SDS PODCAST EPISODE 6 WITH XINRAN LIU

Size: px
Start display at page:

Download "SDS PODCAST EPISODE 6 WITH XINRAN LIU"

Transcription

1 SDS PODCAST EPISODE 6 WITH XINRAN LIU

2 This is episode number 6, with expert financial modeller, Xinran Liu. (background music plays) Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, data science coach and lifestyle entrepreneur. And each week, we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let s make the complex simple. (background music plays) Hey guys, welcome to the SuperDataScience podcast. Super excited to have you on board today. This is a very special episode to me, and not only because I have my dear friend Xinran Liu today on the podcast, but also because this is an episode that completely debunks the myth that data science is only for males or for men. That is not true. There are plenty and plenty of professional female data scientists, financial modellers, economists, and so on. And Xinran is a great example of that. She is a best professional at her job. Xinran is a financial modeller. So I've actually had already students me about inviting a female guest to talk about data science, about statistics, and just to show that anybody can master this profession, and here we go. So Xinran is a great example of that. So what do we talk about in this episode? So we used to work together with Xinran on some projects back at Deloitte, and while I was in the data science department, Xinran was in the financial modelling department, and we did have one project together and then our departments worked on several projects together as well. So we'll talk about that,

3 how data science works together with financial modelling, and moreover, it's not a completely distinct line between the two. In consulting, maybe. But in other companies and large corporations, you'll find that financial modelling and data science, they overlap often. It depends on the structure of the company, and a lot of the skills overlap. So we'll talk about the skill sets. In fact, Xinran actually took some training in data science, and she was exposed to that side of the profession, and she will be commenting on which skills are similar, which skills helped her develop a more intrinsic understanding of financial modelling and data science. And also we'll talk about how data scientists can be successful in the field of financial modelling and vice versa. So this episode is packed with career advice such as that, and also we'll go into a lot of depth on financial modelling. So you'll learn about simulating financial situations, assets, portfolios, businesses. We'll talk about the importance of sensitivity analysis and what it actually is. We'll talk about tornado charts. We'll have a very deep dive into modelling assumptions and why assumptions are very important. And this doesn't only apply to financial models. This applies to any type of models. What's the difference between inputs and assumptions, and why making the right assumptions is so important, how that affects your output. We'll talk about data science skills in financial modelling. So I've already mentioned that. And we'll talk about something that's already come up in this podcast, something that we talked about with Wilson Pok in episode number 3, we'll talk about how do you report distributions. In data science, we think in terms of

4 distributions rather than one single outcome, whereas executives, they often think of KPIs and think of targets and business results as one number. So how do you report a number that actually has a distribution in a way for executives to be able to understand it. And Xinran comes up with a very interesting suggestion, so we'll talk about scenarios, reporting different scenarios as outcomes of modelling, and I found that a very insightful part of our conversation. A very interesting part as well is closing billion dollar deals. Not million, not hundreds of million, but actually billion dollar deals. So Xinran has worked on projects up to that number. So you can imagine what stress and what importance her work carries. And we'll talk about the steps involved in building a financial model, we'll talk about SQL and how that helps develop an analytics mindset. So those are just some of the points that we cover off in this podcast. I could go on for much longer just describing the things we talked about, but it's better if I just leave you to do it and let you listen to this interview. I'm sure you'll pick up lots and lots of great knowledge. So without further ado, I present to you Xinran Liu. (background music plays) Hey guys, welcome to the SuperDataScience podcast. Today I've got Xinran with us today. Hi Xinran. Thank you for joining us. How you going today? Hey guys, welcome to the SuperDataScience podcast. Today I've got Xinran with us today. Hi Xinran. Thank you for joining us. How you going today?

5 Hey, I'm good. Thanks for having me. Thank you very much. Happy to be here. Yeah, it's awesome to hear you. I remember back in the day, we used to work together at Deloitte here in Brisbane, and then you moved to Melbourne. It was quite a while ago, right? When did you move to Melbourne? It was about 2.5 years ago, yeah, early It was a bit of a rushed decision, but I was after something different, and Melbourne could offer the job opportunity. So yeah, it was a quick move, yeah, but quite a while ago. Yeah, yeah. It was great working together here. Do you remember actually how we met? Was it over a project? What were we doing when we met in Deloitte? I think it was some mini project that we were working together. You were with the data science team, and I was in financial modelling. I can't remember exactly what the project was, but there might be something that we needed an expert in the programming and data science area to fix that feed into our financial model. And yeah, I remember it was a good experience. Yeah. Or it could have been the other way. Because I remember doing a project for the Queensland government, and it was about -- no, it was your project. That's right. It was your project. It was a project I think for doing a survey around doing some Queensland government stuff and you guys needed some data science input.

