# Automating Excel Data Analysis with Claude AI [Part1]

## Метаданные

- **Канал:** Venkata Reddy AI Classes
- **YouTube:** https://www.youtube.com/watch?v=l4ziwCj3oPQ
- **Дата:** 25.04.2026
- **Длительность:** 20:19
- **Просмотры:** 262
- **Источник:** https://ekstraktznaniy.ru/video/50870

## Описание

Material File Link: https://drive.google.com/drive/folders/1vjlGL_4QxGIFg6g4vI3CEaMgv7svLQvN?usp=sharing

Access the full Course playlist
Link:  https://www.youtube.com/playlist?list=PL2hXNYim5xvqoI5us8HUXhKmwcYzIhFM8

Stay connected with us and receive regular updates through our WhatsApp channel : https://whatsapp.com/channel/0029Vb4ULFP7YScvhbqlkz47

Other Playlists
ML: https://www.youtube.com/playlist?list=PL2hXNYim5xvpbtGR1C2dgAGCtdnhrZNJ-
DL:https://www.youtube.com/playlist?list=PL2hXNYim5xvp8zQb-vAVxvTzudNiRIkjQ
GenAI & Agentic AI Course : https://www.youtube.com/playlist?list=PL2hXNYim5xvon9nmlpkjx_cppODCZi5XR

Welcome to the first part of our AI Data Analyst series, where we use the Claude Pro Excel add-in to completely automate bank marketing data analysis. Watch how we instantly generate a comprehensive dataset overview and basic descriptive statistics using simple AI prompts instead of manual formulas. Learn to enforce a custom "Google style" color theme across your spreads

## Транскрипт

### Segment 1 (00:00 - 05:00) []

Welcome to the session exploring how to use Claude within Microsoft Excel to analyze bank marketing data, generate insights, and build dashboards. You will learn prompt-driven workflows, statistical summaries, segmentation techniques, and efficient reporting for decision-making effectively today. Let me show you something very interesting. I have this data set with me. This is known as bank marketing data set. I have customer number, age, marital status, etc., all these columns. What is their bank balance, whether they have a house loan, loan, etc. Finally, whether they have opted for the term deposit or not. So, this is bank marketing data set. A lot of people know about it. That is not the point. Anyway, I'll talk about this data set later. So, this is the data set information. So, this is by Portuguese bank, and we are trying to do some market campaign analysis. We want to find out who are the customers who are likely to buy our term deposit. Based on that, we want to do some customer segmentation, and we want to find out who are the customers, what type of job, whether married customers will buy our term deposit, what education level, etc. This is the analysis that I want to do on this data set. The point here that I want to highlight, look at this carefully. I have first created the data set overview and well formatted. And then I have done the categorical data analysis to find out by job description, who are the customers who have the high chance to subscribe to our insurance, buy our insurance. So, looks like there are certain type of customers who are retired have the high chance to buy our insurance. At what education level they have the high chance, some unknown. Housing loan, if they have personal loans, who are more likely to buy, in which month they buy. So, some categorical data analysis, a lot of beautiful attractive graphs, and then some discrete variables and the analysis related to that, and then some continuous variables, what is their impact on our target? Our target is whether a customer will buy the term deposit or not. And then the final multivariable multivariate interaction analysis, and we find that two or three variables put together how are they impacting our target? And then finally detailed dashboard where we have the total number of customers, what is actual subscription rate, what is the best combo rate, etc., etc. The whole analysis with the dashboard with all the 3D charts and the key takeaways in every analysis. We have the key takeaways here and the final quick observations, everything. Now, here is the most important point. I have not even touched any of the features in Excel. I haven't created these graphs by myself. charts. I haven't done any of this formatting. All this was done in less than 15 minutes. I would say max to max half an hour, 15 to 30 minutes. That is it. And it has been created by giving certain instructions. Hardly I have given 10 prompts. So, the secret to this whole dashboard and this whole analysis is lying here. If you go to home page, you click on this Claude here. So, this is the place where I have done all the analysis, and my analysis is not really analyzing the data. I was just giving the prompts. I told Claude, you take my data, do a categorical data analysis. First I have asked, give me an overview of the data, then Claude has generated it on its own. Then I asked Claude, let us follow Google type of color coding, Google type of icons, Google type of styling. Then Claude has taken that instruction, and from there on whatever I'm asking, I have asked Claude to do the categorical data analysis. It has given me these categorical data analysis and the graphs. Then I have asked it to do the discrete variable analysis. It has done it on its own, and the insights were also generated on its own. And particularly here, I have asked for 3D type of graphics, and it has created the 3D charts. And then continuous variable analysis, I have asked Claude to create continuous variable analysis. And you will be surprised I haven't touched anything. I haven't created these tables. I haven't formatted. Everything has been done by Claude just like magic. In fact, we are going to do that right now. You are also going to do along with me. Then you yourself will realize that just by giving these instructions, we can create this. And then finally, multivariate analysis followed by the dashboard, followed by the insights. I'm not saying that this is 100% perfect, but I can tell you what it takes for a data analyst at least 2-3 days of time for an experienced data analyst. Or maybe for a beginner, it may take 1 week time to come up with these type of insights. Claude has given us within 15 to 20 minutes. Let us try to use Claude and create this whole dashboard and all these charts. Let's get started. First, uh you open a blank worksheet. You have to install Claude. For that, you go to home page. You click on this simply home. In that, you have an option for add-ins. You click on add-ins. By any chance, if you don't see add-ins here, you may have to check your Excel version. It has to be 2016 or above. If

