I was surprised when some consultants came to a meeting holding Starbucks coffee cups in their hands.

In my first year in Singapore, we had a business meeting with some consultants. My company was the client, and they were from professional firm working for us. It was our first meeting, and we hadn’t gotten to know each other yet. But when they arrived at our office, they came in holding paper cups of Starbucks coffee. I was quite surprised—because this would never happen in Japan.

Why is that? Actually, it’s a good question. When I think about it carefully, it doesn’t really make sense why it would be considered inappropriate in Japan. Maybe it’s because it could be seen as disrespectful or too casual—not formal enough for a business setting. But more likely, it’s simply because no one does it. And if you do something no one else does, it looks odd. That’s all.

Today’s topic overlaps slightly with #2: Following Rules and Order, but It Also Brings Pressure, but I’d like to focus on a few specific customs and unspoken rules in Japanese business culture that might be interesting to know.

Going back to the topic—what happens if we’re thirsty? Don’t worry, the host of the meeting will typically serve coffee or tea. Guests are not expected to bring their “own” drinks or snacks.

In fact, eating something during a meeting would also seem odd or even rude—unless it’s clearly set as a lunch meeting or the host provides snacks and explicitly says, “Please help yourself.”

Ah, one more thing—regarding the food (assuming you’re offered something) or even tea that’s served to you: don’t start eating or drinking before the host does. That’s a safer approach, especially if your host is from a very traditional Japanese company.

At the beginning of the meeting (especially the first one), start by exchanging business cards.

It’s very common in Japan to begin the meeting with business card exchange when meeting for the first time.

Let’s say your team has 5 people, and the other team also has 5. You are the most junior person on your side. As the exchange starts, you notice that one person on their side is just standing there, waiting and available. So you walk up to him to offer your card.

Ah… wait—what’s his seniority?

Of course, you don’t know until you receive his business card or hear an introduction. But try to guess based on age, attitude, body language, or any other clues you can pick up.

Why?

Business card exchanges are normally done in order of seniority, from highest to lowest. So the person you were about to approach might actually be their second most senior person, waiting for your boss (your team’s most senior person), who is still exchanging cards with their number one.

You’re not supposed to skip your more senior teammates and jump ahead to someone senior on their side. That would be considered rude or inappropriate.

Wow, it sounds complicated?

Yes, it is. But once you get used to it, it’s not so difficult.

Visiting your client’s office now and where you take your seat?

Stepping back a bit... Now you’ve arrived at your client’s office. The receptionist didn’t accompany you to the meeting room or tell you where to sit—so you have to choose a seat yourself. The room looks like the one below. (AI generated a slightly old-fashioned room for some reason—lol.)

Question:

You are the most junior among 3 of you. Where your boss will take a seat and where is your place? More importantly, which side you take? Left or right?

Answers:

(Intentionally adding some blank space below so it won’t be seen accidentally. Please scroll down when you’re ready.)

Basically, the seat closest to the entrance of the room is considered the most junior position, while the furthest back is the most senior. However, in a 3 vs. 3 meeting setup, the most senior person on each side usually sits in the middle. Therefore, the highest-priority seats are the middle chairs on both sides.

The host—meaning the company arranged the meeting room—should sit on the side with lower priority, which is the left side (closer to the entrance). This applies regardless of the business relationship. For example, even if the host is the client and the guest is a vendor trying to sell something, the host still sits on the lower-priority side. Although the client typically holds greater purchasing power in such a relationship, this does not affect the seating arrangement.

One advanced consideration is the window. In this image, the outside view is not particularly interesting, so it doesn’t influence the seating order. However, if there is a beautiful landscape view, the host should seat the guests on the side that allows them to enjoy the view, rather than taking that view for themselves. (so the side will be swapped in this case from the “base” case)

I could go on and on, but let me stop here for today.

What do you think of these cultural practices? Were you already familiar with them?

Actually, at the end of the day, if you’re not Japanese, people generally won’t expect you to know all of this or force you to follow it. Japanese people are usually quite tolerant of cultural ignorance, so there’s no need to worry too much.

But if you want to be successful in business with Japanese counterparts, stand out by knowing the culture!

What do you think about all these points I’ve shared today?

Feel free to share your experiences and thoughts in the comments section.

Subscribe now

Share

Regards,

Toyo Ushioda,

Founder of World Finance Edu

Me as a kid: “Which country did you study English in, Ms. A (English teacher)?”

Ms. A: “Well, I’ve never been outside Japan. I wish I could.”

I’m 45 years old now, so going back to my childhood may not seem relevant these days. But I still believe it holds some relevance. Many people of my generation—Japanese professionals working at the forefront of business—grew up in these circumstances and were shaped by them.

Back in the 1990s, for example, English education typically began in junior high school, around ages 13 to 15. Lessons were held a couple of times per week and focused primarily on written grammar. While there were some elements of pronunciation practice—such as learning through songs—the overall approach was quite academic.

As mentioned above, most of the teachers were domestically educated, and based on my personal experience, some had never even been abroad.

Starting in 2002, changes in the education system introduced mandates for English learning at earlier stages, such as in elementary school. However, I personally haven’t heard of any significant improvement in the overall outcomes—though I’d be happy to be corrected.

I’m sure there have been many positive changes compared to when I was a student. Still, why does it seem that English proficiency remains limited for many people in Japan?

From what I understand, the key challenges lie in the structure of the language, pronunciation, and a perfectionist mindset.

a) Structure of Language challenge:

In English, as grammar or more accurately to say the structure of language, a conclusion comes first and add conditional sentences.

For example, someone asks, “What did you do yesterday?” Then you might answer:

“I went to see a doctor as I was not feeling well, although originally there was a plan to go to a museum that had newly opened yesterday with my friend.”

(Intentionally, I’m adding many conditional details—FYI.)

