, where do I position my cash for these accounts kind of thing.

07:15 Emma Sjöström: How would you do it?

07:19 Amanda Mitt: First, I think I would need the Like current cash position. So the balance I would start getting the opening balance.

07:30 Emma Sjöström: Yeah.

07:30 Amanda Mitt: Yeah. Then I would add the open. like, I don't know the accounts payables data for example. So if the payments that I have in the ERP, I would get a report. like, for example, we could like, for one week, for example, what is I don't know, maybe three months maximum of that open data, but if I think really, really short term, it would be even I don't even call it cash, forecast.

07:57 Amanda Mitt: I think it's for the same day or like, for the following day, I don't call it a forecast. I think I would call it more like At least more than a week like to be a forecast. Not not really, I don't know. Like How do you think We said? I didn't.

08:12 Amanda Mitt: I never called forecast, something that it's like really, really short term.

08:15 Lucía Galán Cáceres: Yeah.

08:16 Amanda Mitt: More like a.

08:18 Emma Sjöström: Yeah.

08:19 Amanda Mitt: Yeah. Like at least a week. Or two even.

08:23 Lucía Galán Cáceres: Yeah. I mean, it is it is a forecast. It's just that the Let's say the room for data that is not in the system, increases the longer you're looking the horizon. So, if I'm looking at c++ 1, so how do I expect to close my bank accounts today? I expected to be very easy equation of starting balance, plus stuff that I'm gonna pay.

09:03 Lucía Galán Cáceres: So, my stuff that I'm gonna pay plus stuff that I'm gonna receive ending balance. And any deviations against that. Is a sign that we're not booking things correctly. So that's what I mean. There is no room for anything else or this shouldn't be whereas when you look one week or two weeks then there is a lot of room and it continues increasing until you go to.

09:08 Lucía Galán Cáceres: I don't know, three months. Six months. A lot of room for hate. There is a lot of stuff that you don't know that's gonna happen. There is not yet booked for absolutely normal reasons so that's where maybe you say Amanda. That's not a real forecast but to me the formula is the same.

09:36 Lucía Galán Cáceres: It's starting balance plus cachines, minus cash out and in balance, that's the forecast. And then in the cachians in cashouts you have different elements or you have different percentages of how much I expect to be booked already versus I live to, I don't know, controlling to give me input ai to suggest.

00:00 :

09:40 Emma Sjöström: Yeah. How do you get the idea like? Is it later intuition or experience? I'm just asking broadly like the Just getting an understanding of how to know what percentage is booked already. For example, like, if you look at the head is that is it, is it possible to say, ish how much booked

09:57 Lucía Galán Cáceres: Yeah.

09:58 Emma Sjöström: or

10:00 Lucía Galán Cáceres: I would say, it depends a lot in the industry and there any, if you have a

10:02 Emma Sjöström: Yeah.

10:04 Lucía Galán Cáceres: company that's been doing exactly the same thing for 10 years, and they book everything six months in advance. Then, I think you can rely a lot in the ERP, but if you have a company that keeps changing or vendors, that invoice, really late customers that are unreliable. Then there is a lot of room for interpretation from day one.

00:00 :

10:23 Emma Sjöström: Right. Yeah.

10:24 Lucía Galán Cáceres: Would say, so that's what I mean. This is very companies specific and I think we're rather in the And the end of the spectrum of the data like historical or even book data is not enough for us to know where we're gonna end. We're not mature enough or stable enough in that sense.

10:43 Lucía Galán Cáceres: Not that it's a bad thing. It's just

10:45 Emma Sjöström: Hello, and growing. And I mean that's

10:48 Lucía Galán Cáceres: yeah, but I think it's when you're in innovative, you have an innovative business model, and you have also like different Business models within the business model that keep popping up and new ideas and so on. It's, it's bound to happen like this. so, I would say there's no Rule of thumb that you say 50% for the first week and then, but I think it's a good sign.

11:15 Lucía Galán Cáceres: Maybe there is some sort of benchmark for the first week, at least that you say, hey for the T plus 7. I don't know. You should be able to have like up to 90% reliability or something like this but that

11:26 Emma Sjöström: Right.

11:28 Lucía Galán Cáceres: I would know which percentage to expect.

11:31 Amanda Mitt: Yeah, and I think a good way as well as I just when you see that historical data, maybe you don't focus too much on the amounts but you're like, of course we'll always pay salaries or taxes in the certain month or certain days. So this will kind of help you get at least some sort of like, where to focus on.

00:00 :

11:48 Emma Sjöström: Mmm.

11:49 Amanda Mitt: Maybe it depends because there could also be new stuff, but Yeah, I think like focusing on. What's? Yeah? The larger like more considerable payments? It's already like a lot. I think when you don't really have a lot of historical data or a lot of things changing, So at least, you know where like where to ask because for example we didn't know our text payments, right? This happened we needed to kind of forecast this year.

12:26 Amanda Mitt: How much we were going to Have to buy in CHF to pay for the tax payments, you know like this is a big payment. So you go when you reach out to the team and you're like investigate further on the exact amount that they are predicting. So I think if you know like the bigger sums you can kind of Work a bit.

12:37 Amanda Mitt: On how to talk to and what you ask if you don't have that.

12:41 Emma Sjöström: and those type of information,

12:42 Amanda Mitt: Sorry.

12:43 Emma Sjöström: typically does not necessarily live in an ERP, right? Or like,

12:48 Amanda Mitt: No.

12:48 Emma Sjöström: that's, I have to talk to the right person to get this number and then put it in myself somewhere in the system or in an excel like she or in an email or

13:02 Amanda Mitt: Yeah. Yeah.

13:08 Emma Sjöström: This is super interesting. There's a lot like unpack. so, I mean, It could we just move on to like kind of? It might be a bit fluffy too but also still are there any parts of this like the The process of forecasting that you find more critical than other parts.

