Build these 5 data analyst projects this weekend and get a job

When I was trying to break into data, I wish someone had told me about these 5 projects. No, not the Titanic dataset or another Netflix dashboards. These are 5 projects that are actually getting hiring managers to call you back for interviews. And they're going to cost you about $0 to get started. All you're going to need is a laptop and one weekend for each project. And every one is built around a problem that companies in the highest paying industries pay analysts to solve. So you're not just following a tutorial. It's like doing the actual job. Make sure you stick around for number 4, because if you build only one project from this video, It's that one. Now, if we haven't met yet, I'm Kedisha. I used to deliver pizzas and became a self-taught data scientist, and I have helped thousands of other career changers land data analyst jobs at big Fortune 500 companies like Microsoft, Roku, and more. And if you're looking for personalized help to pivot into data analytics, click the link in the description. Now, something that I have seen over and over again is people spending weeks or months building projects, putting them on GitHub, making their portfolios look really nice, but but still struggling to land interviews. And a lot of the time, it's not because they don't know SQL or Excel, it's because their projects don't show a hiring manager how they'd actually perform or use those skills on the job. And that's what we're going to fix today. Now, I didn't just pick 5 random datasets and call them portfolio projects. There's a reason I chose these specific problems, and understanding that reason is going to help you choose better projects long after this video. First, I looked where the money is. Right now, some of the highest paying industries for a senior data analyst are financial services, pharma and biotech, and information technology. According to Glassdoor, all 3 have the median total compensation of about $130,000. Now, that doesn't mean you're going to land a $130,000 job just because you complete one project, but it does tell us something useful about the kinds of industries that value data analytics. So instead of just picking projects because the dataset looks interesting, we're going to focus on problems companies in these industries actually need people to The second thing I looked at was skills. SQL, Power BI or Tableau, Python, and AI are all skills worth developing. And I know that list can feel overwhelming when you're just getting started, but you don't need to learn it all for this weekend. Each project focuses on a specific skill or a specific combination of skills, so you can choose one based on what you already know, what you need to practice. And here's the most important part. I don't want you finishing one of these projects and saying, look, I just made a pretty dashboard. I want you to be able to say, here's a high-value problem that I found. Here's what it's costing the business, and here's exactly what they need to do about it. Because that's the conversation that you want to be having in an interview. For every project, I'm going to walk you through the high-value business problem that I found, the data that you need, what to analyze, and how to turn your findings into a recommendation. And when you're done, I'll show you how to package that work so recruiters and hiring managers can actually see it. You can spend Saturday night doing the analysis and Sunday putting everything together. So let's start with one of the most common problems companies hire analysts to solve, and that's customer churn. Think about how many subscriptions you pay for every month. Maybe it's Netflix, Spotify, YouTube Premium, your phone plan, your gym membership. Now imagine you're the company collecting these payments. Every time someone cancels, you're losing revenue you were expecting to receive next month. That is customer churn. And for a subscription business, understanding why people leave is a big deal because keeping an existing customer is cheaper than finding a new one. Bayes Research found that increasing customer retention by just 5% can increase a company's profits anywhere from 25% to even 95%. So if you can help a company understand which customers are leaving, how much revenue it's losing, and where it should focus its retention efforts, you're solving a problem that actually matters to the bottom line. And this is a great place to start if you already know just a little bit of SQL. Here's what you can do. You can find a free customer subscription dataset, ideally something that includes customer ID, signup date, cancellation date, subscription subscription plan and monthly payment amount. If it includes things like contract type or how the customer originally found the company, even better. And once you've got your data, I want you to answer a few questions. How many customers are leaving each month? Which customers are likely to leave? And how much money is the business losing because of it? Now, a beginner might just run a query and see that 300 customers canceled and put that number on a dashboard. And technically, that's analysis. But if you're the person running the department, knowing that 300 people canceled doesn't tell you what to do next, and they probably had an idea of what the number was already. You need to understand what's