6 Something like that. Yeah, it was a race -- I think something relating to a planning for one of the government departments and there was a large amount of data involved. Correct. I think it was a STEM department, the Science, Technology, Engineering and Maths, and they were researching how many people are studying these skills across Australia. Something like that, right? Yeah, yeah, could be. You are talking about like 4 years ago! But it was cool. Yeah, and it's always fun working at consulting. I don't know if it's across all consulting firms, but at Deloitte, you gotta agree, we had a lot of this where departments would share expertise, right? We would -- Exactly. If you need something from us, you would come over and we would work together, or you need an accounting person, you always have the right skills in the company. That was really cool. Yeah, yeah, that's actually the really good part about this consulting industry. Like to start working there, it's like every project I was working on as much as like working hard there, contributing my expertise and ending up learning something new from different parts of the service lines. Like I learned some programming things from you guys, a bunch of accounting, and even forensics skills from other service lines. It feels like talking to experts every day is great. It's truly a really good opportunity to learn working there.

7 Yeah, yeah, definitely, for sure. And remind me, what was the name of the service line you were in there? I was in financial advisory, and within which the financial modelling team is quite a specialised area within the financial advisory team. Ok, yeah, that's pretty cool. And you were building models. What kind of models were you building at Deloitte? Various types of models, primarily based on Excel and spreadsheets. Financial modelling is I guess a field that's relatively niche, sort of like data science a few years ago. It's to build a mathematical model to simulate any financial situation, performance of an asset or portfolio or business. Primarily also it helps to assist in business decision making to help you make some reasonable decision, maximise your return, whether that's a business valuation, whether that's an investment analysis, project finance, or simply your financial performance. That's the reason for most people to build financial models and to inform and interact with stakeholders and executives as well. So that's a tool to use as a tool in the finance field. Yet behind that is a sound financial decision to make, if that makes sense. Yeah, that's really cool. And is that the same thing that you went to do in Deloitte Melbourne when you moved? Yeah, I was in the same financial modelling team in Melbourne as well. Melbourne probably has a slightly different industry focus compared to Brisbane. In Brisbane, it's primarily energy and resources at the time. It would be the public sector, etc. Melbourne, however, has a larger focus in healthcare, telecommunications. So the same skills, different exposure to a wider range of industries.

8 Ok. Alright. Pretty interesting. And did you find that moving inside consulting from one industry to another, did you find your skills were transferable? Or did you have to re-educate yourself in a lot of that space? Yeah, really a bit of both. The financial modelling skill is certainly transferable, because finance principles are more or less similar across industries, with some differences. But other industry knowledge is very different. And the way of doing business, the thinking in terms of the priority to be considered when looking at investments are very different. And government regulation as well is a key consideration. For example, energy and resources and lots of the infrastructure relating to a trend distribution or electricity transmission are highly regulated. So the ways of considering them are completely different from a free industry like retail. So yeah, like I mentioned before, there was a lot to learn at every project. But the financial modelling skills helped a lot with it. I find the same thing in data science. That a lot of the skills are transferable that you learn and that you know about dealing with data. Same, I guess, in financial modelling. The actual building the model part. But then whenever you move, you just have to catch up on the domain knowledge, the industry specifics, and off you go. It's not like moving to a completely different type of job. Yeah, yeah, exactly. So that always helps. It's always a bonus of being in these industries, in this type of work. That you can move around and you already have the expertise, even if you move to a different industry very quickly.

9 Exactly. And I guess interestingly, for me at least, you get the opportunity to look at different industries being able to offer the financial modelling skills, but also able to learn a lot about different expertise in different industries. That was a key bonus for me. Yeah. Totally. And our listeners might get a bit confused thinking that we're talking about Deloitte so much, thinking that you're still at Deloitte. You recently started a new job. Congratulations on that. What are you doing now? Yeah, thank you, thank you. Yeah, I started a new job, and moved from consulting to industry. This new company called Gemini. It's an electricity and gas distribution company within the infrastructure, energy and resources area. The primary markets are Victoria and New South Wales, but also has asset exposure, portfolio around the rest of the states. Yeah, it's for me a really great move, because I've always wanted to do energy resources. And this new role allows me to use financial modelling skills, but also has given me the opportunity to look at lots of other areas of the finance function, like investment analysis, assist with mergers and acquisitions, assist with impairment testing, or long term target setting, etc. It's really about understanding one industry, and using, I guess, still my financial modelling skills. So that's good. Yeah. So you can kind of like dive deep into this industry and become very proficient at it and develop those financial modelling skills specific to the gas, energy, and resources industry. Yeah. And extending from financial modelling, I guess, and access to really having the experience and understanding of