13:32 Emma Sjöström: Like sudden. Yeah, the start question makes sense to you.

13:40 Lucía Galán Cáceres: I mean for me, that Like I said, the forecasting is special when it's short term, it really should be. A very simple, you could you should be able to do it in a simple excel. So you have per bank account, starting balance cash in cash out and imbalance and then since you know that the source for the cushions and cashouts is gonna be Coming from ERP.

14:02 Lucía Galán Cáceres: So you need to get the data from ERP, but then you need manual entries, right? So tax will give you manual entries. You also know from controlling that you expect 30% growth on sales. So you type this in this kind of stuff, but for me, the difficult part is getting the right data from the ERP.

14:19 Lucía Galán Cáceres: like,

14:20 Emma Sjöström: What, what is difficult about it? Or like how in what way is it difficult?

14:28 Lucía Galán Cáceres: More from a data mining point of view, I would say. So identifying the right table and then Yeah, making sure that it's comprehensive enough. I mean it's but I think this goes more towards The fact that In accounting. I mean the accounting system is not designed as a treasury management system, so the bookings

14:47 Emma Sjöström: No.

14:48 Lucía Galán Cáceres: there to make it easy for accounting or easier for accounting to prepare the financial statements at the end of the month, but that doesn't mean that it's very Intuitive for us, to access a table and say, Okay, this is exactly the table that I need. This represents all of the expected payments to be made.

15:07 Lucía Galán Cáceres: It's rather a combination of three tables. And then those tables don't include which bank account is going to be used for this payment, so then you have to link it to another table so that's the difficult part, that's a critical piece and it's usually the one you spend the most time when you build a forecast, The manual part.

15:19 Lucía Galán Cáceres: You try to reduce as much as possible, and again, you could replace it by AI to a certain extent. But it's the data mining from the ERP that is difficult for me, please.

15:32 Amanda Mitt: And I think always keeping that. Flexibility as well. If you need to change something last minute and to make a decision, you need that. I think sometimes

15:42 Emma Sjöström: Yeah.

15:45 Amanda Mitt: so yeah, maybe some sort of Manual. Like a yes, scenario thing like a little scenario that you could add

15:57 Lucía Galán Cáceres: Yeah.

15:57 Amanda Mitt: minute to kind of calculate a bit and make a decision. In case you just want to kind? Yeah, for example, right we're paying 8 million to today but then you're like, Oh no actually Like, Oh, we're paying tomorrow but then I don't know. Someone actually Um, approved the payment before or something something happened? I don't know.

16:19 Amanda Mitt: You need to make a chart as decision and even if you have like really nice automated report sometimes, you need to add that little manual change to your Plan A

16:29 Emma Sjöström: Yeah.

16:31 Amanda Mitt: scenario. I think in the short term

16:34 Emma Sjöström: Makes sense.

16:35 Amanda Mitt: and anything in the long term, you like, What I always learned is like, to be very kind of conservative, especially in your cachines. Yeah, to never overestimate anything and try to be very pessimistic in a sense.

16:50 Emma Sjöström: So pessimistic about enclosure.

16:53 Lucía Galán Cáceres: Yeah.

16:54 Emma Sjöström: Okay.

16:55 Lucía Galán Cáceres: You have to pay everything on time, but you won't receive everything on time.

17:00 Emma Sjöström: This is super, super nice. I'm gonna I'm gonna move on to a little bit more like Pain points, Related Questions. So I'm just gonna start one and if you feel like, you feel free to drift away a little bit, if you want to, if something that's important pops up in your head, okay? it's Like what types of inaccuracies or surprises? Do you not want to have in your cash forecast? Something that keeps happening or has happened a lot to you historically that you would like to avoid moving forward.

00:00 :

17:53 Amanda Mitt: Yeah, like that last minute. payment from whatever like I don't know you discover from Monday to another that you have like this huge payment coming.

18:04 Lucía Galán Cáceres: Totally.

18:04 Amanda Mitt: That was not in anyone's radar, like that's the nightmare.

18:10 Emma Sjöström: so like things, you just couldn't predict

18:14 Lucía Galán Cáceres: Couldn't or didn't I think payments that are booked on the last day in the year or even worse for me, our payments that are made outside of the system. And then you only see the debits the next day or you see the failed payment because you didn't have enough cash.

18:36 Lucía Galán Cáceres: So I don't know. Tax or let's say a tax payment made from online banking, That's not in the system. You don't have that data. It should have been delivered as a manual inputs by the team so that you had it into account but no one remember to include you in the loop and then you see the debit That's a huge problem.

19:03 Lucía Galán Cáceres: Speaking of debits, direct debits also an issue sometimes. So Unfortunately, it's not very transparent the direct evidence that you've set up historically in a bank account so you can be surprised by a radio bill that you forgot was there so and it's very difficult to get this information from banks.

19:09 Lucía Galán Cáceres: The direct debits that you've set up.

19:11 Emma Sjöström: All right, all right.

19:12 Lucía Galán Cáceres: So, direct debits are always. Yeah, very hard to to have a new radar. So that's that's not a good surprise when you're working title cash.

19:27 Emma Sjöström: So how do you guys, how does it typically? How do you typically mitigate these risks? Like for example that the the big top tax payments is that a process issue at the company or more

19:39 Lucía Galán Cáceres: Yeah.

19:41 Emma Sjöström: how could like yeah.

19:45 Lucía Galán Cáceres: Do you think Amanda? I have to answers but

19:49 Amanda Mitt: Mmm. right now, I think, because of how our processes are it's based on like really, That we have an experienced team, it's kind of a bit worried like, it's more. Yeah. Like we have this mindset that like, for example, But I I also like the cash flow structure as well.