But in the same context, the order of the sentence in Japanese would be something like this:

“Yesterday, with my friend, to a newly opened museum, there was a plan to go. But I was not feeling well. So, to see a doctor, I went.”

In this example, the difference may not be very noticeable. But the plan to go to the museum is not directly related to the answer to the question “What did you do?” In English, you usually hear the conclusion first, followed by additional information. But in Japanese, you often have to wait until the end of the sentence to understand what actually happened.

Here’s another example.

This is based on Japanese conversational structure:

“Growing up surrounded by relatives who ran music instrument stores, and with my father doing music sound engineering as a side job in addition to his full-time desk job, I was lucky to have plenty of opportunities to see, touch, and play instruments as well as sound machines. I was quite good at singing and playing the guitar, and even composed my own songs and won some prizes in competitions when I was in college. Then, after graduating, I also spent a lot of time and effort trying to become a professional musician—but I couldn’t make it.”

Actually, this is my real story. While you’re reading it, you might expect that this person became a professional musician. But only in the last sentence, you did find out that it wasn’t the case.

Maybe this type of storytelling—even in English—makes sense to structure this way. But my point is that in Japanese conversation, you need to listen carefully until the very end to know what actually happened.

This type of fundamental structural difference presents challenges when learning English, especially in speaking and conversational contexts. And I would say—culture affects language, and language affects culture.

b) Pronunciation challenge:

To keep this section brief—as it’s not strictly a cultural aspect—in my own definition, there are two distinct challenges in speaking English.

1) Japanese does not have the “R” sound (and some others like “TH” and “V,” but “R” is the biggest challenge). There isn’t a clear distinction between “R” and “L” in Japanese, so words like rice and lice are pronounced the same. For those who studied abroad—in the US, UK, Australia, or elsewhere—some of us have confidently mastered the “R” sound. However, sometimes it ends up being over-pronounced, especially in American English with its darker “R” sound. It can sound awkward, as it still touches parts of the mouth in a way that feels unnatural.

2) Japanese words almost always end in vowel sounds. We don’t typically end words with consonant sounds. For example, a book becomes booku, books becomes bookusu, and accent becomes acucento—if I try to mock the sounds in English spelling.

These are not really issues in a global environment where everyone has their own accent. But they relate to the final point: the perfectionist mindset, which I’ll explain next.

c) Perfectionist mindset challenge:

This is the most challenging aspect and the one most closely related to culture, and it connects with the two points mentioned above.

When Japanese people try to speak English, one thing that often comes to mind is: What if I make a mistake in grammar, structure, or pronunciation? This thought can cause us to hesitate or get stuck, unable to say what we want to say next. As a result, it may give the impression that we are not confident, being quiet, or lacking opinions.

Why does this happen? I believe it’s somehow related to the cultural principle of “Don’t disturb others.”, which I wrote in the 1st article in this newsletter. If I make a mistake in English, will it cause trouble or disturb the person I’m speaking to or the flow of the meeting? Will my poor English negatively affect the progress of the discussion? This kind of mindset is a serious challenge for us—more so than grammar, structural differences, or pronunciation, which in reality don’t matter as much.

It’s a kind of perfectionist mindset, and in the worst cases, it can lead to peer pressure, where we pressure each other to be perfect. (Related to the 2nd article in newsletter.)

Note that I made this intentionally as “stereo type image”

---Now that today’s writing has grown longer, I’d like to wrap it up here. Going back to the question I raised in the title—#4: Why Do Japanese People Find challenges in English ? Does Culture Matter? My answer is yes, I believe culture matters a lot.

Over time, we’ve seen progress, with more people gaining experience in different cultures, studying abroad, and working in multinational companies. I hope we continue moving in the right direction. But at the same time, we should, of course, respect and preserve our own culture.

What do you think about all these points I’ve shared today?

Feel free to share your experiences and thoughts in the comments section.

Toyo Ushioda

Founder of World Finance Edu

Subscribe now

Showing respect to others requires changing how you speak to them

In Japanese culture and language, there are several different levels of politeness in speaking styles and even distinct sets of vocabulary, known as Keigo (敬語). I won’t go too much into detail here, but will focus on helping you understand the concept.

When you watch or listen to two people speaking in Japanese as an example, you’ll often notice that one person is using casual language while the other speaks in a much more formal way. In many cases, the person using casual way is older or holds a more senior position in a business context—such as when both people work in the same company.

This custom of adjusting how we speak based on relationships and social hierarchy starts when we are still children. Among family members, it’s acceptable to use casual language, even when speaking to your parents, who are obviously older and “senior” in a sense. However, when speaking to someone outside your family, you’re taught to use a more polite or formal way of speaking—especially with older people or those you don’t know well, regardless of their actual age.

If you speak casually in an inappropriate situation, it may come across as rude or overly assertive, as if you don’t care about social norms. You could even be seen as a strange outlier.

Yes, well... it may sound difficult or cumbersome — and I would agree with that.

But somehow, we are so well trained to do this that it usually isn’t very difficult to adopt the style.

That said, some people do find it hard to fit in, which can create challenges when living in this culture and society.

“How old are you??” The reason why Japanese people tend to ask this question

You may notice that Japanese people often ask about age. However, it’s not considered very polite to ask someone’s age directly. So instead, we tend to ask questions like “Where are you from?”, “Where did you grow up?”, or “What generation are you in?”—indirectly guiding the conversation toward figuring out whether the person is older or younger than us.

Let’s have a look at an example. (I created this myself, but it sounds like a very real conversation.) Note that this is written in English as a translation to help you understand.

A and B are meeting for the first time. Both start by speaking in a formal level of Japanese. Let’s also assume that there is no business relationship between them—otherwise, the business context would define the hierarchy, and both would likely continue speaking formally throughout the conversation.

Toyo: Nice to meet you. I’m Toyo.

Masa: Nice to meet you. I’m Masa.