happening behind that number. Numbers without context doesn't matter. So the very first thing I want you to do is group customers by the month they signed up, then use SQL to check how many customers from each group are still paying after 1 month, 3 months, 6 months, etc. That is called cohort analysis. It sounds complicated, but you're basically just following groups of customers over time to see when they start leaving. And this is something subscription companies analyze all the time. And once you do that, you can start breaking down cancellation rates. Are people on monthly plans leaving faster than those on annual plans? Are customers more likely to cancel during their first 90 days? Does one subscription have a higher churn rate than others? Make sure you are comparing rates, not just total numbers of cancellations. Because a bigger customer group will naturally tend to have more cancellations. And then I want you to connect all your findings to revenue. If a customer is paying $100 a month and then cancels, that is $100 in monthly recurring revenue the company can no longer receive. Start adding those numbers up across customers who left, and then suddenly you're not just looking at cancellations, you're looking at revenue at risk. Now imagine your analysis shows that the company is losing about $40,000 in monthly recurring revenue, and most of that loss is coming from customers on month-to-month plans who cancel within their first 90 days. Your recommendation may sound like something like this. We're losing $40,000 a month, and customers who are on the monthly plans are leaving 3 times faster during their first 90 days. I'd focus our retention efforts on those customers before they reach that point. Now, those are just examples. Of course, your dataset will tell you what your real findings are, but do you see the difference? Instead of just walking into an interview and saying that you just built a random churn dashboard using SQL, you can explain exactly what you found, why it matters, and how you tied it to revenue impact. And most importantly, what you'd recommend. That's how you start showing a hiring manager that you understand the business side of analytics. And once you understand how to measure the money a company loses when customers leave, the next project introduces a different challenge, stopping money from disappearing in the first place. If you're interested in working in banking or financial services, this is a project I'd seriously consider putting in your portfolio. Because one of the biggest problems these companies deal with is fraud. In 2025, Americans reported losing $15.9 billion to fraud, according to the FTC. And that's just the fraud that was reported. Banks, credit card companies, and fintech companies are constantly looking for ways to identify suspicious transactions before money gets lost. But there's a catch. They can't just decline every transaction that looks unusual. Think about the last time your bank blocked your card when you were just trying to buy something. Maybe you were just traveling or making a purchase you normally don't make. You know it's your card, but the bank thinks it's something suspicious happening. Now you're frustrated because you can't access your own money. And from the bank's perspective, that's also a problem. So fraud teams are trying to balance 2 things at the same time, catching fraudulent transactions while avoiding unnecessary declines for legitimate customers. That is a challenge I want you to solve with this project. All right, quick pause. Now, if you want to land your first job in data analytics and avoid the lengthy mistakes that I did, click the link in the pinned comment where you can find the exact steps that my students use to land their job in data analytics without going back to school. All right, now let's get back to it. Start by finding credit card transaction datasets that identify which transactions are fraudulent. It's normally hard to find real data around this, so you can use synthetic data for this. Which means transactions are just made up but designed to resemble real financial activity. That way you can work on realistic patterns without needing anyone's private banking information. Now, before we start building anything complicated, I want you to understand where the fraud is happening. Using SQL or even Excel, you can calculate the fraud across different groups of transactions. Look at things like transaction amount, time of day, type of merchant, and account age, depending on whatever your dataset includes. You might find, for example, the fraudulent transactions are more concentrated among newer accounts or in certain transaction ranges. But finding a pattern is only the beginning. The interesting part is deciding what to do with it. Let's say you notice a lot of fraud happening on accounts that are less than 30 days old. You could recommend blocking every transaction from a new account, but that would be a terrible experience for legitimate customers. So instead, I want you to come up with 2 or 3 simple fraudulent detection rules and then test them against the data. Maybe one rule flags transactions over $500 on accounts that are less than 30 days old. Another might combine transaction amounts with high-risk merchant categories. Then you can use Excel to compare what happens when you apply each rule. How many fraudulent transactions