20:21 Amanda Mitt: I think it's a big Way to mitigate this because then you have an automated flow of money between bank accounts of us not. We don't have this at every entity at the global level. But I mean, I think this helps a lot to give some comfort and to mitigate last-minute payments, but it's not a root cause mitigation it's just

20:34 Lucía Galán Cáceres: Yeah, exactly. Actually feel the

20:35 Amanda Mitt: A way to cool.

20:36 Lucía Galán Cáceres: opposite about this. It's like I don't know giving candid to a kid every time it cries, we're we're not really solving you or just making it

20:44 Amanda Mitt: Yeah.

20:46 Lucía Galán Cáceres: stop for a bit. So if you have a connected cash pulled and the payment will go out because another account will fund it but you're not addressing the root cause which is we should have known about this payment. So For me, it's education. We stakeholders like letting the team know.

21:03 Lucía Galán Cáceres: Hey you cannot make a payment without informing us. It's just kind of happens. It's a process topic of like stakeholder management topic and the one it's something you need to do as a Treasury team is calculate the buffer. So you shouldn't operate the accounts at zero and I think the calculation of the buffer Should be.

21:19 Lucía Galán Cáceres: You know, needs to be more conservative, the more unforeseen events that you see your company. Having then you have a big buffer and

21:26 Emma Sjöström: Mmm.

21:28 Lucía Galán Cáceres: the goal is, of course, to reduce it to say, Hey, I should be able to operate with a small buffer because it should really be small last-minute things. It cannot be a tax payment which is huge. Because that's, you know, lost investment opportunities for the company.

21:46 Emma Sjöström: That makes a lot of sense as well to preventative, feels fairly like Yeah, it was super straightforward in my mind in that sense. but, I'm trying to think like a little bit out. You can stop me also If you think it's like how is there any where you see that a TMS would help somehow in this process of getting this information, right? So like

22:13 Lucía Galán Cáceres: Yeah.

22:15 Emma Sjöström: Was.

22:16 Lucía Galán Cáceres: I think a very use of friendly TMS can be super key because then you can give stakeholders access and say every time you have a payment to see here, this is the template, you put the amount, the account, the date. A little text and then Treasury team CC directly, right? So this is this could be super helpful.

00:00 :

22:35 Emma Sjöström: Are these tax payments typically like? Solar recurring within like the same hilarities. So I'm just trying to like

22:43 Lucía Galán Cáceres: Yeah, it depends I think sometimes it tax for example is something they have to calculate month by month and so on. So it's not so much of a recurrent amount but for example, salaries can be The TALENTINE can tell you already for the next year, expect 1 million in salaries per month, of course, it can go up.

23:08 Lucía Galán Cáceres: It's no down, but it's a good estimate. And I mean in general I think the as a company you should push everyone to book this in the ERP. So not to have different sources of truth. But if you cannot have that, then at least having a user-friendly TMS helps.

23:18 Lucía Galán Cáceres: Because the problem sometimes is that TMS is really not user friendly. So you cannot afford to give access to people because they would not know where to put it. So, then you need them to send an

23:27 Emma Sjöström: Yeah.

23:28 Lucía Galán Cáceres: email and then you need to be input in this yourself. So in that sense that EMS can help with the, with the manual input, and then also with the buffer calculation, I think Previously, what I did is, I came up with my own formula for how to compute the buffer.

23:52 Lucía Galán Cáceres: So let's say, 30%, I calculated deviation of questions with 30%, so, it's a 30% of machines plus minus this and I was a little Formula. I wouldn't even call it an equation. No. But I think that TMS could come up with this instead and say Hey, I suggest that you leave 50 million.

23:58 Lucía Galán Cáceres: Why? Because XY said so

24:01 Emma Sjöström: Right. That's where some machine learning magic. Current coming, you could do combination of like your

24:04 Lucía Galán Cáceres: Yeah.

24:06 Emma Sjöström: historical transactional, data your books, and maybe your trends in how your book looks. So I'm many things

24:11 Lucía Galán Cáceres: Exactly.

24:13 Emma Sjöström: Yeah. This is super cool, but do you feel like if let's say a team us would have? Now, I'm just because this is a bit news to me, which I find interesting. Like you, Do you feel that if a team mess would have a way to collect, stakeholder input, that would go into your Data, right? That produces your forecast.

24:34 Emma Sjöström: So and do you feel like

24:35 Lucía Galán Cáceres: No.

24:36 Emma Sjöström: that would be enough to help like stay at a process to what you needed to be? Like, would that be enough to say?

24:44 Lucía Galán Cáceres: I mean.

24:46 Emma Sjöström: Okay, you are responsible now, their tax team to please input this information like what some

24:51 Lucía Galán Cáceres: Yeah, so I think this from, from a tax team, it depends on the company. How? I think in liquid, liquidity planning. if the company has a culture of collaboration and cash awareness and so on, Then, this can really work that you say, you are responsible for telling me what cash you need.

25:09 Lucía Galán Cáceres: Otherwise, I cannot guarantee then they understand it. Can also be in a company where they say. No, this is the Treasury team. I you have access to my calculations. You can translate this to your TMS. I'm not going to touch your system, so it really depends. But even if you have to do the translation yourself having a user-friendly manually input template helps you with the Treasury team as well.

25:31 Lucía Galán Cáceres: and then you can see that this is manual input, you know, so if regardless of which stakeholder is inputting it, but I think this is what's difficult sometimes that it says Okay If you want to do a manual input you need to create an Excel as csv and then upload and then you forget what the format was like and you know, so that's then you stop using it and then you just include this in the buffer calculation say, yeah, plus 2 million you know, and is that Being patient with this.

00:00 :

25:55 Emma Sjöström: Yeah.