10 that industry such that I could make assumptions to make my financial model more accurate to reality. So it's financial model, I guess lots of people understand it as sort of a black box. But really, it's just like a tool that you build in, spit out a bunch of outcomes, and that's primarily driven by your methodology and assumptions. What the assumptions are is really -- how you make the assumption makes the key difference to your outputs. And that's really where the experience and knowledge kicks in in that industry. So yeah, it's about really modelling is a tool, but it's really about making the sound financial decisions behind it that's the key. Yeah, yeah, and interestingly, that's actually the area I think where data science could be working really well together with the financial models, or financial modelling area, where it's to help making the assumptions more profound and more accurate, closer to reality. It sounds very interesting, and especially like gathering those data insights to create those assumptions in the first place. And can you tell us a bit more about what are your models about. Because there's so many different types of models that come to mind, especially like from a data science perspective. What models do you build, and what are their purpose? Yeah, sure. I think I'll give you an example of some of the models I'm recently building. So say Deloitte's working for a company that is looking to acquire another business in a similar industry. The business is publicly listed and we are really building a model to simulate that situation to valuate the business and also combine that business, the target company, with the existing company to see what the

11 synergy outcome is. So my model would have an input section where it's the target company's financial statements, and the potential different scenarios of your contract price, which is how much you pay to buy the business, and the legal fees, etc. And then with a range of calculations in the middle, depending on what scenario we're going with, and the output, it would be under different scenarios and different market situations, what are the outcomes after the acquisition? We will be looking at a range of metrics like earnings per share and FFOT. A range of financial metrics that's fairly standard to compare with the other transactions to see whether this kind of deal would make sense or not. Alright. And so in that case, one of the companies, the acquiring company, is the company that engaged Deloitte in order to make this assessment and prior recommendations, is that correct? Yeah, a bit of both. Sometimes in this example, it was the acquiring company. In some other cases, it's also the target company. If they know that they're going to be sold, they want to obviously have their own prices in mind as well, that is to help with the negotiation on the table. Normally, both sides would hire investment bankers, financial modellers, accountants, and legals to proceed with the deal. And the financial model is a key part of it to come up with a number. And we're talking about massive deals, right? So if you're able to disclose approximately, not specific to this deal, but other deals you've worked with, what are the approximate amounts that are paid in these cases?

12 Oh, they could be a range of different cases. Like for some of the potential deals we're looking at, they could be billion dollar businesses. Billion dollars. Yeah, yeah, absolutely. Woah. That's why people get very sensitive around the financial models, like how the number is coming from and is making off, and where it's coming from, what's the sensitivity, etc. It's a very stressful process. Often in the transaction, it's about the financial models. Because that really changes a lot about the price, and there are various cases where the financial model could have an error, which happens. And then the prices change a lot, and end up having the clients sue the company. Not Deloitte per se, but some of the financial modelling companies for the case because if there's anything happening in the model that's wrong, and if the prices go into the contract that's later found out to be wrong, there could be a big impact. So we have got to be really careful not making mistakes. Alright. That's really cool. And so like a lot of due diligence is involved, and auditing, and that kind of stuff, yeah? Yeah, exactly. That's a billion dollars! I just can't -- it doesn't escape my mind. A billion dollars. I remember working on a project back at Deloitte that I think our department worked with your department, and we were creating a model like that together. But that was only like $48 million or somewhere around that mark, that ballpark.

13 Oh yeah, that could happen too, depending on the size of the business. Billion dollars obviously is big cases. But other smaller transactions, depending. Sometimes they are not buying the whole company, they are just buying some of the assets, say a component of the assets for strategic reasons. That could be a million dollars, or a couple of hundred million dollars, sort of business. Yeah, the bigger the deal size, the more DD, due diligence, involved, I would say, obviously. Yeah, that's true. I'll just give an example of when our department worked with your department. I think it wasn't you. I think you already left to Melbourne at that point. But it's a good example of when data science can work closely with financial modelling. And what we were doing is, there's a company that was also being sold like that, and Deloitte needed to create a finance model to forecast the profits of the company for the next couple of years in order to calculate its current value. But in order to calculate the profits for the next couple of years, Deloitte had to calculate the amount of customers, or clients, that the company is going to have in the future. In order to calculate that, because this was a facility that was dealing with a specific medical condition, we had to create a model that would forecast the spread of this disease across the population of Australia in order to calculate how many. It was pretty cool. So we had to first create a model for the spread of this disease, in order to calculate how many people would contract it in the future, in order to calculate how many customers this company would have over the coming years. Then subtract the amount of customers that would go to