Toyo: Kenji told me about you, and I wanted to meet. Thank you for making the time today. Was it a hassle to get to this area?

Masa: Not at all. I live nearby—it only took me 10 minutes to get here.

Toyo: Ah, I see. Did you grow up around here as well?

Masa: Yes, I’ve never really left this town.

Toyo: Really? Same here. I went to AA (name) Elementary School.

Masa: I see. I went to BB (name) Elementary School. It’s actually not far from yours. I know someone from AA—her name is Yuki.

Toyo: Yuki? Hmm, not sure... maybe I don’t know her. But I guess we’re from the same generation, right? I used to watch Dragon Ball anime when I was a kid.

Masa: Oh, I really liked that too back then. I was wondering.. maybe you’re older than me? I was born in 1978.

Toyo: Oh, close. I was born in 1980. So I’m actually two years younger than you.

Now… gradually, Masa may switch to a more casual way of speaking. Meanwhile, Toyo (as the younger person) would likely continue using formal language (Keigo).

Next time they meet, it’s very likely that Toyo will still speak formally, while Masa will use casual language.

Sounds funny? They’re only two years apart in age, which doesn’t really make much difference in terms of experience, seniority, or maturity in life. But this kind of situation happens a lot in Japan.

In a conversation between two people, it’s common for one to speak casually while the other speaks formally. Alternatively, both may use formal language—especially in a business context. A situation where both use casual language usually only happens when they are exactly the same age (called “tame” in Japanese), which is relatively rare if think about what’s the possibility—or when they’re very close friends and don’t care about the age difference (or among family members).

Also, What I find especially fascinating is how this works in conversations with more than two people—something that almost everyone in Japan can do quite naturally.

Let’s consider a group with three people: Person A, B, and C.

A is older than B.

B is older than C.

For A, it’s simple—A can speak casually to both B and C.

For C, it’s also straightforward—C uses formal language when speaking to both A and B.

But for B, things get interesting. B has to switch the way of speaking depending on whether they’re talking to A or to C.

And amazingly, we can do this almost without thinking—it happens almost automatically.

Going back to the title of this article—why do Japanese people tend to ask about age, either directly or indirectly?

It’s mainly because we want to figure out whether we should use formal or casual language when speaking to the other person. If we’re unsure, it can feel uncomfortable. Actually, if you are clearly older but continue using formal language, the other person might feel even awkward. That’s what makes it a bit complicated.

Identifying the relationship early on helps set the tone for the conversation—it acts like an unspoken “rule” that makes both sides feel more comfortable.

What do you think about this kind of culture? Does it sound funny or interesting to you?

Do you have something similar in your culture?
Feel free to share in the comments!

Toyo Ushioda
Founder of World Finance Edu

Subscribe now

<Skip this introduction section if you’ve been following the series.>

I’m Toyo Ushioda, currently based in Singapore (as of April 2026), and have worked for several companies with a global business footprint, where I encountered many cultural differences. I also attended the Executive MBA program at INSEAD from 2021 to 2023, studying alongside classmates from a wide range of backgrounds.

Through these experiences, I’ve come to reappreciate how unique Japanese culture is compared to others. (Of course, all cultures are unique and different — I’m not saying one is better or worse, just different in fascinating ways.)

With this weekly article series, I’d like to introduce some fundamental aspects of Japanese culture that shape how people think and behave. If you have Japanese friends or business partners — whether clients, suppliers, or colleagues — I hope this will help you better understand their perspectives and strengthen your relationship.

Also, in this newsletter series, I aim to keep each article brief and easy to read — not a long thesis. The interpretations I share are based on my own perspective, not academic research or historical analysis. However, I believe many Japanese people would resonate with these views.

Following Rules and Order

You may be surprised that many Japanese people follow rules and order without complaining. For example, most employees in Japanese offices arrive on time—or even 30 minutes early. It was very rare (at least before more flexible time arrangements were introduced after COVID-19) to see people arriving late to the office.

Most people commute by train or subway, as driving is not a viable option in Tokyo. In cases where trains are delayed (which happens less frequently compared to some other countries due to highly punctual operations), train staff issue official delay certificates to commuters. These certificates can be shown to their employers to justify being late.

Continuing with the example of commuting behavior, Japanese people tend to strictly follow the order of lining up at the train station staircase. They typically keep one side clear for people coming down after the train arrives and use only the other side to ascend to the platform. This allows the flow of people to remain smooth, without getting stuck in the middle.

The same applies inside the train. If you’re standing near the door and the train is crowded, you’re expected to step off temporarily at the next station—even if it’s not your stop—to make room for those who are actually getting off.

This kind of behavior doesn’t usually happen in Singapore, which often frustrates me, even though I know it’s not something I should expect. Normally, I’m a pretty patient person (you know if you know me personally), but this really upsets me.

What’s Behind These Behaviors

If you read the first article in this series, you may have noticed that these behaviors are based on a core principle: “Do not disturb others, and respect them.”

I believe the reason why we Japanese tend to follow rules and maintain order so well is because this principle lies at the heart of our collective mindset. It also reflects our own desire not to be disturbed by others. (Yes.. that is why I mentioned that I tend to be upset on the train in Singapore as example.)

This last point is actually very important, as it reveals another side of Japanese culture—peer pressure.

Peer Pressure

Because we don’t want to be disturbed by others—and we expect others to behave the same—this increases the social expectation for everyone to act in certain ways. It mentally enforces people to follow rules and maintain order. Those who do not conform stand out and are often labeled as “outliers,” “strange,” or something similar.

This can be particularly tough for those who look different, think differently, or come from a different cultural background.

The reason you may not have felt “strange” or treated differently while in Japan (assuming you are not Japanese or didn’t grow up in Japanese culture) is likely because you were seen as a guest—a tourist. Even if you’ve lived in Japan for many years, you may still be treated as an “exception,” for better or worse. (To be clear, I have no intention of debating this point here.)