25:58 Amanda Mitt: Yeah. And I remember doing cash forecasting and I was like, what I had this checklist kind of of data that I had to have every month. But this was like for one year. So 12, like maybe one year cash forecast and then it was yeah, a lot of following up because we had this different reports from different teams, so it kind of like you start to follow up as well.

26:20 Amanda Mitt: It's not super efficient so if they could do it, just use the ERP. Of course would be ideal. I mean if I didn't even have to Follow up with them, would be better. Because I know of course sometimes they can't, they have to calculate in an Excel or yeah.

00:00 :

26:36 Emma Sjöström: yeah, no, yeah, it makes sense to

26:37 Amanda Mitt: Yeah. or, for example, if just what I imagine, for example, even the bank had this like you see all of the books payments and you see some sort of visual of The days or the weeks and kind of like the amount that you have booked. It's already super helpful because at least then you have like if something new comes up that you're not expecting and it's a really high amount kind of It will.

27:03 Amanda Mitt: Catch in.

27:03 Emma Sjöström: It was somehow you will be drawn to it kind of you expect or

27:08 Amanda Mitt: Yeah. Yeah. Because you can yeah like just a more visual way because then the year depends sometimes it's not super visual. And Yeah.

27:18 Emma Sjöström: Yeah.

27:18 Amanda Mitt: because, I don't know, something like

27:22 Emma Sjöström: That makes sense like you would need to regardless if that input comes from the ERP from a coworker, like a colleague on the tax team, you'd need a way to be notified or like, Hey this this payment is coming up, right? So it's not just silently

27:35 Amanda Mitt: Yeah.

27:36 Emma Sjöström: waiting and forecast somewhere. And then, you'll still be surprised.

27:40 Amanda Mitt: Yeah, patient. You knew like above 10

27:41 Emma Sjöström: Or like, yeah. That means.

27:48 Amanda Mitt: million payment was included in your AP.

27:51 Emma Sjöström: Right, well, yeah, that's actually like amounts with. Would that be like with you would it be like, in general, like, Hey, we want to be notified for any account. Anything about this. This amount in our functional currency, or would it be like, or would you also want to be like, Hey, for this account this amount for that account, that amount, like how granular when something like that? Be For you guys.

00:00 :

28:14 Lucía Galán Cáceres: Having you could go crazy to be honest, with the granularity like it's helpful to be able to do it per account. But I think per entity could be. A good. A good enough solution, right Amanda. So I think Just just to calibrate the magnitude, right? So maybe 1 million is nothing for onage, but it's a huge deal for a little subsidiary in Italy, obvious.

00:00 :

28:36 Amanda Mitt: Right. Mm-hmm.

28:37 Lucía Galán Cáceres: Obviously, why are you doing a 1 million payment?

28:40 Amanda Mitt: Yeah, true. I agree.

28:41 Lucía Galán Cáceres: When your revenue is X, right? So maybe per entity could be Entity and currency could be a good combination.

28:49 Emma Sjöström: Oh yeah.

28:50 Amanda Mitt: It would be great. Yeah.

28:53 Lucía Galán Cáceres: now, and in general,

28:56 Emma Sjöström: Very cool. Nice. I'm gonna start, I'm gonna keep, I'm gonna repeat myself a little bit now, so please bear with me. But like, if you could improve one part of cash, forecasting, process like in general, what would it be? Like, what would be your number one? Part of the process.

29:15 Emma Sjöström: That you would like to improve. I think I can guess but I still want to just

29:27 Amanda Mitt: for me, that the cleaning and that structuring

29:31 Lucía Galán Cáceres: Yeah, it's a very obvious one for us. If I try to copy on this, what else could be useful in cash forecasting. I think. It's maybe one of us like a secondary thing but to meet super important. Like the visualization Having an intuitive outcome with, you know even stupid stuff like coloring and graphs and question mark here and action but really making it because you when you do cash forecasting special in the short term, you do it for bank account.

00:00 :

30:02 Emma Sjöström: Yep.

30:03 Lucía Galán Cáceres: And it's so difficult on. You have to do this on a daily basis. And at some point, you become blind, you become fully blind. You see the numbers, but you don't know. where to draw your attention, if you have 150 accounts, so, I think this utilization element,

30:13 Emma Sjöström: Oh yeah.

30:16 Lucía Galán Cáceres: helps a lot. If the tool is already telling me these are the top 10 accounts I would start with today.

30:24 Amanda Mitt: True.

30:24 Emma Sjöström: And that would based off of like yeah, large changes in the forecast. So like How would you like I'm based on? What would you choose to like, what would make you want to look at a forecast first over another forecast?

30:39 Lucía Galán Cáceres: Negative balance. And so if you expect the negative balance, that's always the first unit to check. or if in the second would be, if you're going, what I, how I did in the past, if it goes below the what you define, as a threshold, So for us it was very it was part of a policy that we said.

30:55 Lucía Galán Cáceres: If it goes to zero you need to have a meeting immediately.

30:58 Emma Sjöström: Hmm.

30:59 Lucía Galán Cáceres: Goes below what we've defined as a security threshold. Then you need to review this week doesn't need to be today. But you know, maybe in two days, something like this.

31:10 Emma Sjöström: a way to add the account level, say this is the security threshold of this account. You would want to be notified or alerted or at least, have your attention drawn to The forecast is, you know, saying that this week it will go beneath it. So It was something like that.

00:00 :

31:28 Lucía Galán Cáceres: Yeah.

31:28 Emma Sjöström: or even if it's an overdraft, of course, that that would be like a lot of my alarm at right like

31:33 Lucía Galán Cáceres: Exactly. I mean, if you focus an account to go overdraft that should be your first action of the day, Maybe even kind of already just thinking ahead, but you could do the tool could already offer suggestions like Easy. Like, Hey, you, we see that another account in the entity in the same currency has cash.