14 competitors. And then supply all of that information to your department, from which you would create a model. Yeah, yeah, yeah. I can imagine how that's coming together. Yeah, that's really cool. I think that's where I was also thinking of how data science could come up with utilising the mass amount of data behind to come up with more profound assumptions that go into the financial model, which significantly helped with the accuracy of our models as well. So yeah, sounds like a really cool project. Yeah. At the same time, it's not such a distinct border between data science and financial modelling. Because a lot of the time, data scientists in other companies, and specifically in smaller companies, or in the industry, might find themselves doing the financial modelling. Or like what you're going to be probably, I'm assuming, doing in the industry now is going to be more broad than what you were doing specific in Deloittes. So you're going to have more exposure to the data science elements. And why I'm bringing this up is because it's good for data scientists to know what goes into building a financial model, what steps you go through. And I was wondering if you could walk us through the steps that you go through when you're developing a financial model, or any other types of models that you've been working with. Yeah, yeah, of course. I think different models would probably have slightly different procedures. But they do have something in common in terms of the procedures that we would look at putting a model together. And different modellers would have their own habits of where to start with and how to control the accuracy of the model in the middle.

15 So normally I would start with talking about the outputs, so what is the main purpose of that financial model? Obviously if it's a transaction, what's the key metrics, what are the key outputs, having that set up, and then going into the input area. To achieve that outcome, obviously what inputs do we need? And what inputs are available there? So as you set up the input accounts, normally include the financial statement, P&L, balance sheet, and cash flow, capital investments, and things like regulation changes and other market condition changes, things like the spread of the disease, as you mentioned. So all these special events would have a material impact on the output of the model. Having this to set up, I would say, takes up 80% of the time in a lot of the cases. It's about going back and forth between us and the client, whether that's internal or external clients. External clients when I was a consultant, or whether that's internal clients now that I'm in the industry. But it's really about locking in the output of the model and making sure the inputs that go into the model are right. And then we link the calculation in between. For most experienced financial modellers, that would then take as long as the output and the input section. It's really about linking the calculation, apply financial principles, and the mathematical principles depending on what case this is. And I guess the key part in my understanding of where data science could kick in is really the assumption and the input section, where say for example, we need revenue forecast for the next five years. Or if we need detailed revenue build up, or opex build up, and the drivers. We know the drivers, we know the key factors that influence it, normally we would probably take historical numbers in the past five years. But

16 now with data science involved, we could do a more bottom up build up approach such that it's using the real operational data and closer to reality. Very interesting. And interesting that you mention that the set up of the model takes about 80% of your time. It's very similar to what we experience in data science. From my estimates, based on my experience, setting up the project, getting the data, cleaning it, and preparing it for analysis, and actually understanding what you want as the output, it also takes about 70% of the time. And then the fun stuff, the actual modelling, and the visualisation, and preparing results, that's only 30% of the time. So that's a very interesting trend. That's very interesting, right. I remember I was working a bit on some of the data science projects [25:20 ] but I was trying to write some SQL procedures. But it's really about setting up the input and output, making sure that the structure was right. Well, that's what I was told anyway, before I went in and really build up the details in between, and the calculations. So there's also a lot of similarity there for sure. Oh wow, that's very cool. And how did you find doing a bit of data science work? I would say in a training programme from Deloitte for a couple of weeks, I learned a lot about SQL. Well, I learned the basic fundamentals about SQL, and a bit of other programming languages, and really about writing procedures. And what I find is really a lot of the projects are similar to what a financial model could do, but on a more operational focus. Which is really great. Our financial