What I want to highlight is that there is significant peer pressure among Japanese people. Because everyone behaves in a certain way, there’s a strong social expectation to do the same. This often has a tendency to limit individuality, creativity, or uniqueness in personality and expression.

Living on Both Sides of the Coin

My message here isn’t necessarily negative or positive. I simply want to point out that there are two sides to the coin. We Japanese are very good at following rules and maintaining order—but that strength can come at the cost of pressuring others to conform.

What about your culture? Do you see similar “two sides of the coin” in certain aspects?

Feel free to share your thoughts in the comments.

Subscribe now

Share

Regards,

Toyo Ushioda,

Founder of World Finance Edu

I’m Toyo Ushioda, currently based in Singapore (as of March 2026), and have worked for several companies with a global business footprint, where I encountered many cultural differences. I also attended the Executive MBA program at INSEAD from 2021 to 2023, studying alongside classmates from a wide range of backgrounds.

Through these experiences, I’ve come to reappreciate how unique Japanese culture is compared to others. (Of course, all cultures are unique and different — I’m not saying one is better or worse, just different in fascinating ways.)

With this weekly article series, I’d like to introduce some fundamental aspects of Japanese culture that shape how people think and behave. If you have Japanese friends or business partners — whether clients, suppliers, or colleagues — I hope this will help you better understand their perspectives and strengthen your relationship.

Also, in this newsletter series, I aim to keep each article brief and easy to read — not a long thesis. The interpretations I share are based on my own perspective, not academic research or historical analysis. However, I believe many Japanese people would resonate with these views.

Today is the 1st topic:

1. Respecting others is a core principle of all

Back in our childhood, most Japanese were repeatedly told, “Do not disturb others and respect them,” whether by parents, kindergarten teachers, or educators in elementary school and beyond. Literally — they said those exact words to us.

As a result, certain behaviors tend to form as we grow up in Japanese society. For example:

We do not directly say “No” to others but imply it softly. Why? Because we do not want to disturb others’ feelings by making them feel rejected or opposed.

We do not interrupt when others are speaking, but wait until the current speaker finishes. Then, the next person will take over. Interestingly, there are even soft cues or words people use to show they want to speak next — instead of jumping in. Why? Because someone else may also wish to speak, and we want to avoid overriding them.

We do not skip the queue in lines (this might be common in many countries too). Why? Because we do not want to disturb others by cutting in or appearing disrespectful.

We “do” care about the quality of our work — for examples, whether it’s food, presentation, or packaging like how we place food on a dish.. Why? Because we don’t want others to feel they are being treated carelessly or without respect.

We use different forms of language when speaking to clients, elders, or seniors at work. This respectful language is called Keigo. Why? Because we do not want them to feel they are being treated without proper respect.

I could go on — and many of these will be explored in future topics of this weekly newsletter.

But the key takeaway for today is this:

“Do not disturb others, and respect them” is a deeply rooted principle that underlies many aspects of Japanese culture.

(Again, this is just my personal interpretation — not academic or historical theory.)

Please subscribe to receive notifications, and feel free to comment or share your own cultural experiences.

Subscribe now

Share

Regards,

Toyo Ushioda,

Founder of World Finance Edu

It sounds simple. But anyone who has built one from scratch knows what "quick" actually means. You open Excel. You start a new sheet. Then you spend the first 30 minutes just thinking about structure: what tabs to use, how to handle the debt schedule, where to put the assumptions, how to link the exit valuation to the cash flow section. You haven't typed a single number yet.

Then the inputs start to feel thin. You have an acquisition price and some revenue projections, but the WACC is missing. The tax rate is missing. You're not sure whether to use straight-line debt repayment or a custom schedule. You make a call on each one, note it somewhere so you can defend it later, and keep building.

Two hours later, you have a working model. Maybe two and a half. And now you need to check it.

I wanted to see whether Claude in Excel could compress that process. Not replace the thinking, but handle the basic step so the analyst can focus on the judgment calls instead of the construction work.

Here is what happened when I tried it.

The Starting Point

Before typing a single prompt, I filled in a sheet called "Initial Info" with the deal parameters I had on hand.

The company was WFE Tech, a US corporation. The deal size was $5,000,000. I shared some info such as revenue figure of $1.2M, EBITDA of $500K (41.7% margin), and both entry and exit EBITDA multiples of 10.0x. I knew 40% of the deal would be debt-financed at 4% interest, and revenue was expected to grow at 4% per year, assuming we have only this of info at this moment.

One field I left blank was WACC. I put "?" intentionally. I wanted to see what Claude would come up with on its own. This was a test: if I give Claude an incomplete input, would it make a reasonable call and keep going, or would it stop and ask?

The First Prompt

I typed one sentence into the Claude panel inside Excel:

"can you make IRR model based on this info?"

No structure. No instructions on what tabs to use. No guidance on assumptions. Just that one line.

Claude Read the Sheet and Made a Plan

Claude did not start building immediately.

Instead, it read the spreadsheet and laid out a 5-step plan for what it intended to build:

1. Sources and Uses (how the $5M acquisition would be split between equity and debt)

2. Projected Income Statement (revenue growing at 4%, with debt service factored in)

3. Exit Valuation (Year 5 EBITDA multiplied by the exit multiple, minus remaining debt)

4. Cash Flow to Equity and IRR (Year 0 equity outflow, annual free cash flows, Year 5 exit proceeds, then the IRR formula)

5. Formatting (clean layout with finance conventions)

At the end of the plan, it asked one question: "Shall I proceed, or would you like to adjust anything (e.g., debt repayment schedule, cost of equity assumptions for WACC, or any other details)?"

This step matters. Claude did not assume it understood correctly and start building. It confirmed its interpretation first and gave me a chance to correct anything before any work was done.

One Response: Just Proceed

I replied:

"Please proceed. I do not have info more than this. If there are missing info, please guess and compliment based on your knowledge."