31:50 Lucía Galán Cáceres: You just want to make an easy transfer

31:52 Emma Sjöström: Yeah.

31:53 Lucía Galán Cáceres: Or you don't have enough cash in the entity. Do you want to initiate intercompany financing? But that would be like a solution.

31:58 Emma Sjöström: Right.

31:59 Lucía Galán Cáceres: So, next step.

32:00 Emma Sjöström: So even, let's say, let's say our system doesn't currently support payments, right? But even if we could suggest even hey, this account, you know. So okay. So quick

32:10 Lucía Galán Cáceres: Yes, yes, like hey, there is it's also silly but if If there is entity, if there is cash in the entity, then it says about pulling. If there is not cash in the entity, then you need in their company financing.

32:23 Emma Sjöström: Right.

32:24 Lucía Galán Cáceres: And that is an easy conclusion to make by looking at the total cash for entity. So just that pop up and saying, Hey, you need in their company, financing here. Hey, you need just being here, that's what it

32:33 Emma Sjöström: Yeah.

32:34 Lucía Galán Cáceres: saves you one step of checking yourself.

32:37 Emma Sjöström: Very smart. Did you want to add something more Amanda or like,

32:47 Amanda Mitt: no, I was just thinking this visual part, but remember I used to include comments, for the minute, like, for example,

32:54 Emma Sjöström: Yeah.

32:55 Amanda Mitt: Treasury like the train, the head of Treasury, I would do the cash forecasting, and I would have to add some more comments in the forecast, to explain what that was. Sometimes if there was like because we have the categories in the report, why on the reporting is a little bit more long term, And then the person would want to kind of click in and see what was that payment but because it was an excel.

33:20 Amanda Mitt: I would just add a comment of like the vendor or I don't know. Something like this. So for example, Yeah, I think this was cool. That's the CS said about the visuals because people will ask so many like follow-up questions and Then you kind of start to build the monster year before, or

33:38 Lucía Galán Cáceres: Yeah.

33:39 Amanda Mitt: Yeah.

33:39 Emma Sjöström: That's fair. Yeah. Would you say it would be helpful to have like some again, something that shareable from, for example, palm like, if you were the forecast that you share with,

33:48 Lucía Galán Cáceres: Yeah.

33:51 Emma Sjöström: like a limited limited access to pollen version, kind of well we could

33:56 Lucía Galán Cáceres: Yeah. Yeah. Or even a P, an email,

33:56 Emma Sjöström: just

33:57 Lucía Galán Cáceres: right? But I think what you can What is super helpful is. If you can predefine the kind of report or email you want to send every day and say Hey I want total cash. I want Deviation versus Forecast For the previous day. I want to increasing cash or decrease like I want this.

34:13 Lucía Galán Cáceres: And then the report is prefilled, right? So when you go in the morning, you do your checks and say, Yeah, reports good to go, No changes. Otherwise, I'm manually at the sentence because I said, Amanda I see that someone is gonna ask a question about this and probably didn't pick up on it.

34:27 Lucía Galán Cáceres: So I add it and then send. But I think an email is often enough maybe some graphs or something, but

34:34 Emma Sjöström: Yeah. Yeah, that makes a lot of sense. it's I would like to just move back a little bit to just like the forecasting bit now and just ask you like if you think so I agree on all those things and that's really nice inside for us. and I, I hope that our reporting tool that we're building will be able to actually provide a lot of that for you guys, but we need the forecasts to be something that you would use right to meet feel so central that even if we were to surface them in our reporting tools.

35:07 Emma Sjöström: The variances, for example, wouldn't be valuable to you unless you believed in the forecast, right? so, I would just based on that kind of like, How would you identify like a great focus like what or yeah. What would make you go? Yeah, I try again. We're just going back to like,

35:26 Lucía Galán Cáceres: You know.

35:27 Emma Sjöström: what? What make you go. Just dance. I trust it. I will present it to my stakeholders. I will just my

35:32 Lucía Galán Cáceres: Yeah.

35:33 Emma Sjöström: decisions off on it. We've covered Including ERP data input, is there.

35:39 Lucía Galán Cáceres: Yeah.

35:40 Emma Sjöström: anything more besides like the ERP date and manual inputs? And you can also think in terms of machine learning if you want but just

35:48 Lucía Galán Cáceres: Exactly. So, for me, it's that the forecast is splitting, where it comes from there. As you say, Hey, 80% of the forecast, is your own ERP data. 10 is the manual tennis machine learning adoption so that you have transparency. And you know, which percentage is a machine learning.

36:04 Emma Sjöström: Yeah.

36:05 Lucía Galán Cáceres: guess. And you can see that percentage increasing over time and that's normal. But I think this is an important piece because, yeah, it gives you like it makes it trustworthy. And then, for me, for example, It's also about being able to double click and see what the machine learning assumptions are.

36:23 Lucía Galán Cáceres: Right? So but from a non-data, scientist point of view so I can see the graph and, you know, that's nice. I can see some statistical measures. That's nice. But also some explanation and saying, Hey, we believe that this is something that you're missing in your AP data. And this is why we added this on top of the forecast, something like this, right? So it's like explanations on the machine learning outcomes.

00:00 :

36:46 Emma Sjöström: So, so making the machine learning

36:47 Lucía Galán Cáceres: that's,

36:49 Emma Sjöström: bits explainable and transparent would make make would that would make you trust the forecast and therefore you would use it.

36:57 Lucía Galán Cáceres: Right. Yeah.

36:57 Emma Sjöström: is that the correct like assumption from my

36:59 Lucía Galán Cáceres: Just I mean if you asked what would it make it reliable for you to share it to your stakeholders and the

37:04 Emma Sjöström: Yeah.

