In this instructional video, Kedeisha Bryan argues that traditional data analyst interview preparation—heavy focus on LeetCode, memorizing SQL syntax, and timed coding tests—is becoming obsolete. She attributes this shift to the rise of AI‑generated code, which she calls the "vibe coding crisis," leading to increased code churn and "comprehension debt" that make technical screens unreliable. Employers at high‑paying companies now prioritize logical reasoning, judgment, ETL thinking, and the ability to handle messy, unstructured data over raw syntax knowledge. Bryan advises candidates to articulate assumptions, edge cases, and trade‑offs, and to use pseudocode and verbal walkthroughs during interviews. She highlights the growing importance of take‑home assessments from firms like Airbnb, Stripe, and Roku, urging candidates to treat these like consulting engagements—delivering slide decks or memos, anticipating stakeholder questions, and providing actionable recommendations. A central theme is data storytelling: a four‑step framework (state the problem, show the data, provide insight, give recommendation) and precise, insight‑driven chart titles. Bryan stresses a mindset shift from mastering tools to mastering ambiguous problem solving, positioning analysts as trusted advisors rather than replaceable code generators. Throughout, she intersperses personal background, statistics, and calls to action for her community resources.
Chapters
Everything that you have been told about passing data analyst interviews is officially dead. Now hear me out. You're probably grinding through LeetCode every night, memorizing SQL syntax like your life depends on it, or stressing over a live Python test, treating the whole process like a high school exam where one wrong answer means that you fail. But the highest paying companies aren't even using that as a main filter anymore. So today I'm going to show you what the new interview processes are looking like right now and why the people who are actually adapting to it are landing high paying roles while everyone else is still not succeeding in their interviews. Now, if this is your first time here. I'm Kedeisha, and 6 years ago I was delivering pizzas and Amazon packages for about $8 an hour. And today I am running a community of over 80,000 current and aspiring data professionals, and I've helped thousands of other career changers break into data analytics without going back to school and wasting money on courses that just don't work. If you're ever interested in working with me to land your first role in data analytics, click the link in the description. Let's get started. All right, so the first major shift is something the interview prep industry wants you to ignore. You're spending hundreds of hours trying to write a flawless code from memory. And you think that just mastering syntax, every comma, every parentheses, every semicolon will get you the job. And you think that your value as an analyst is directly tied to your ability to act as a human compiler. But here is a massive problem no one is talking about. AI has now made technical skills incredibly easy to fake. We are living in the era of the vibe coding crisis. Andrej Karpathy, the former director of AI at Tesla and the co-founder of OpenAI, coined the phrase vibe coding in early 2025 and described it as fully giving into the vibes and just forgetting that the code even exists. Just seeing stuff, saying stuff, running stuff, copying, pasting, and thinking that it mostly works. And when he said that, people thought it was a joke, but now it has become a crisis. According to GitHub, 41% of all new code written globally is now AI generated, and companies are paying for it the hard way. An analysis found that code churn, which is code written and then immediately reverted or fixed in 2 weeks increased 39% in projects heavily using AI tools. Google Chrome's engineering lead gave it a name, comprehension debt. It accumulates the moment you ship code you didn't write and you don't fully understand. So people are copying, pasting from ChatGPT, throwing it in production, and hoping that it works. When a data pipeline breaks at 2 AM on a Friday, AI cannot jump on a troubleshooting call. Only a human with critical thinking can do that. So hiring managers know that a candidate can fake their way through a basic technical screen using AI. So while technical screens still exist, They are no longer the ultimate deciding factor for roles that actually pay well. What employers are now testing for is your logic, your judgment, and your ETL thinking. Basically knowing how to think and structure projects end to end. They want to know if you can understand why data is being transformed a specific way, not just how to write SQL statements. They want to see your approach to messy and unstructured problems where the requirements are vague and the data is dirty because it always is. And the best way to really develop these skills is to stop doing isolated perfect scenario coding problems or data projects. LeetCode is not going to save you. I was reading a thread on Reddit where someone went from months of rejections to finally getting an offer. They said the switch was this: they stopped treating SQL rounds as a coding test and started mentioning assumptions, edge cases, trade-offs, and performance considerations in the interview. They focus on explaining their thinking, Not just trying to impress everybody. So that is a shift that you need to make. Before you write a single line of code on a messy dataset, write out the pseudocode. Write out the steps in plain English. When you're in the interview, talk through your thought process out loud. Say, here's a business problem as I understand. Here's how I would extract the data. This specific table usually has null values, XYZ. Prove your logic, prove your critical thinking, and prove that you are the human who governs the tools and fixes the mess when AI breaks something. But even if you can explain your logic perfectly, there is a new gatekeeper waiting for you because it is still easy to fake technical skills on a 30-minute Zoom call. Thanks to all the new cheating tools out there, even the highest paying companies are now leaning heavily on take-home assessments and business case presentations. Real take-home assessments from companies like Airbnb, Stripe, Roku are the new standard benchmark. One hiring manager put it bluntly in a Reddit thread. Some people interview really well, but cannot do the job. Assessments have saved me from bad hires. The company's offering $100K. $120K, $150K aren't just handing you money just because you pass a 15-minute SQL quiz. I went through take-home assessments myself. They were always from the higher paying companies. And