That was the key moment. I handed over the gap-filling entirely to Claude and trusted it to use reasonable assumptions for anything I had not provided.

What Claude Built

Claude put everything on one sheet. Not three tabs, not a separate assumptions page. One clean sheet with a logical flow from top to bottom.

The model had four main sections:

Sources and Uses. Equity $3,000,000 and Debt $2,000,000, totaling the $5,000,000 acquisition price. Straightforward.

Projected Income Statement for Years 1 to 5. Revenue starting at $1.2M and growing 4% per year. EBITDA margin held constant at 41.7%. Then D&A, EBIT, interest expense on the declining debt balance, taxes, and net income.

Debt Schedule. $2M repaid in equal annual installments of $400K, fully paid off by Year 5. Interest expense declining each year as the balance reduced.

Free Cash Flow to Equity (FCFE). Net Income plus D&A, minus CapEx, minus debt repayment. This is the actual cash available to the equity holder each year.

What stood out was how Claude handled the missing inputs. It filled them in on its own and highlighted them in blue so I could see exactly which numbers were estimates:

• Tax Rate: 21% (US corporate tax rate)

• D&A: 3.0% of Revenue (typical for tech companies)

• CapEx: 3.0% of Revenue (maintenance CapEx set equal to D&A)

• Cost of Equity: 12.0% (typical mid-market PE target return)

• WACC (computed): 8.5%

Every cell in the model was formula-driven. Nothing was hard-coded in the output area.

The Results

IRR: 13.1%
NPV at WACC (8.5%): $717,409
MOIC: 1.9x

To translate these simply:

IRR is the annualized return on the equity invested. At 13.1%, the model indicates the equity investment would earn 13.1% per year over the 5-year hold period. Since this exceeds the WACC of 8.5%, the model shows a positive return above the cost of capital.

NPV of $717K means that after discounting all cash flows at 8.5%, the deal generates $717K of value above what was required to justify the investment.

MOIC of 1.9x means the equity of $3M grows to roughly $5.9M at exit. Not quite 2x in five years.

One detail worth noting: FCFE is negative in Years 1 through 4. This is not an error. It correctly reflects that debt repayment ($400K per year) and taxes consume more cash than the business generates in the early years. The equity holder sees no cash until the exit. Year 5 is when everything resolves: the final FCFE plus the $5.85M exit equity value combine to $5,865,476, and that is what drives the IRR.

The model handled leverage correctly. Early years are cash-negative because of debt service. The return is back-loaded, concentrated in the exit. This is exactly how leveraged buyout math works.

Key Takeaways

1. Claude reads your spreadsheet directly. It picked up all the inputs from the "Initial Info" sheet without me having to re-enter or explain anything. The context was already there.

2. It confirmed its plan before building. One clarifying question, then it waited. This is better behavior than jumping into work based on a guess.

3. When told to proceed with missing info, it filled the gaps with reasonable assumptions. The 21% tax rate, 3% D&A and CapEx ratios, and 12% cost of equity are Claude's starting points. They are not presented as the correct numbers. They are a reasonable basis to begin with, and you should review each one against the actual deal.

4. It marked its own estimates clearly in blue. You always know which inputs are yours and which are Claude's. The ones in blue are the first things you should challenge.

5. Every cell is formula-driven. Change one input in the assumption area and the entire model recalculates. Revenue growth, margin, exit multiple, debt level. Any of them.

6. The model correctly handled the mechanics of leverage. Negative FCFE in Years 1 to 4. Exit proceeds dominating Year 5. This is not simplified math. It is a structurally correct FCFE-based IRR model.

The total time from entering the initial inputs to a working model was a few minutes. Of course, then you continue to work on to change or simulate the results.

What I Would Do Before Using This in a Real Presentation

The model gives you a solid starting structure, but there are a few things worth reviewing before taking it to a board or client:

• Confirm the debt repayment schedule matches the actual term sheet (straight-line repayment is an assumption here)

• Review the D&A and CapEx ratios for the specific company, they can vary widely within "tech"

• Check whether the EBITDA margin should stay flat or compress or expand over the forecast period

• Consider adding a sensitivity table on IRR vs. exit multiple and revenue growth

Claude built the skeleton. You still own the judgment.

This is Part 5 of the series "The New Age Financial Analysis with AI," published on Substack by World Finance Edu.

If you found this useful, follow along at toyoyukiushioda.substack.com for new articles in the series.

Share

Subscribe now

World Finance Edu helps finance professionals build practical AI skills for real-world financial analysis.

This one doesn’t.

This is for those who haven’t started yet. You see the discussion happening and you’re not sure where to begin. Simple task, step by step, real screenshots.

If this sounds familiar, you’re in the right place.

The Problem

I had a 5-year financial model for a company with 5 tabs: Assumptions, Revenue, Income Statement, Market Sizing, and Headcount.

After some restructuring (renaming tabs, reorganizing sheets), the model was full of #REF! errors. And the thing about #REF! errors is that they spread. One broken link in Revenue breaks Gross Profit, which then breaks EBITDA, which then breaks Net Income. Before you know it, half of your Income Statement shows #REF!.

You’ve seen this before. You know the pain.

The Income Statement was a wall of #REF!. Five columns, nearly every row.

The Revenue tab had broken regional breakdowns too:

And the Headcount tab had broken staff cost calculations:

In total: 5 distinct broken references spread across 3 tabs, turning into 20+ cells showing #REF!.

If you do this manually, you’d trace each formula, figure out which sheet got renamed, and update every reference one by one. That’s 15-20 minutes if you’re experienced. Longer if you’re not sure what the original references should have been.

What I Asked Claude

I selected the Income Statement and said in Claude in Excel:

“This financial model has multiple #REF! errors across several tabs. The errors are caused by sheet references pointing to old tab names that no longer exist. Please scan all tabs, identify every #REF! error, diagnose the cause, and fix them.”