37:05 Lucía Galán Cáceres: things. If I share a forecast that is 50% machine learning, I need to be able to stand behind this machine learning algorithm as a non programmer as a non statistician, right? And be able to say, still makes sense, because what I see is that the machine learning is taking into account.

37:26 Lucía Galán Cáceres: The fact that in the last months we had so many urgent payments that went outside of the system. That's a fair point. I see this happening in this entity a lot. So yes, you know, so if I'm asked they can justify it.

37:33 Emma Sjöström: Yeah. And make it easy for you. Also to distinguish between what was the model's fault? What was this reality?

37:41 Lucía Galán Cáceres: That's true, right? Yeah, I mean a thing also something.

37:43 Emma Sjöström: Yeah.

37:44 Lucía Galán Cáceres: that makes any model reliable is how good the variance analysis is. So not just saying, Oh, I was 20% off, but rather saying I was 20% off and I think this is the reason why I was off additional payments that were not in the system. If you don't have any machine learning, right? You can say it was, was it payments that were not in the system? Was it payments that were in the system that failed? And then you have an operational topic if it's stuff that was not in the system, What is it? Is it missing machines? Missing cashouts? What category does it repeat over time, right? This is there a pattern of you missing tax payments every month.

38:20 Lucía Galán Cáceres: So I think giving you more confidence in the very ends analysis and help you close. That gap makes the model more reliable.

38:29 Emma Sjöström: That makes a lot of sense. Nobody would be very helpful to get that context on the variance analysis especially if you're doing it across 150 accounts, right? Yeah.

38:41 Lucía Galán Cáceres: you know, yeah, I think at some point it's also You can go one step further and say, I do varians analysis on an entity level, right? So, I tell you the most problematic entity. So to say, Hey, your entities in America, their forecast is s***. It's always off. You might want to align with the stakeholders.

39:01 Lucía Galán Cáceres: I don't think they're getting the task, right? So, Just providing that sort of. Analytical support to the Treasury team makes you trust the forecast because it's like tool is helping you improve it as well. And it's giving you suggestions, Have you thought of X You thought of why

39:17 Emma Sjöström: Amanda. You look like you're having very deep thoughts.

39:20 Amanda Mitt: Yeah. No. I'm just wearing my own room and icing and my first experiences. Yeah. Um, no I agree. I think with all of the points but yeah, I Know, I'm just thinking what would make me like? Get annoyed or frustrated with the technology or with the system. On the day today, I think as well is like Yeah.

39:47 Amanda Mitt: Again this flexibility like what if something doesn't happen? And I need to check like I think that's the only thing I think like that would make me want to do. rough calculation or like, you know, like A quick fix or something. Like what would make me want to not? Look at the TMS because I think this happened a lot.

00:00 :

40:01 Emma Sjöström: it's

40:02 Amanda Mitt: And you like to excel you and I think this is like, what a big pain, something that you could just delete the same. Like what if I don't receive this?

40:09 Emma Sjöström: right, like a large inflow for

40:10 Lucía Galán Cáceres: Synonyms.

40:11 Emma Sjöström: example, that didn't when then you

40:12 Amanda Mitt: Yeah.

40:13 Emma Sjöström: want to be able to

40:14 Amanda Mitt: Yeah. So

40:15 Lucía Galán Cáceres: It's a really good point. Like if the tool is not flexible or intuitive free, to make scenarios or adjusted adjustments, you end up doing it in Excel. And I think that's, To me. That's the words that can happen to TMS that you end up doing stuff in Excel. It's a bad sign of the usability of TMS

40:35 Emma Sjöström: Right.

40:36 Lucía Galán Cáceres: That's a good point. Amanda like if it's not easy to make it an adjustment or a scenario,

40:42 Amanda Mitt: yeah, you feel like

40:43 Lucía Galán Cáceres: Getting up, yeah.

40:46 Emma Sjöström: so but when you when you say scenarios like that can be to meet feels like it's such a broad like you can potentially you know, ask for a huge like scenario studio where you can you know like where Can you help me understand? Like what would be a minimum viable scenario?

41:04 Amanda Mitt: Very simple. Like Yeah, it's not even really. It's a really like, an addition calculation, basically of the balance. So you're just kind of want to check like for example in the in this because you have to do this operations. So for example, this week lucía and Rodrigo had to transfer an invest like a big amount of money, and then you

41:23 Emma Sjöström: Right.

41:24 Amanda Mitt: need to check for example, which bank accounting will need to move around or you need to move the money from an account to another account. So, for example, what would happen like in my operations? How would I coordinate with the team? Because sometimes it's a little bit of an effort.

42:06 Amanda Mitt: So what would happen if, for example, This transfer doesn't, I don't know, it's delayed. I don't get the money on this date. How would the account, how I would do my cash operations. So, I think it's really simple. It's just basically a sum so, for example, to reach that balance in x day, Yeah, a series of events needs to happen but it's very simple.

42:23 Amanda Mitt: It's basically just adding and subtracting amounts. So what if something goes wrong in this edition, what would happen, you know. So it's very simple like It's a very simple scenario. So think about your bank account if you don't salary next day, how would I would cover the rent and you just kind of need to plan, you know, it's very simple like this

42:26 Emma Sjöström: Hmm. Yeah.

42:27 Amanda Mitt: would be awesome. It's not even like, Oh my God, the whole business like, in this sense, for example, like What if I don't know the full market change is? No, it's not like this is not what I mean.

42:36 Emma Sjöström: so I I would lead a way to like just

42:37 Lucía Galán Cáceres: Yeah. Yeah.

42:41 Emma Sjöström: influence the bat would set the balance for a specific day. I either manipulating the balance.

42:46 Lucía Galán Cáceres: Yeah.

42:48 Emma Sjöström: directly or saying, Hey, this and this, this transaction which would then add into the balance.