17 models are primarily in, like we mentioned, the investment and financial areas. Whereas in the operational side, there's also a lot of data to be utilised, and a lot of decisions to be made. So it's good that the data science could actually fill in this box and help the decision making in the operational side, things like your customer focus, your products, preferences, etc. Which is really inspiring to see. Oh wow, that's great. So you've had some exposure both to the financial modelling side of things and now you've slowly gained some experience with the data science side of things. That's a very rare thing to have, that you can now speak from both ends. I'd like to ask you a couple of questions around that. Just for people who might be in a similar situation as you, and might be wanting to make the move to just raw data science, and see how they can transfer their skills. What would you say about SQL? Was it hard to learn SQL? What do you think of the programming language and how is it different, for example, to Excel, and was it was generally just a challenge to grasp the concepts of SQL? I honestly don't think it's going to be too hard for a financial modeller to move into data science and vice versa because like say in financial models we've got lots of VBA stuff, Excel macro work, where it's similar. Even though it's a different programming language, it's very similar to when I learned SQL and R and Python, even though I didn't do much of that. But that certainly helped with the way of thinking. And both building financial models and building SQL models, I feel, requires a systematic way of thinking about approaching a problem, breaking down a question, and

18 really structure it in a way that it could be answered by questions, by numbers, and calculations. For me, the key challenge probably was the amount of data I was dealing with. Obviously in the financial modelling field, there will be quite a bit of data. Nothing compared to the amount in the data science area. And the different level of details as well. So in the financial modelling area, we normally look at things at aggregate level, total revenue, or revenue at different segments. Whereas in the data science part, it's more a build-up approach of what's the detailed transactions that happened at an everyday level. So that's a really big transition in terms of mindset. Yeah, and I agree, lots of data, especially if you're looking at daily transactions for a huge organisation. It's interesting that you mention that SQL builds up a way of thinking. I also find that that, to me, was one of the biggest advantages of learning SQL. Not that I got to know a new programming language. Not that I got to know how to prepare data sets, even though that's important. For me, SQL helped build up my mindset in a way that I can think about business problems in a data perspective. I can think about how this data set would be structured. What would be the inputs, what would be the outputs, what would be the columns, how many rows it would have? Yeah, when you solve a business problem, you instantly translate that into a data science structure model, yeah. Yeah, that's right. Totally. So tell me this. I had this interesting question, because in one of our previous podcasts, it was in episode 3 with Wilson Pok, he mentioned

19 a very interesting thing that when you're dealing with forecasting or modelling in data science, it's a bit different to when you're dealing with forecasting and modelling in the finance area, or in the financial models. And specifically that in data science, we tend to think about modelling and results, specifically outputs of our models, as a range. It's not enough for us to say that the value of this metric is going to be, let's say, profit is going to be $500,000. We say $500,000, but it's in the range between $490,000 to $510,000. This is the standard error. This is the mean. We think of distributions for our outputs. Yeah. Is that the same with your financial modelling? Or are you more used to actually delivering specific numbers for whether it's profit, or it's capex, or other things that you find in your financial statements? Yeah, I think we deliver exact numbers, but with different scenarios. Obviously, it will be high, low, and medium scenarios. For each, there will be an output. So in some sense, we are still looking at a range of outcomes. But to help with the decision making and also in the acquisition, it's got to be a contract price. There has to be a number to be delivered. But I think that's similar to data science, that you're looking a range of outcomes where the management might look at that and go, "I'll go with the midpoint." Or "I'll go with the high end." So it's quite similar. We do sensitivity analysis in the financial models as well, where as your driver goes up or down by 5%, how that would impact your outcome or your key dollar outputs. That's sort of similar to how data science would approach it, I think.

20 That's actually a good thing, and I think data science can learn a bit from that. The distribution approach is not always easy to convey, easy to explain to stakeholders, right? It's distributions. People don't think in terms of distributions, unless they're trained to. Whereas your scenarios approach, I think it's easier to explain, right? You just go up to a stakeholder and say, instead of just saying, hey, we've got this distribution, you can say, hey, we've got the low, medium, and high. Yeah, yeah. We have got three numbers for you to choose from. Exactly, yeah. That's actually a really good point because there's a key problem in the financial models which I suspect data scientists might be facing in a similar way, is that to talk to executives and especially people in the more senior positions, say. They think very much from experience. And from a non-scientific, mathematical model, data science perspective very much not like that. So if you start talking to them about distribution, talking to them about your range, your sensitivity, when it comes into technical terms, they will not be interested in that and quickly get bored from it, and they're just like, I want to know my answer. So it is very important, I guess, to translate the financial model concepts into a way that's very much wrapped by the business concept and the business background such that people in the business can actually understand and relate to. Ok, yeah, that's definitely true. And can you tell us a bit more about sensitivity analysis? Because I know that this is