One prompt. That’s it.

What Claude Did

Claude scanned all 5 tabs and came back with a diagnosis:

Error 1: Income Statement Revenue (row 5) referenced ‘Rev Detail’, but the actual sheet is named Revenue. Fixed all 5 cells.

Error 2: Income Statement COGS (row 7, FY2026-2028) referenced ‘Old Assumptions’. Should be Assumptions. Fixed 3 cells.

Error 3: Revenue regional breakdown (rows 12-13) referenced ‘Mkt Sizing’. Actual sheet is Market Sizing. Fixed 10 cells.

Error 4: Headcount staff cost (row 13) referenced ‘Staff Plan’. This was a self-referencing formula that should stay within the Headcount tab. Fixed 5 cells.

Error 5: Income Statement G&A (row 15) referenced ‘HC Plan’. Should reference Assumptions. Fixed 5 cells.

Total: 28 cells corrected across 3 tabs.

But here is what surprised me the most.

Claude didn’t just do a simple find-and-replace of sheet names. It actually understood the structure of the financial model. It knew that G&A costs should be linked to the Assumptions tab because they are budget-based fixed costs. It knew that the staff cost formula in the Headcount tab should reference its own sheet, not an external one. It understood that Revenue regional breakdown should pull market share data from the Market Sizing tab.

In other words, Claude understood what each line item means in a financial model and where the data should come from. That is not a simple text replacement. That is financial model comprehension.

The Result

Revenue flowing. COGS calculating. Gross margin at 38-45%. EBITDA positive from FY2025. Net income growing from $152K to $1.8M over the projection period.

Regional revenue breakdown fully reconciled to the total.

Headcount staff costs calculating correctly again.

All from a single prompt. No manual tracing. No formula bar detective work.

What I Took Away

1. Name your problem clearly. Telling Claude “there are #REF! errors from renamed sheets” is far more effective than just “fix the errors.”

2. Claude understands the model, not just the formulas. It doesn’t blindly swap sheet names. It figures out where each line item should pull data from based on what that line item actually means. G&A from Assumptions, staff cost from Headcount, regional revenue from Market Sizing. That level of understanding is what makes it useful.

3. Let Claude scan broadly. Don’t point it at one cell. Ask it to check the entire model. It can trace cross-tab dependencies faster than you can.

4. #REF! errors look scary but they follow a pattern. Once Claude identifies one renamed sheet, it can fix every reference to that sheet across all tabs.

5. The spreading works in reverse too. Fix 5 root causes, and 20+ downstream cells get resolved automatically.

6. Always verify. Check that the numbers make sense after the fix. Claude fixed the formulas, but you should confirm the output is reasonable.

Well, it is exactly like managing a junior analyst on their work as a manager.

World Finance Edu provides finance-related training and seminars for both corporate and individual learners, targeting junior- to mid-level professionals.

Its programs cover not only finance but also broader business topics such as AI, statistics, leadership, and communication.

Please subscribe to get notified when new articles come out!

Subscribe now

There’s no shortage of posts about how powerful Claude in Excel is. Most of them jump straight into impressive, complex use cases.

This one doesn’t.

This article is for those who haven’t started yet — who see the discussion happening around them and aren’t sure where to begin. My focus is simple tasks, shown step by step, with real screenshots of what actually happens.

If this sounds familiar, you’re in the right place.

The Setup

I’ve been building out a financial model for Novatech, a fictional company I use for teaching. The Income Statement lives on the FS tab, and all the raw expense transactions sit on a separate “Other expense” tab — a running journal with hundreds of line items covering everything from Corporate Cost to Depreciation of Computer Equipment.

Note that the first two columns cover full annual periods, and the last four break FY2026 into quarters. This kind of mixed template is common in practice — you’ll likely recognize it.

The challenge: getting those expenses to flow into the FS automatically, by period, by category — without hardcoding a single number.

The Ask

I selected the expense section of the FS and gave Claude a simple instruction:

“Can you work on the expense section I selected to link up from the other expense tab? You need to do it by formula base. Summarised account may not be enough if so, please add them. I am ok to insert more rows. But make sure it does not affect the sub total. Also, can you make a new tab to show reconciliation works to proof you have picked up correct numbers from the source tab?”

Ask Claude to work on a formula basis — otherwise it may come back with hard-coded figures. Also ask it to build a reconciliation tab upfront, so the first output is already verifiable. If you skip this, you’ll end up doing the reconciliation yourself.

What Claude Built

Claude constructed SUMIFS formulas that filtered the “Other expense” tab by both account category and date range, using the period start and end dates from the FS header row. The column references were made relative, so the same formula pattern worked across FY2024, FY2025, and all four quarters of 2026 without adjustment.

For the four summary categories (Corporate Cost, Facilities & Utilities, Technology & Subscriptions, Other Operating Cost), Claude used a broader “Summary Category” approach. For the depreciation and amortisation lines, it matched on exact account names.

The Reconciliation Tab

Claude also created a dedicated Expense Recon tab with three sections:

A. Source — what the “Other expense” tab actually contains, summed by category and period.

B. FS Pick-Up — what the FS formulas are pulling in, expressed as absolute values.

C. Difference — Section A minus Section B. All zeros. Fully reconciled across all 6 periods and all 9 expense categories.

A completeness check at the bottom confirmed that 100% of source debits were allocated — nothing left unaccounted for.

Then I Spotted Something

Looking at the grand total row on the reconciliation tab, the numbers didn’t quite add up at first glance. The “Other expense” tab had a total net debit of $2,131,580.30 across all dates. But the sum of what the FS had picked up across all periods was only $1,794,289.54. A gap of $337,290.76.

I flagged it to Claude: “I have done a high level check. Notice that the total of expenses in the Other expense tab does not match to the total of expense in FS for the period. Why is this happening?”