42:54 Amanda Mitt: Yeah, delete the inflow. You delete one. Certain info.

42:58 Emma Sjöström: We're deleting. Of course. Yes, exactly. Okay. So deleting shows or adding new actuals or what?

43:04 Lucía Galán Cáceres: But it's to create two parallel scenarios, right? So to say I have scenario one, realistic scenario, two worst case, like Amanda says adjusting one manual payment. Say What if this doesn't happen or it's

43:13 Emma Sjöström: Yeah.

43:14 Lucía Galán Cáceres: saying What if we have 30% more of expenses, What happens then? So I increase 30% of the cash out everywhere and I see what happens with the balances just to like,

43:23 Emma Sjöström: Hmm.

43:23 Lucía Galán Cáceres: stress test. Yeah, I want to say, I think with scenario even the simplest scenario like this. Like What if you received 20% less of cash in that helps already?

43:35 Emma Sjöström: Trip.

43:36 Lucía Galán Cáceres: Because it makes you also prove how good how strong your buffers are, right? So if I want to have buffers in the accounts that allow me to have 20% more expense and that's a simple way to check. I say, Okay, 20% more expense is every account in green or are they in yellow?

43:55 Emma Sjöström: Nice. And then it's more like, just a final. Very, very concrete question. Do you have like any Kind of specific specific metrics or type of feedback that you get, that helps you evaluate whether you did a good job with this forecast. Just that question, makes sense.

44:17 Lucía Galán Cáceres: The variance analysis.

44:20 Emma Sjöström: Yeah.

44:20 Lucía Galán Cáceres: Matrix.

44:25 Amanda Mitt: didn't have to explain anything like Because if you go to the balance, of course, then maybe it's something can happen. And like the balance, coincidentally, you have like the exact balance that you forecasted. But then, of course, people start to kind of break it down into inflows outflows and then you have to explain like, like three and then you're

44:42 Lucía Galán Cáceres: Yeah.

44:44 Amanda Mitt: Yeah. She don't have to explain anything, it's so nice.

44:49 Emma Sjöström: So, it's like kind of self-explanatory in a Muslims, or

44:56 Lucía Galán Cáceres: zero, varians, then you don't have to explain anything. You say, my focus was amazing. Of course.

45:01 Emma Sjöström: Yeah, that's that's fair. That's fair. And I reckon it will be different for, like, the very short term where you could in theory, practice rely more on the book notes. And then for the longer,

45:13 Lucía Galán Cáceres: For the long term, I think no one expects or I think it's a reasonable to expect TMS to replace the function of controlling.

45:19 Emma Sjöström: Of course, no. But

45:20 Lucía Galán Cáceres: I think it's a waste of time, but I think for the short term, there's a lot of value to be added.

45:25 Emma Sjöström: Yes. But also like you said for the T plus 7, you might might have about that 90% of your actual snow like you're right. And would that be considered Good enough. And then coming week, two weeks, it might be a bit more uncertain. But the forecast would update closer.

00:00 :

45:44 Lucía Galán Cáceres: Yeah.

45:46 Emma Sjöström: to today. You get, right?

45:47 Lucía Galán Cáceres: Yeah, and I think that kind of insights into the data, right? Help a lot as well. So if you can see that you can see a graph and saying this is the percent, the composition of your forecast from now until six

45:57 Emma Sjöström: Yeah.

45:58 Lucía Galán Cáceres: months and you see the graphic decline in when it comes to the actual letter increasing with the machine learning.

46:03 Emma Sjöström: Exactly.

46:04 Lucía Galán Cáceres: And I don't know, like one step further would be benchmarking and saying, Hey, we believe you have two little actual data. Why is your accounting? Not booking things on time. I don't know. Or you have too much manual input. You have 30% of your cash is actually manual. Have you considered automating that? Or I don't know like that, that kind of composite that it gives you trust that the tool is working with you in order to avoid money input, enhance the machine learning.

46:29 Lucía Galán Cáceres: Also, I think for the machine learning it's a tricky piece because again, I guess it's much more powerful like it's based on historical data, right? So If your historical data has nothing to do with what's coming.

46:42 Emma Sjöström: of course, now

46:44 Lucía Galán Cáceres: And I think being able to also give context of rather going direction AI and being able to give context to that machine learning and say, Hey, actually, whatever you see, multiply times two because we have a huge marketing campaign in the UK. So for the UK, everything that has to do with sales, you will increase and the machine, the model acknowledging that I'm saying, Okay? Thank you.

47:03 Lucía Galán Cáceres: I noted down and I change.

47:05 Emma Sjöström: So would be fine. A starting off, let's say what a simple way to just throw in your broad assumptions, like like it wouldn't have to be super

47:12 Lucía Galán Cáceres: Yeah.

47:14 Emma Sjöström: duper nitty gritty but like a way to at least can Hey, 2x team close and

47:17 Lucía Galán Cáceres: now

47:19 Emma Sjöström: and it's, yeah.

47:22 Lucía Galán Cáceres: Yeah.

47:22 Emma Sjöström: And I guess if it's a completely new bank account, you might want to set it up with some assumptions or

47:29 Lucía Galán Cáceres: What we did in the past for new entities in the model that I had at salana was that we took an entity as a reference. So we said this entity or this bank account is going to behave like that one. So take the historical data from that until you have enough.

47:44 Lucía Galán Cáceres: of your own

47:45 Emma Sjöström: Smart. Yeah.

47:46 Lucía Galán Cáceres: If you can do that, then I think at least you copy the behavior of a different one.

47:50 Emma Sjöström: Yeah, that's a good point. Would you say that's the typical thing that you can do or is, I guess it's very context and company specific. But

47:57 Lucía Galán Cáceres: I don't know if I would say typical do you think Amanda?