it's the same for my students. The companies that pay very, very well want to see how you perform in the wild, not how you perform with a countdown timer. And here's what I've seen over the years. You have one group that gets the take-home prompt, answers specific questions, and then dumps all their code in a spreadsheet and then sends it back and then waits for a rejection email 3 days later. But the other group treats it like a consulting engagement. They don't just answer the questions, they actually solve it like an actual business problem, and then they get the offer. One of my students landed a senior people analyst role at Roku, and the take-home project that Roku gave her was almost identical to a milestone project that we do inside my program. She didn't just send them blocks of SQL code, she sent a comprehensive solution. She identified trends in employee turnover, visualized the cost of that turnover, and presented an actionable recommendation directly to the HR leadership team. Proof is the product. Companies use these take-home assessments because it is the only way to really see that you can do the job. AI cannot sit in front of a VP of marketing and explain why customer churn is increasing based on a messy dataset. AI cannot defend its methodology when a stakeholder pushes back, and AI cannot build trust. So to really prepare for these take-home assessments, you first need to read prompts and identify the core business objective immediately. What problem are they actually trying to solve? Is it increasing revenue, cutting costs, improving retention? If you don't know the business objective, your analysis is useless. Second, do not submit your Jupyter Notebook or SQL script alone. That is a bare minimum. Submit a slide deck or written memo that summarizes your findings for non-technical audiences. Your hiring manager might be technical, but your stakeholders that you'll be working with on the job are not. Show them that you can communicate with both. Third, anticipate their questions. If you found an anomaly in your data, explain why you think that it's there and what you should do about it. Did you exclude it? Did you average it out? Why? Bridge the gap between raw data and business strategy and show them what it would be like to actually have you on their team. 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. And this brings me to the single biggest reason people bomb the final round. You can have the perfect SQL query, write all the perfect code in the world, but if your data storytelling is terrible, you will fail, period. I know what you're thinking. I'm applying to be an analyst, not a public speaker. That's the exact mindset that keeps people stuck at entry-level, low-paying roles or not even getting hired in the first place. A survey found that 57% of hiring managers now value communication and storytelling more than technical coding skills when evaluating candidates. And 76% of those human-centric abilities are simply difficult for AI to replicate. So these aren't soft skills. This is the marketing telling you what it's paying for. Data without a story is Noise and useless, and even working data professionals still struggle with this. Saw another post on Reddit that said, "Hey, pulling data and cleaning it and building dashboards they can do all day, but they struggle with the so what part. When it's time to present to stakeholders, they completely freeze. They just end up reading the numbers on the slide like a robot. This is exactly why people fail. They pull up the screen and just read numbers out loud. They just point to a bar chart and say, as you can see here, sales went up 10% in Q3, and then the stakeholders just sitting there thinking, so what? Every chart that you build, every bullet that you write, every tool that you mention has to pass." one test. How does this make the business more money or save them time? If your visualization shows inventory costs are rising, the so what isn't just inventory costs are rising. It's what the company needs to do to negotiate vendor contracts immediately, or they lose $50,000 this quarter. If your dashboard shows customer retention dropping after month 3, the so what can be the marketing team needs to trigger a re-engagement email campaign at day 75 to save $100,000 in recurring revenue. Skills are the currency of freedom today, but communication is how you cash it in. So use this framework for every presentation, take-home portfolio, real-world meeting, whatever. 1, state the problem. What is a business issue? Set the stage immediately. 2nd, show the data. Provide evidence that the problem even exists and keep visuals clean. Remove the clutter. 3, provide the insight. Explain why it's happening. This is where your analysis really shines. And then 4, give the recommendation. Tell the business exactly what to do and be prescriptive. One more thing that will instantly elevate elevate your work, never show a chart with a generic title. Sales by region tells me nothing. It forces me to do your job for you. Instead, title it West Region Sales Dropped 15% Due to Supply Chain Delays. Tell the stakeholder what to think before they even look at the graph. Guide their eyes to the truth. Now I want you to stop and really internalize this. Everyone online is telling you to learn more tools. Python, R, Snowflake, dbt, the newest AI plugin. This BI tool, that BI tool, that certificate, whatever. They keep you on a hamster wheel of continuous learning and consumption of tools, so you never actually feel ready to apply. But it's not about mastering tools. It is about mastering thinking in ambiguity. Tools will always change. The syntax will change. There's going to be a new AI advancement every single week, but a business's fundamental need for a human who can look at a chaotic situation, find the truth hidden in the data, and explain it simply to the people making that decision, that need is forever. You are not a human calculator. You are a problem solver. When you shift your mindset from I need to know my code to I need to show them how I think, the entire interview process changes and you stop being a commodity that can be replaced by a cheaper candidate or an algorithm, and you become a trusted advisor. So stop focusing on trying to be perfect and start focusing on trying to be valuable. Stop worrying about whether you memorize the exact syntax of a window function and start worrying about whether you can explain why the window function is needed in the first place. Now, if you want a further breakdown of exactly what to do step by step to land your first role in data analytics, I have a video that shows a roadmap on exactly what you need to do, and you can watch it here. I'll see you in the next one.