Then Claude comes back with analysis.

The Root Cause — And Why It Was Actually Fine

Claude’s diagnosis was clean. The mismatch was not a formula error. The “Other expense” tab contained transactions running to 2026-12-29, but the FS only had period columns through 2026-03-31. There were 281 transactions totalling $337,290.76 sitting in the source tab with no FS column to land in — future-dated data in a model that hadn’t been extended yet.

Claude added a Section E (Grand Total Reconciliation) to surface this explicitly:

• Total net expense (all dates): $2,131,580.30

• Total picked up by FS columns: $1,794,289.54

• Transactions outside FS date range: $337,290.76 (Not yet in FS)

• Unreconciled difference: $0.00 ✓ Fully Reconciled

What This Demonstrates

A few things stood out from this exercise:

Instructions matter. Be specific about what you want — formula-based, not hard-coded. Vague prompts get vague outputs.

Ask for reconciliation upfront. Claude can build the proof alongside the work. If you don’t ask, you’ll be the one doing it manually after.

Reconciled doesn’t mean done. Even with a clean reconciliation, a high-level human review is still needed. A number that looks off might be a real error — or it might have a perfectly good explanation that isn’t visible at first glance. You won’t know unless you look.

Claude can help you investigate. Once you spot something, point it at the issue and ask why. It will work through the logic with you.

Well, it is exactly like managing a junior analyst on their work as a manager.

Closing Thought

If you’re building financial models in Excel and haven’t tried Claude in Excel yet, this kind of formula construction + reconciliation workflow is a good starting point. The combination of “build it” and “prove it” in a single session is genuinely useful.

World Finance Edu publishes practical financial modelling content for finance professionals and learners. If you found this useful, consider subscribing.

Toyo Ushioda, Founder of World Finance Edu

Subscribe now

The typical rule of color coding is as follows:

• Blue: Hard-coded (manually entered) figures

• Black: Formula-driven cells

• Green: Linked from a different tab within the same file

• Purple: Linked from an external file

This rule helps significantly when analyzing existing models as well as managing your own models.

However, the point of this article is not the color-coding convention itself.

Instead, the question is: Can we automate this color-coding process using Claude?
The answer is of course, YES.

Before testing that, it is helpful to understand the existing solutions available for this type of coding.

One well-known solution is Macabacus, an Excel add-in widely used by finance professionals. It includes this functionality. With a preset shortcut such as Ctrl + Alt + A, you can instantly apply the color-coding rules to a model.

Of course, Macabacus requires a paid subscription, although it offers many other powerful functions beyond this feature.

But what about Claude integrated with Excel? Claude also offers an add-in that can be used with a Pro account.

Can Claude replicate the same functionality?

Subscribe now

Let’s test that here.

1) Make sure you have Claude in Excel installed.

2) Select the sheet you would like to test

Actually, the selection of the sheet itself has not yet been done.

This means that in the chat instructions to Claude, we need to specify which sheet or which cells we want the changes to be applied to.

But we need to “look” at which sheet we work on.. so select it first.

Next, give an instruction to Claude. You can copy and Paste this but actually it does not really matter. Only point is to specify the rule.

«
Can you change font colors in figures based on the rule below?

Blue: Hard-coded (manually entered) figures
Black: Formula-driven cells
Green: Linked from a different tab within the same file
Purple: Linked from an external file
»

After about one minute or so—depending on the size of the sheet (and the file)—it may come back with this. In fact, sometimes Claude asks you, “Which sheet do you want to apply this rule to?” This time, for some reason, it did not ask me to select one.

Click “Allow Once” for now.

Then the result will appear.

However, there are some issues here. The summary line, such as TOTAL ASSETS, is shown in black font, which is correct based on the rule. However, due to the cell color design, the numbers are now hard to see—for example, black font on a dark blue background.

In this case, we need to give another instruction and rule to Claude.

After proceeding this, it will be perfectly completed. And You can ask Claude to apply the same rule to rest of the sheets.

But here is a big question as follow up.
Technically if you close the Excel file and open it again with another model either even the same model, Claude does not remember what it learned this time.
Claude comes back like this.

In this case, the best solution is to use Skills in Claude.

When you open Claude and go to the Customize section, navigate to Skills and follow the instructions. Basically, you can define what rules you want applied whenever you type the name of that Skill.

For me, it will be like this below. The name is excel-autoformat.

Apply Macabacus-style AutoColor font coding and financial number formatting to Excel cells via Claude in Excel. Use this skill whenever the user asks to: format numbers in Excel, apply AutoColor, color-code cells by type, add thousand separators, show negatives in parentheses, apply finance/financial modeling formatting, replicate Macabacus formatting, or says anything like "auto font color", "color code my cells", "format my numbers", "apply thousand separator", "negative in brackets/parentheses", or "WFE formatting". Always trigger this skill for any Excel formatting request involving color coding or number formatting.

In the Claude in Excel window, type the name of the skill you created. Mine is excel-autoformat — a simple name. Next time you type this, Claude in Excel will work according to your pre-set rules.

This article ends here.

Please subscribe for the next one, as I am planning to make this a series focused on AI and the practical considerations for the future of finance work.

Toyoyuki Ushioda
Founder of World Finance Edu

Subscribe now

Author: Toyoyuki Ushioda

With 20+ years in financial and operational management, he has held senior leadership roles including Global Head of FP&A at a US$125B asset management firm (Real Estate/PE/VC).

Building on this experience, he founded World Finance Edu and is also an Angel Investor, VC Limited Partner, Venture Partner, and Financial Advisor.

He is a Certified Public Accountant (Washington State) and holds an INSEAD Global Executive MBA as well as certificate in business excellence at Columbia Business School’s Private Equity & Venture Capital Program.

AI progress has been drastic. Just one year ago, AI was still relatively weak when it came to number-driven, finance-heavy tasks.