21 a very specific type of term to financial modelling, and from my experience, we don't tend to do that much sensitivity analysis in data science. And at the same time, I think it could be helpful sometimes. What is sensitivity analysis, and maybe you can give us a couple of examples. Yeah, yeah, sure. Interesting, because yeah, sensitivity analysis is an extension to our financial model output, I would say. Say you build a model in a very simple way. Your volume times price gives you the outcome discounted by your discount rate to give you a current value, just as simple as that. You could run sensitivities on different drivers. It could be your volume, say you suspect the customer subscriptions could be 10% more, 10% less, to see how that would impact your output. And you could run a sensitivity, just increase the discount rate by 5% or add it up by 3%, because your assumptions, I suppose, can never be really accurate, and hence you look at a range of the assumptions to see, if I change that, what's the output of my model? When a lot of drivers come together, I guess you could roughly what the direction of the outcome should go with. But you wouldn't know unless you play the sensitivity analysis to actually change the inputs in the model to see, you wouldn't know the output of it. So that's where the financial models as a tool come in to help, is it tells you the output, and obviously in a more sophisticated model there are lots of easier ways to do the sensitivity analysis. You could set up things like tornado charts, like key drivers being lined up, and play sensitivities are recorded to give you a chart that looks like a tornado, where people can see what are your drivers that impact your business. For, say, energy it could be probably your discount rate, or your

22 regulated price. For other industry, it would be something else. So it's good to know what your key drivers are so that you keep an eye on those items, and at the same time, it's good to understand what a best case scenario or worst case scenario is for your business as well. Wow, that's really cool. I've never heard of tornado charts. Definitely, after we finish this interview, the first thing I'm looking up is tornado charts! And so it sounds pretty cool, right? Tornado chart. So the point of the tornado chart is to line up your inputs in terms of your output's sensitivity towards the input. Correct me if my understanding is wrong. And then from the tornado charts, you will be able to see which of your inputs has the most impact on your outputs. So to which inputs your output is most sensitive, right? Exactly. Ok. That's a really cool concept. Definitely going to have a look more at that. And so would you say, I'm just thinking here, from what I learned back in economics, would you say that price elasticity for a given product is a case, or an example, of sensitivity analysis? Yeah, I would say price elasticity is a sensitivity factor. So if your revenue is very sensitive to your price, then the price as a driver, or as an input, will come up on the top of your tornado chart as a sensitivity factor, and it will see a big range in that. So you change price by, I don't know, 2%, you might see your revenue or your profit change by 10%. That could implicate a high price elasticity. But you know, that's the beauty of a simplified chart without having to bring in the concept of price elasticity. Because you don't have to

23 explain about it. Now this chart will just tell you that your profits are very sensitive to your price. Ok. That's a very good concept. And the other question I had was in terms of assumptions. What role do assumptions play in models? How important are assumptions, and how do you go about finding the right assumptions for your models? What's the whole process? Because for a lot of people, assumptions are not that important, or a very new concept. So what can you recommend to people who are just starting to deal with modelling, and just starting to learn about assumptions? Cool, assumptions I suppose are a very wide range of stuff. You could call it like inputs, or for the inputs that you aren't too sure, you call it assumptions. Things like CPI could be your assumptions. Like how will prices change over the next 10 years, your CPI. Do you use 2.5%, do you use 1.5% cash rate, or do you use a higher rate? That's really a judgmental call based on your experience and your economic knowledge of that specific industry and that specific country and region as well. So that's so called the assumptions. How do you get the assumptions right? That I would really like to know! Like, really talking to experts in different fields, for us, I guess when we're in consulting, we spend a lot of time with clients from different departments. From treasury, from finance, corporate finance, groove, talking to people up on the ground, even those people who are facing the customers, so that from talking to them, they understand how the business is really running such that they would actually be the people who know about what assumptions to make. So from that sort of conversations, normally it's in the form of workshops, or even casual conversations, we get some sense

24 of what assumptions we should make about specific problems. And in the industry area, it's similar. We still go out and talk to people at different business service lines, and people who understand the real problem the most such that we could make the decisions. And there will be more scientific ways of making assumptions as well. We go up to databases like Bloomberg, like Thomson Reuters, to find out about your bank rate yield, the cash rate, your foreign exchange rates, to obtain the data, collect data inputs or assumptions, but to get more accurate information that feeds into a financial model. But it's certainly a very, very important part to get your assumptions more or less in line with reality to make sure your financial model works. Yeah, definitely, assumptions would play a very big role. Absolutely. Absolutely. For sure. Alright. Moving on to more questions around how you got into this field and what are the prerequisites, what would you say is an important part, like something to study, in order for somebody to get into the field of financial modelling? Sure. Financial modelling is a really practical area. People -- like data science, I imagine -- people come from different backgrounds. Personally, I studied economics and finance at uni, but I know people who studied engineering and finance, people who studied commerce come in to become a financial modeller. It's a relatively niche and small community around the world. There is a really good competition called Model Off, where the financial modellers come together every year, have a competition, and there is a