### Segment 2 (05:00 - 10:00) [5:00]

you have the right Excel version, then you can search for Claude. Then installation of the Claude or just like an add-in, you're just adding it. So that is not going to be very difficult. You can just install Claude and you can log in. And here is the most important point. You must have the Claude Pro account, which costs you around $20 per month as of today, which is fine. Maybe after this session, if you feel that you need a Pro account for 1 month, if it is worth, then you can go ahead and try. Otherwise, you just watch this video. At the end, I'll try to give you another Google Gemini alternative. It is not as effective as Claude plus Excel. You can also try this on Google Sheets plus Gemini. We will try that right at the end, but when I compared Gemini plus Google Sheets versus Excel plus Cloud plus Excel is working fantastic and the results that we are getting from Cloud plus Excel, they are much more accurate than Google Sheets plus Gemini. So, if possible, if you feel that this dashboarding is something or the Excel data analysis is something that you really want to try it out, maybe take for 1 month, you can take the paid subscription. But, first let's observe whether it is going to excite you or not. All right then. First, let's get the data set. Can go to data and then from text file, I will be passing on this data or you can also search in internet for bank market data. You can get it or simply I will attach this data set. You can access this data and otherwise, I'll just give you the file with this two tab with this tab of the data and the second tab is data set explanation. Otherwise, you can get this data with the delimited and then let's say next. Comma separated values, this is a CSV file and then we'll say finish. So, this data set has these columns. The columns are customer number, age, job, what kind of job a customer does, marital status, education, whether they have been a default or not earlier, credit default, whether they have a housing loan or not, personal loan or not. Basically, what we are trying to do is what kind of customers are more likely to buy our term deposit. This is called market campaign analysis. When I'm doing a market campaign, I don't want to send one type of strategy for every customer. I want to find out who are the customers who are most likely to buy my product or respond to me positively, who are the customers who are less likely to respond. That is the analysis that I want to do. So, let's name this rename this as data set. This is my data sheet. Now, the description of the data, we can attach it here. I'll be passing on the description. You can read that description. So, I'll be just copy-pasting this description. I'll be getting it from our previous file. I'll just move that description into this file. So, I will I have opened one file where this data set description is there. This is something that you will get. You don't need to do any analysis for this. So, you will be getting the data set as well as the description. I'll just move it. I'll try to create a copy. Our file name is just new spreadsheet. So, I have the data set and description. The data set description This data set contains information of direct mail marketing campaigns and the goal is to predict whether a customer will subscribe to a term deposit based on their personal profile and previous market campaign interaction. How the customer has been behaving based on his profile information, based on his previous behaviors, we want to find out whether they will subscribe to our new term deposit or not. And we have the data set structure. We have the column names. So, what are the different values a column takes? A detailed data set description is given here. Now, the thing is we do not want to put a lot of effort in understanding this data, in analyzing this data. So, what I want you to realize here is even if you do not understand this data, what we want to do is we want to use Claude to help us in analyzing this data, getting all the basic insights. So, the work that we may have to do within a week, we'll try to achieve that within half an hour. Let us get to our first exercise. We will go to home. I will click on Claude. And then I will try to ask the first question. Try to describe the data for me, each column, and try to give me the details that are very important. So, let us write our first prompt. Now, you're giving the prompt, the quality of the output depends on how detailed you are writing your prompt. Here, I'll try to write in intentionally a shorter prompt, and then you will see the results are not that great. Then if we give a very detailed prompt with a lot of clarity, what exactly you're looking for then the results will be very accurate. So, explain or describe the data set. Explain all the basic descriptive statistics about all the columns. So, I'll just give that then the thinking process has been started. Let