However, with recent updates and progress—particularly with tools like Claude—capabilities that were not ready even a few months ago are now practically usable.

This article is part of a series in which I will share, step by step, how to use AI for:

• Creating financial statements

• Performing financial analysis

• Building projections

• Conducting valuations

• Other core finance-related work

Note that there are many advanced ways to use AI, such as integrating Claude in Excel that allows you to work on within Excel itself or Claude Code that can automate the process by directly accessing the file in folders. I will cover these approaches later.

For now, I am starting with the basics to ensure the series is easy to follow, regardless of prior experience with AI.

We will still use Excel as the final deliverable.

But AI can now function much like a junior financial analyst, handling the basic preparation work for you.

This is the key point.

This “junior analyst” is extremely talented and highly disciplined—it never complains about workload.

(Although, to be fair, even Claude’s Pro plan has usage limits… it gets tired at the current pay grade—very human-like behavior, haha.)

That said, AI still requires:

• Clear guidance

• Careful review

• Explicit feedback on what went wrong and why

We need to point out incorrect outputs and explain how to avoid similar mistakes next time.

In this series, I will show:

1. What actually happened in real workflows

2. How I reviewed AI-generated work

3. How I instructed AI to “grow” its skills as a financial analyst

How to Create an Income Statement and Balance Sheet from Journal Entry Data:

Creating financial statements is the most basic task in financial analysis. Let’s start with this area and see how AI can help with the work.

First, we are using Claude’s Pro plan.

(At this moment, based on my own trials, ChatGPT’s Pro plan was not able to do the same task.)

Below is a set of journal entry data (dummy data).
(In fact, this random data was created using AI. The company and business are assumed to be a real estate investment, but this does not affect any of the arguments in this article.)

*Above is not showing all data but only snap shot

Let’s try to ask in rough instruction first

Attaching this excel data and I asked:

”Make Trial Balance and move on to create Income statement and balance sheets. Forget about CF for now.”

It comes back by this outcome.

Trial Balance:
First, let’s check on Trial Balance.

There is an obvious red flag in the output.

If you look at the last line of the table, it says “Balance Check”, followed by numerical values and it says “should be 0” This does not make sense.

Presumably, the AI attempted to perform a balance check to ensure that the total debits and credits equal zero.

It is actually impressive that the AI tried to conduct this balance check on its own, as I did not instruct it to do so.

However, despite the correct intention, the result of the test is incorrect.

What was wrong?

You need to carefully review the output and identify where the AI made mistakes.

That is the better approach, as it allows you to properly assess and become more comfortable with the AI’s work.

You may also ask the AI to identify the mistake and correct it by itself.(I am confident that AI can resolve it independently.)

In this case, the mistake was as follows:

In the balance sheet section, retained earnings were already included with their corresponding figures.

However, the table then continued with income statement items.

Since income statement items include revenues and expenses, profit is calculated as the difference between the debit and credit sides. Naturally, this creates a net difference. That net income should then be transferred into retained earnings — that is how accounting works. (Most readers will understand this.)

The issue is that retained earnings had already been presented in the balance sheet.

The current year’s income statement net income should be closed into the current year’s retained earnings as part of the closing process.

After explaining the error to the AI and instructing it to be careful next time, it was able to avoid making the same mistake.

(I have tested the result and it was correct.)

Balance Sheet :

More interestingly, similar to the trial balance, the balance check does not work.

In the first year, the 600k appears to represent a declared dividend, as the same amount is used.

However, the trial balance does not explicitly state that a dividend was declared, which is another point worth noting.

To avoid making this article too long, I will jump into the main issue.

The mistake relates to accumulated depreciation. Instead of deducting accumulated depreciation to arrive at net investment property, it was added to the amount. In other words, the formula was incorrect.

If we correct the summation formula in row 39, the issue is resolved.

By the way, if the accounting principle is IFRS, investment property should not be depreciated. However, let’s not focus on that for now—this is an issue with the random data itself rather than with the task of creating financial statements.

FY2024 has still imbalanced numbers:

To jump straight to the answer, the opening balance of retained earnings in 2024, shown as –1.2m, has the wrong sign.

It appears as a negative figure because the formula refers directly to retained earnings in the trial balance, which is stated as a positive amount. For some reason, however, the AI added a minus sign in front of it. The opening retained earnings balance should therefore be positive, not negative.

After I shared my feedback and explained how the AI made this mistake, I ran another round using a different set of randomly generated journal entries. In that next round, the AI did not make the same mistake again—it had improved.


This is how AI worked and made mistakes and why.. and how to correct the next work.

Is it like a real junior analyst? In many ways, yes.

AI can make similar mistakes, and after receiving feedback, it can be more careful the next time. However, there is an important difference. If we restart the conversation—unless it is within the same project that retains historical context—the AI may make exactly the same mistake again, or even a different type of mistake.

I have tested this logic directly, and the AI itself acknowledges this limitation.

Unless the AI has genuinely learned and internalized the underlying financial statement logic, there will always be a risk of errors. This is something we should all be aware of when using AI as a tool.

So, your knowledge of finance—and how financial statements connect and work together—remains extremely valuable at this moment.

This article ends here.

Please subscribe for the next one, as I am planning to make this a series focused on AI and the practical considerations for the future of finance work.

Toyoyuki Ushioda
Founder of World Finance Edu

Subscribe now

Author: Toyoyuki Ushioda

With 20+ years in financial and operational management, he has held senior leadership roles including Global Head of FP&A at a US$125B asset management firm (Real Estate/PE/VC).

Building on this experience, he founded World Finance Edu and is also an Angel Investor, VC Limited Partner, Venture Partner, and Financial Advisor.

He is a Certified Public Accountant (Washington State) and holds an INSEAD Global Executive MBA as well as certificate in business excellence at Columbia Business School’s Private Equity & Venture Capital Program.


World Finance Edu Website