25 ranking around the world where you can see who are the top financial modellers in this world, and people get together and exchange ideas and new techniques, improvement in your tools, etc. So it's a very helpful community to keep an eye on. It's called Model Off. Model Off, right? Model Off. Just had their latest competition, I think, not too long ago. Ok. Did you get first place? Haha, I didn't participate, unfortunately, but I've seen quite a few participants from Australia that win the top 10 spaces, data scientists. Interesting. So that sounds like something similar to Kaggle for data science. Yeah, yeah, very similar to Kaggle in data science. Yeah. Ok. That's interesting. And what are the normal tools, or the tools that you use on a daily basis? Like, is it Excel? Or do you use any other tools for your modelling? Yeah, I use Excel with a range of various add-ins, Excel add-ins primarily. So we use Spreadsheet Detective for model audit. I actually use that for building my model as well, because Spreadsheet Detective helps with shading the model such that you can see whether you can drag the formula across or not. At Deloitte, there are some specifically developed add-ins that help you to give you the short cut flip through your dependents and precedents of the formula, etc.

26 I wouldn't say that's the most important part of the financial, though. Anyone with a spreadsheet could start building a model, that's what I think. Because really, modelling is a tool. How you link the formula is just a part of it. What's really more important is the structure of your model, and how do you approach the business questions, and how does your model help with answering it, is a bigger question. Alright, it makes sense. And what is the biggest challenge in modelling that you've ever faced, personally, in your 5+ years at Deloitte. What was the biggest challenge that you can think of right now? There are challenges in the financial modelling field. My probably personally biggest challenge so far faced is a very blurred business question that we don't know what we are asking, or we don't know what the model is trying to answer, such that we develop the model step by step and the question and the purpose of the model kept changing, whereas you need to quickly make sure the structure of the model changes so that you could answer the question. So I think that could be probably a pain for lots of financial modellers, which is the change in structure. Minor changes are fine, but bigger changes in terms of the purpose of the model, if that changes, it's really about counting on your experience of how to create the model to make sure it's still working in a very robust manner. It's very interesting that you mention that, because in data science, it's also one of the biggest challenges, identifying the question. So it's not as fluid, so once the question is identified, it probably won't change as much, from my experience. But at the same time, going about identifying

27 the question and asking the right question is very important. Because at the end of the day, if you ask the wrong question, or if it's not precise, you'll end up doing the wrong analysis. Yeah, absolutely. And about helping develop the right question to ask as well. Because a lot of the time, there is a problem, but you have no idea what we are trying to achieve, or how to answer that, how to solve that problem. We're there to help with that, but that's a whole different topic, that's our key challenge in this field. Yeah, totally. Alright. And what is your aspiration in terms of modelling, in terms of what you've seen, the experience you've been through, what is your career aspiration, where do you think you want your career to go from here? For me, I guess having spent about 5 and a half years at the Deloitte financial modelling team, I've really wanted to go into industry, which I have just done, to become more of an expert, to develop my expertise in one field and one industry. Being able to utilise my financial modelling skills and help with real business decision making, being able to understand everything about energy and resources, being able to, I guess in a very simple sense, being able to come up with assumptions in my financial model that are very close to reality and really help with a real business valuation and investment analysis. That's sort of where I guess I'm going. Because I guess as I mentioned earlier, financial modelling, as much as a it's a tool, it's really helping with business decision making. You could have your model absolutely perfect and beautiful, but if that doesn't help with the decision making process, then it won't be as helpful. I guess it's a two-way sort of thinking about that.

28 One is to build a model, the other is how to use it as well. That's, the second part, is where I'm heading to at the moment. Ok. That's a great way to be going. So how do you find the difference between industry and consulting? Because it's funny that on this podcast, we have people going both ways. We have people going from industry into consulting. From consulting to industry is like the natural transition, I guess. But still, we had people that go from industry back into consulting. So what is your feel? What is the biggest change for you? What do you think, are the hours better, is it less stressful work, or is it the opposite? What do you feel? Generally the hours are better, but in my opinion, whether the hours can be better or worse is really up to you. I've seen people who work really hard in consulting, and I've also seen people who work really hard in industry. And different parts of industry are different as well. There are certainly these types of opportunities in the industry itself. In consulting, what we were after mostly is a variety of projects, getting exposure to different types of things. I guess the trade off of that is I don't get to spend enough time in one industry and one area. Going into industry though is about one business, but a different side of that business itself. If you look at consulting being looking at a different industry from one angle, working in the industry is really looking at one business but from different angles. That is definitely the case. I also find that when are working inside a specific company, because I also made the move from consulting into industry, a major difference that you see that you are able to follow through with your project.