### Segment 3 (10:00 - 15:00) [10:00]

me zoom in. I think the zoom in is not working. Let me just stretch this. Cloud is working and this is a thing reading all the columns like you can click on this. So, the code that it is writing in the back end you can see that 45,212 rows and then it is reading some chunks to give you the insights. Basically, you have asked for basic descriptive analysis that means at overall level how many rows are there, how many columns are there, what are the column names, values? Are there any null values in any of the column? Are there any outliers, glaringly visible outliers? Are there any problems in this data set? That is what we are asking and it is thinking. Usually, this process of doing the basic analysis on the data overview, it may take 2 3 hours for a data scientist but here all this like literally there is an agent that is doing the job of a data analyst here right in front of our eyes. Let us see, it has done all of that. It is trying to respond and you will see that it will create a new sheet. Now, this almost works like magic. It will create a new sheet and the sheet name as of now is temp calculation. Let us see, it may give a different a data set name. It says that do you want me to add the sheet? It is asking for our permission. So, Cloud wants to compute descriptive statistics. Do you want to allow it or do you allow once or allow always? I will say always allow it. We can trust it. It is not a very big problem but yeah, little bit careful you can be. You can do allow once. Deny anyway you cannot do it because you want Cloud to answer this question. So, I will say allow always. So, it is going to compute the descriptive statistics. each step will take certain amount of time depending on the complexity of the question that we have asked. On an average I have observed every time you're asking a slightly larger question, it takes around 5 minutes of time to actually generate your output. Now it's been what I think it's been like 2 3 minutes that we have asked this question. Let us see. It's going by each and every variable. So it depends on what prompt that we have asked. In the prompt if I have asked for detailed analysis or in the prompt if we have asked for just give me uh quick overview, not a detailed analysis, accordingly it will give us the result. In my system it is taking this much of time. Maybe in your system it may not be taking because since this is AI, it doesn't follow one single route to give you the answer. So in my case it is going through each and every column and trying to find everything and it is finding the key observations right away here itself. But intentionally I'm not asking for it because right now I don't want any key observations. I just want a quick overview. I should have mentioned that. So basically when I'm giving the prompt I have given a very short prompt which may be ambiguous. Because of that Claude is taking a lot of time. If I would have mentioned it very clearly, written that prompt very clearly saying that you precisely give me this and nothing else, then Claude have would have given me a right output. Looks like it has completed everything and it has given the data set overview. This is These are all the numerical columns etc. It has given the data set overview here. Ideally it should be adding it over here. Let us see what does it say. So accept all the edits. So I will say ask before edits or accept all the edits. I will just say go ahead, accept all the edits and let us see what is the output that we are going to get. So ideally it should be adding one sheet and showing it here. It has not added. So what I will ask Claude is it hasn't understood. Maybe I would have mentioned in the beginning itself add all these insights to a sheet. Right now it not had added. Let us give that instruction. Add these insights to a new worksheet. So, I have just asked you have found all these add these insights to a new worksheet. So, it is creating a new sheet called insights. I should have given that sheet name as basic details of the data. But anyway, insights is fine, not a problem. So, sheet one we will be deleting it. I don't need that sheet one. Insights is what we will see. Taking time, thinking. Should not be thinking this once I have just asked add these insights. Bank market data set insights and analysis. It has directly gone ahead with uh full analysis, which I'm not really looking for. Just the data set overview is what I'm looking for. That is the prompt that I'm talking about. Your prompt has to be very clear, precise, very detailed. So, it is adding total number of records 45,000. Numerical columns are these many. Categorical columns are 10. Target variable Y subscribed or non-subscribed, missing values, etc., etc. Numerical columns descriptive statistics. For each numerical column, probably in the descriptive statistics it may give minimum value, maximum value, first quartile values, third quartile value, mean value, average value, median value