would each rule catch? How much fraud-related money would it identify? And how many legitimate transactions would it incorrectly flag? Because a rule that catches almost every fraudulent transaction might still be a bad recommendation if it also blocks a huge number of good customers. So you need to look at both sides. And this is where your project becomes much more than just a simple fraud dashboard. Imagine that your analysis shows that one rule identifies 60% of fraudulent transaction dollars while flagging only 2 legitimate transactions. You could recommend this rule catches 60% of fraud dollars while only flagging 2% of transactions. That is building something immensely valuable to a company. You're not just identifying a problem and just saying, here's the numbers. You're evaluating possible solutions and explaining the trade-offs. And that's exactly the kind of thinking hiring managers want to hear. A lot of the work in data analytics is ambiguous. No one's going to be holding your hand. No one's just going to be giving you a data dictionary. No one's going to be telling you what to do step by step. You have to be able to utilize your critical thinking skills to navigate the ambiguity. But this is exactly the kind of thinking hiring managers want to hear and see. Now you might be thinking, okay, that's great, but I don't want to work in tech or banking. Well, the next project is for you then, especially if you're interested in healthcare. Healthcare has a lot of opportunities for analysts than people realize. And while pharma and biotech made our list for high-paying industries, hospitals and healthcare systems also need analysts to help them solve expensive operational problems. And one of those is hospital readmission rates. Imagine someone goes to the hospital, gets treated, goes home, and then ends up back in the hospital just a few days later. That's a readmission. Now, obviously hospitals want patients to recover and stay healthy. But there's also a financial reason they're paying close attention to this. Medicare can reduce payments to hospitals that have higher than expected readmission rates for certain conditions. In fact, more than 2,300 hospitals are facing Medicare payment penalties under the fiscal year 2027 program. So helping hospitals understand where readmissions are happening and which conditions are driving them is valuable work. And the best part of this project is that you can use real public available data. Medicare publishes hospital readmission data for facilities across the United States on data.gov. I want you to go there, choose your state or area you'd potentially like to work in, and download that relevant data. Then you can open Power BI or Tableau, whichever you choose. Your goal is to build a dashboard that helps a hospital executive understand how their hospital is performing compared to Medicare's expectations and other nearby hospitals. And that's where I see a lot of people go wrong with dashboard projects. They learn how to create 10 different charts that they put all 10 on a dashboard. A bar chart is here, a pie chart is there, a random map in the corner, and a bunch of filters everywhere. And yes, that might show that you know how to use Power BI or Tableau, but who's supposed to use that dashboard? And what exactly are they supposed to do with it? That's what I want you to be thinking about for this project. Imagine you're building the dashboard for a hospital's chief medical officer. This person is busy. They're not going to spend an hour clicking through 15 different visualizations trying to figure out what actually matters. And they need answers yesterday. So I want your dashboard to focus on 3 things. First, which hospitals are performing worse than Medicare's benchmarks? Second, which conditions are contributing to those excess readmissions? You might be looking at heart failure, pneumonia, or hip or knee replacements depending on the data. And third, how does one hospital compare to others in the same area? Once you've answered those questions, think about how you would recommend the hospital to investigate first. Imagine your dashboard shows that heart failure is responsible for most of the hospital's excess readmissions, and that hospital is performing worse than 4 out of the 5 nearby hospitals. Then you can wrap your recommendation around that finding where you can recommend investigation of the discharge and follow-up process for heart failure patients. Notice how much more useful that is than simply presenting a bar chart for readmission rates. You're helping someone understand where to focus first. And here's something else I really like about this project. If you're applying for healthcare analyst jobs, you can build this dashboard using data from hospitals in your own area. Think about how different that interview conversation would be. Instead of saying, I just built a Power BI dashboard using healthcare data, you can say, I actually looked at your hospital's publicly available readmission data and compared it to local hospitals. And here's something interesting that I noticed. That is a much stronger way to demonstrate your interest in the organization. But there's one product I think is more versatile than this. Because so far we've talked about identifying losses and helping companies prevent them. But project 4 is about something almost every company has to