29 In consulting, you deliver a great project to the client and then you move on. And you never see the results, you never see the change it drives for the business, for the world, you never feel that meaning. The impact. The impact, exactly. You don't feel that impact. I know exactly. Whereas now, you're working in industry. Every model you build, you'll be able to always see what it changes and be proud of your results because you're actually working in that same company. Exactly. And in some other sense, you can't wash your hands off it. Exactly what you said, just that often in consulting, we build a financial model and at the end of the project, we hand it over to a client. What happens after, the business decision being made, whereas in industry, every project's a long term project. So in the industry, for the financial model, we're on it for a long time. It's not very often that we deliver that to someone else. Therefore, we do need to be careful about what to suggest, what to say, about every step we're taking and the reward for that is you get to really see how the projects evolve, how before the decision was made and after, what impact it has, and even the long term impact as well. A lot of times, the project might be put on hold, the acquisitions or investment may not go ahead for the moment, but in a year's time, it might come back. So it's a long term strategy working in the industry. Yeah, yeah, definitely. Hopefully your experience of moving to the industry will be a great one and you'll find your true

30 passion and develop that expertise that you're looking for. Thank you very much, we're coming to the end of our podcast here and really appreciate everything you've shared with our listeners here. In terms of contacting you, is there any way that our listeners can get in touch in case they want to follow your career or learn more about what you're doing in the world and how you're impacting the different sides of businesses? Yeah, I use LinkedIn. So if you search Xinran Liu in Jemena, then that should have my name come up. I'm happy to put a link through of my LinkedIn profile. If anyone has any questions relating to financial modelling, more than happy to answer them and I will do my best. Awesome. Thank you very much. We'll definitely include that in the show notes. And so to finish off, I have one more question for you. What would you say is your one favourite book that can help our listeners develop their careers? Yeah, there are various books I come across that help, and one in particular that's recommended by one of my mentors is called "The Effective Executive", written by Peter F Drucker. That book is really talking quite generally about what the key issues you could be facing at work, and how to address them and really I guess to some extent, expanded my vision about how I could go with a professional relationship and develop my personal skill in terms of management, managing myself and managing a team. So that would probably be the book I want to share. It would definitely be helpful, not specifically a data book or modelling book, but at the same time, if that is the one that impacted your career the most, then definitely our listeners

31 should consider picking it up. "The Effective Executive" by Peter Drucker. Check it out, sounds like a great book. Haven't read it myself, but that's another one that's going on my to-do list or to-read list. Thank you very much, Xinran, for coming on this show. We appreciate you, we appreciate you sharing your expertise and knowledge about building models. It was great having you on the show. No problem, thank you for the opportunity. Alright, have a great day. See you next time. Bye. See you. So there you have it. That was Xinran Liu, ex-deloitte and now a senior corporate financial analyst at Jemena. And we had a great chat, it was good to reminisce about the good old days, about the past, and also it was great to understand how financial modellers and data scientists can work together, and what are the overlaps between the two professions, what are the differences, and what are the similarities. What are the skills that you can invest your time into so that you can be proficient in both sides. Personally, I picked up quite a lot of value from this conversation, for instance, sensitivity analysis was one of my favourite parts, of course closing billion dollar deals, I didn't even know about that before, so I was a bit surprised and very impressed. And another valuable thing I picked up personally was the reporting of different outcomes as scenarios. So instead of reporting a distribution of outcomes, why not report it as scenarios to the executives, it's easier to understand, it's easier to get your head around

32 scenarios and say listen, this is the worst case scenario, expected scenario, and the best case scenario. So I think that's a valuable tip. And when you think about it, it's pretty much an essential thing to do to not over-complicate your findings and reports. Instead of reporting a distribution, why not report some scenarios. And definitely, of course, I'm interested in learning a bit more about the tornado charts and book that Xinran recommended. So hope you enjoyed this podcast as much as I did. You can find the show notes at There you'll find all of the links that we talked about, recommendations, you'll find how to follow Xinran. And while you're there, don't forget to subscribe to this show. It's now available on itunes, so you can just go ahead and subscribe on itunes, and you will always get these fresh updates, fresh new episodes, as they are released. I can't wait to see you next time. Until then, happy analysing.