### Segment 4 (15:00 - 20:00) [15:00]

standard deviation, skewness, kurtosis. It is trying to calculate all of them and trying to good give us those results. These are all the values. Now, that is what I generally say it's almost working like magic. Somebody to create all these or somebody to find all these, it will definitely take good amount of time. And then keeping like this inside a sheet with all these formatting and all that will definitely take time. So, usually what setting that I prefer is I go to view and switch off the grid lines, then it looks slightly better. But that is personal choice. So, numerical columns age slightly in this range skewed slightly skewed and the range is this, mean is this, median is this. For numerical column the interpretation is given target variable class distribution. How many customers have taken the term deposit? responded positively? How many customers responded negatively? So, no did not take the term deposit 39,000 88% did not take 11% only have taken. Heavy class imbalance. Class imbalance means one class is dominating. The other class is having very less number of records. Column distribution. So, categorical columns now it is done with old data set overview. Numerical column descriptive statistics done and then numerical columns interpretation done. Honestly, I'm not looking for this much of analysis in the beginning itself. I'm just looking for the data set overview. I could have mentioned that carefully. But let us go ahead and see what it is trying to give us. Categorical columns distributions blue collar these many people are there etc. I do not want all this. Maybe I can just stop it here because it is doing too much because categorical columns distributions I want to do that separately. So, let me just stop here because I got the basic idea of the data. Then this is what I just want the columns and the column names. That's it. Not beyond that. Now, what I'll do is I'll try to go through each and every variable and try to analyze that. First, I'll start with categorical variables. That is what I will ask Claude. This time I'll try to give a very detailed instruction. Let me see. Analyze all the categorical variables. All the categorical variables and give me the insights. Maybe before this I will try to give an instruction the styling and all that. Let's follow the Google style colors from here on for all the upcoming charts and dashboards. All the colors should be in Google style. All the charts that we are creating should be in Google color coding. All the charts, icons, everything I want to be in Google style. So that it will keep in mind. I think it will give me what are all the distinct color Google uses, what is the color coding, etc. So that I will be working inside a particular theme. So Google yellow, Google blue, etc. Or you can give your own company's color coding. company color theme and ask plot to follow the same color theme. Now we will give our instruction. Analyze all the categorical variables and give me the insights. Create charts. Create maybe bar charts or any other relevant chart charts. Find out the variables that have very high impact on the target. What are target? Like who are the customers who are more likely to buy? So find out the most impactful drivers or segments inside this categorical variable list. I can give a better prompt. I can actually go to another LLM and ask for a better prompt. We can do that as well. But I have a feeling that if you spend some time, maybe give five, six lines, or maybe 10 lines, explain ChatGPT or here in this case Claude, explain it in a better detailed manner. Obviously, it should understand our requirement and give us the result. So, let us go ahead. So, I have given instruction and I should have added one more line. So, let me just stop this. Add the details to a new sheet. I think it should do it, but still to be on the safe side I'm giving add all the analysis to a new sheet. Should do it automatically, but you never know. Better to be on the safe side. I'll analyze all the category variable, identify key drivers, and I'll build a comprehensive analysis sheet with charts. Let me first read the data, compute cross-tabulations with target variable. So, it is doing now this step takes time, but definitely much lesser time than a general human. If a human takes around half a day of time, it may take around 5 minutes of time. Let us observe.

### Segment 5 (20:00 - 20:00) [20:00]

By completing this session, you can leverage Claude in Microsoft Excel to analyze data sets, identify actionable segments, and create dashboards. In the next session, you will learn how to automate data analysis efficiently. —