decide, where to spend money to generate more revenue. Okay, so this is the one I told you to stick around for. If you only have time to build one project from this entire video, this would be my pick. And the reason is simple. It doesn't matter whether you want to work in tech, finance, retail, or some other industry. If a company spends money trying to acquire customers, someone needs to figure out whether that money is being spent effectively. And that's a problem that you can help solve with data. According to Gartner, large companies spend an average of almost 8% of their revenue on marketing. And that is a lot of money. And the people managing those budgets often need to know which marketing channels are actually bringing in customers, how much those customers are worth, and whether they should be spending more or less on each channel. In this project, we're going to be analyzing marketing performance, but we're not going to be just stopping at which channel gets the most clicks or even which channel gets the cheapest customers. You're going to be figuring out which customers are actually worth the acquiring, and that's way more difficult than it seems. You can start by finding a marketing dataset that includes spending across different channels. That could be Google Ads, social media ads, email marketing, or affiliate marketing, whatever it is. Ideally, you also want information about the customers or orders generated by each channel, including customer IDs and purchase dates if you're going to be analyzing repeat purchases. Then you can use Python to calculate 3 things. If you're comfortable with SQL, you can also do that instead. First, calculate customer acquisition cost, or CAC. That's how much money the company spends to acquire one new customer. If you spend $1,000 on a marketing channel and acquire 20 customers, then you're spending $50 to acquire each customer. Pretty straightforward. Then your next job is to calculate return on ad spend, or ROAS. For every dollar spent on advertising, how many dollars in revenue comes back? This helps you understand which channels are actually producing revenue relative to the money going in them. Now, if you just stop here, you've already built a decent project, but I want you to take it one step further because this is where the analysis gets more interesting. I want you to look at how much those customers spend over time, not just their first purchase. Look at what they spend over the next 3 months, 6 months, or whatever period your data supports. And here's why this matters. Let's say social media ads are bringing in customers for $20 each while another channel is bringing in them for $35. At first glance, you probably think social media is better investment, but why would you spend $35 to acquire a customer when you can just spend $20? But then you look at the customer purchase history, and then you discover that customers from social media ads tend to buy once and disappear, while customers from the more expensive channel come back and purchase again and again. And suddenly that $35 customer might actually be more valuable. If you only analyze the acquisition cost, you would have missed that completely. So then I want you to use those findings to recommend where the company should spend its marketing budget. Don't just rank the channels from best to worst. Think about what would happen if the company moved part of its budget from one channel to another and the impact it would have on revenue. For example, imagine your data shows that social media ads generate the cheapest first-time purchases, but those same customers spend 40% less over the 6 months than customers acquired through email. If the numbers support it, you can definitely recommend that the company should spend more money through email. Versus just social media. That is a recommendation that a marketing director would actually like to A/B test. And obviously the specific numbers are just an example. Your job is to find whatever your dataset supports and be able to explain your reasoning. This is also why I like this project so much for people changing careers. Even if you've never held a title as a data analyst, you might already understand marketing, sales, customer service, or business operations from your previous career. Or heck, you already understand it as a customer. We're all consumers and we all see marketing every day. And because virtually every company cares about how it spends money and how it customers, you can adapt this project to the kinds of jobs you're applying for. But there's one more skill I want you to demonstrate before we finish, because we talked about identifying customers who leave and figuring out which customers are valuable. But what happens when a company wants to understand why those customers aren't happy in the first place? And that's where project 5 comes in, AI customer feedback. Because companies are looking for people who actually know how to use AI, but most people don't. PwC found that workers with AI skills earn on average 62% more than people in comparable roles without AI skills. But the thing is, using ChatGPT to generate a SQL query does not mean you know AI. The real value comes from knowing when AI can help you analyze something, how to use it effectively, and how to check when results are trustworthy. So let's work through a problem where that actually