48:04 Amanda Mitt: Yeah. No I think it depends a lot on. The business model I think. It depends, I think really hard.

48:14 Lucía Galán Cáceres: If you have I think it's it does depend but if you have a business model where you replicate the same, for example, like we replicate ENT. entities in different countries, but

48:24 Emma Sjöström: Right.

48:24 Lucía Galán Cáceres: the purpose of the entities exactly the same just Intel instead of Spain that I think it's helpful to say. Hey, it's a twin company just copy the data the behavioral data of pattern of Spain to Italy. Of course if it's a completely different business model, you need to be able to sell the model.

48:42 Lucía Galán Cáceres: Hey, you've not seen anything like this in our company before these are the broad assumptions but start from

48:48 Emma Sjöström: Yeah.

48:48 Lucía Galán Cáceres: Right? So the more flexible, the model is The better for that.

48:53 Emma Sjöström: Yeah. I'm currently a bit interested in, in ways of sort of, keeping the human in the loop, and just interested in, in what ways? You as humans would want to be kept in the loop and ways. You want to be able to influence, right? What I'm

49:08 Lucía Galán Cáceres: Yeah.

49:08 Emma Sjöström: here. What I'm hearing is like Inputting large one-offs. right, or the scenario bits or even deleting permanently like Payments or stop not gonna happen. So actually updating the forecast based on you having knowledge, the models don't

49:26 Lucía Galán Cáceres: Yeah.

49:27 Emma Sjöström: Scenarios. and then like maybe throwing in some broad assumptions, that would actually also affect the model and those don't really inputs would not

49:37 Lucía Galán Cáceres: Yeah.

49:37 Emma Sjöström: it would be more like you're saying please like everything for this

49:41 Lucía Galán Cáceres: Context.

49:43 Emma Sjöström: for this group of category just double or stuff like that.

49:45 Lucía Galán Cáceres: Yeah. Yeah. And I think A lot of the times with with forecast models, the problem is that you throw all of that stuff and then you see the outcome. But it's difficult to bridge. What you throw in with what you see, and that makes it for me and trustworthy if I cannot bridge it.

50:02 Lucía Galán Cáceres: So if I threw in 100 million in sales and I see 150 in or one million sales distributed over six months. My question is why six months why, why not three like where. So and I mean, if you the model that I use in the past, we define the logic.

50:18 Lucía Galán Cáceres: So I could do step by step table by table the calculation, myself. And then I would see where and I would define how the tables will be computed. But the moment you're going to machine learning, you're letting that go and

50:29 Emma Sjöström: Yeah.

50:29 Lucía Galán Cáceres: instead of the rule you have set of logics that you can follow, you have A model and this is where I think. Honestly. I said, Treasure you. You want to be able to understand this without being a statistician, you know? But You need to get to, I think that's as you want to be in the loop with the model.

50:49 Lucía Galán Cáceres: You have to feel like you're contributing and understanding what the model is doing. Otherwise I wouldn't present this to my management.

50:54 Emma Sjöström: No.

50:55 Lucía Galán Cáceres: So, this was created by a machine, but I don't want to take responsibility.

51:00 Emma Sjöström: Makes sense.

51:00 Amanda Mitt: Know, like, Who would you blame in the end? Write them so funny. Yeah.

51:03 Lucía Galán Cáceres: Yeah. Yeah.

51:06 Amanda Mitt: Could you chat like with the model? For example can you please like create some sort of Like, I don't know.

51:13 Emma Sjöström: Why not, right? Yeah.

51:14 Amanda Mitt: Yeah. Yeah, like us outside chat.

51:20 Lucía Galán Cáceres: Yeah.

51:21 Emma Sjöström: Like almost like a chat tpt like but

51:24 Amanda Mitt: Yes.

51:25 Emma Sjöström: but it would update stuff in the app for you.

51:28 Amanda Mitt: Like, why did you forget? I

51:30 Emma Sjöström: Questions.

51:31 Amanda Mitt: Yeah.

51:34 Emma Sjöström: I mean that would be really awesome. For sure. It might not get there in the first iteration or two, but

51:43 Lucía Galán Cáceres: That's the goal.

51:43 Amanda Mitt: Yeah.

51:46 Lucía Galán Cáceres: When do you think you'll have the next iteration of this Emma? What's your timeline?

51:49 Emma Sjöström: Oh, the forecast. So, I mean, I'm hoping that we will know what to

52:11 Lucía Galán Cáceres: and I, Yeah, for enough.

52:18 Emma Sjöström: which awesome for us, you know, on our velocity next year. But it's

52:19 Lucía Galán Cáceres: Yeah.

52:19 Emma Sjöström: slowing us down a bit right now. Yeah. So But yeah, I'm actually quite excited that we're getting you on board on the product. So,

52:24 Lucía Galán Cáceres: Cool.

52:25 Emma Sjöström: Yeah, and happy to yeah, I know.

52:26 Lucía Galán Cáceres: Thank you.

52:29 Emma Sjöström: out of time but I just want

52:30 Lucía Galán Cáceres: You know.

52:31 Emma Sjöström: Yeah, happy too. If you have feedback, like when you're using it more actively, please.

52:34 Lucía Galán Cáceres: We'll do, we'll do will do and we have the stand-up study next week, right?

52:40 Emma Sjöström: That's true. Good. Yeah.

52:42 Lucía Galán Cáceres: I just have to run because I have another meeting at three years to prepare something before.

52:46 Emma Sjöström: Okay, thank you much.

52:48 Lucía Galán Cáceres: but, Nothing so much, Emma. Thank you. Amanda by

52:50 Amanda Mitt: Oh, thank you for interesting.

52:50 Emma Sjöström: place to Thank you, likewise.

52:54 Amanda Mitt: Bye.

52:55 Emma Sjöström: Hi Amanda.