matters. Imagine you're working for a subscription company that receives thousands of customer reviews, complaints, and cancellation comments. Some customers are complaining about the price, others are frustrated because the app keeps crashing, and some are just having problems with customer service, and others feel like the product is missing features that they need. Now, if you're working with thousands of comments, someone could spend days reading through every single one and trying to identify patterns, or you can use AI to help organize that information better, and that's what you would do here. And you might already have some of the data that you need. If your churn dataset from project 1 includes cancellation reasons or customer comments, you can use Otherwise, you can find publicly available App Store reviews for a subscription-based app, or you can find public Google reviews. Once you have your comments, you can use an AI model to categorize them. For example, you might ask it to identify which comment is about pricing, product bugs, customer service, etc. Or you can ask it to identify and create buckets itself. You can also ask it to do sentiment analysis and identify which are positive or negative. If you know Python, you can automate this by sending batches of comments through an API, or you can use a simple free AI tool and work in smaller batches to keep the project at a low cost. But here's the part I don't want you to skip. Once AI gives you its answers, your job is not finished. I want you to select 50 comments and label them yourself. Read through what each customer actually said, decide which category it belongs to, and compare your answers to whatever AI produced. Did it correctly identify the problem? Did it misunderstand sarcasm? Did it put comments about the app crashes into general customer service categories instead of product bugs? If the labels aren't accurate, go back and improve your instructions. And then test it again. Because the goal isn't to prove that AI can generate a bunch of labels, it's to demonstrate that you understand how to evaluate its work and direct AI. You are still responsible for the analysis. And if a hiring manager asks why they would trust your results, you would be able to explain exactly how you check them. And once you're confident in your categories, you can start looking for the patterns. Which issues are customers complaining most often? And if you have cancellation data, Are certain complaints more common among customers who leave? You'll find your patterns and then you'll make recommendations when the data supports it. But look what this project does. We're taking large amounts of unstructured data or unstructured feedback, using AI to make it easier to analyze, checking the quality of the results, and turning those findings into business recommendations. That is a much stronger project than simply saying that you know how to use ChatGPT. And when you combine this project with the churn project, you can actually start answering 2 connected questions. Which customers are leaving and what makes them leave. Now, once you've built your projects, don't just upload it to GitHub and hope a hiring manager figures out what you did. You need to show them why your work matters. So I want you to turn each project into a one-page case study. Explain the business problem, what you found, and the financial impact, and what you'd recommend. Include 2 to 3 visuals to support your findings, and then add it to your no-code portfolio. I always recommend Canva, and then add it to the featured section of your LinkedIn. And don't forget your resume. If you put your projects on your resume, don't say analyze customer churn using SQL. Say something like identified $40,000 in monthly revenue at risk and recommended a customer retention strategy. Just make sure you're clear that that came from a project and not actual work. And then finally, share what you learned in a short LinkedIn post. Because the goal isn't just to prove that you know SQL, Python, or Power BI. It's to show employers that you can actually use those skills to solve high-value business problems, which is what data analytics is about. And that is exactly how to make your portfolio stand out. Now, here's what I want you to remember. You do not need 20 random projects sitting on your portfolio. You only need a few high-value projects and making it easy for someone to understand what you can do for their business. And if you can build one of these this weekend, you'll have something concrete to talk about that goes far beyond just listing technical skills on your resume. But building the project is only part of the process. You also need to know how to explain it in an interview, how to connect it to the experience that you already have, and how to show a hiring manager why you're qualified, even if your past job title doesn't say analyst. And that's where a lot of career changers get stuck. They've been learning the skills, they've been putting in the work, but they don't know how to position themselves and their experience in a way that gets employers interested. That's exactly what I help people do every single day. So if you want help choosing projects for your right background and just want help breaking into data, click the link in the description and book a call and let's talk about it. Otherwise, I'll see you in the next video.