Data Analyst Interview Questions: Top 25 With Answers

Walking into an interview room to prove you know how to handle complex data can feel terrifying. You might know how to write advanced queries in your sleep, but explaining your thought process out loud to a panel of strangers is a completely different game. Hiring managers want more than just a human calculator.

They actively look for someone who can look at raw, confusing numbers, figure out what went wrong, and explain it to the marketing or sales team without putting them to sleep. Preparing for the right data analyst interview questions gives you a massive competitive edge over other candidates. You get the chance to practice your stories, refine your technical explanations, and figure out exactly how to sell your unique skills. This comprehensive guide breaks down exactly what you will face, from standard background checks to tough whiteboarding scenarios, helping you secure that job offer.

General and Behavioral Data Analyst Interview Questions

Behavioral rounds exist to test if you fit the company culture and if you communicate clearly with non-technical staff. Managers want to know if you panic under tight deadlines or if you argue when stakeholders reject your findings. You have to prove you are easy to work with and that you understand the core business side of the job. These non-technical data analyst interview questions carry just as much weight as the coding tests.

1. Tell me about yourself and your background.

Hiring managers use this opening question to gauge how well you pitch your own career story and whether you can communicate concisely. They absolutely do not want to hear about your childhood, your pets, or unrelated personal hobbies. Instead, they are looking for a clear, chronological summary of your past jobs, your current technical skills, and exactly where you want to go next in your career.

You need to start by mentioning your degree or foundational education, transition quickly into your past roles, and highlight one massive win that proves your competence. For example, mention how you used SQL and Power BI to uncover a bottleneck in an app that saved your previous company thousands of dollars. By structuring your answer to focus strictly on data-driven achievements, you instantly prove that you understand the business value of your role. Finish by explaining why this specific senior position is the logical next step for you.

Framework Stage	What to Include	Example Approach
Past	Education & Early Career	Mention your statistics degree and junior roles.
Present	Current Role & Big Win	Discuss building dashboards that optimized ad spend.
Future	Goal Alignment	State your desire to focus on predictive modeling.

2. Why do you want to be a data analyst?

Companies want people who actually enjoy the frustrating process of digging through endless spreadsheets and databases. This job involves a lot of tedious cleaning, formatting, and double-checking numbers. If you do not have a genuine interest in problem-solving, you will burn out fast and leave the company. Talk about the thrill of finding answers hidden in messy information and how you love acting as a detective for the business.

You can explain that you get a huge rush when you finally clean the data and a clear, undeniable trend emerges on the screen. Knowing that your charts and insights directly help the executive management team decide where to invest their budget makes the hard work completely worth it. Emphasize that you love turning chaos into strategy.

Core Motivation	Why It Matters	Interview Impact
Problem Solving	Shows you enjoy the core work	Proves you won’t get bored
Business Impact	Connects your work to revenue	Shows executive-level thinking
Curiosity	Drives continuous learning	Highlights your adaptability

3. What is the difference between an analyst and a data scientist?

Job titles in the tech industry are notoriously confusing and often overlap depending on the company size. The hiring manager asks this to make sure you know what you are actually signing up for and that you have realistic expectations. You need to show that you clearly know your boundaries and responsibilities. As a general rule, analysts look backward at historical events, while scientists look forward to predict future outcomes.

Your job is to explain what happened yesterday and what is happening right now to answer direct business questions. A data scientist usually works on forecasting what will happen tomorrow by building complex machine learning algorithms and writing heavy production code. Make it clear that you focus on actionable business insights, while they focus on advanced predictive modeling.

Role	Primary Focus	Key Deliverables
Data Analyst	Past and Present	Dashboards, Reports, Business Insights
Data Scientist	Future and Predictions	Machine Learning Models, Algorithms

4. How do you handle pressure and tight deadlines?

Stakeholders frequently demand complex reports at the very last minute right before massive board meetings. You will inevitably get urgent emails asking why sales dropped overnight and demanding immediate answers. The interviewer wants to know you will not freeze up, panic, or make careless mathematical errors when the clock is ticking loudly.

Whenever a massive request hits your desk, explain how you handle the pressure by forcing yourself to stop and prioritize immediately. You ask the stakeholder exactly what specific metric they need to make their immediate decision. Usually, they do not actually need a massive fifty-page report right away; they just need one single number. You deliver that single number first to buy yourself the necessary time to finish the rest of the analysis properly without rushing.

Strategy	Action Taken	Result
Triage	Identify the most critical metric needed	Stops immediate panic
Communicate	Set realistic timelines with stakeholders	Manages their expectations
Execute	Deliver a partial win quickly	Buys time for the full report

5. Tell me about a time you had to persuade someone to trust your data.

People naturally hate being told their gut feeling or personal strategy is completely wrong. As an analyst, you will frequently have to tell executives that their favorite passion project is actually losing the company money. You must show extreme empathy and use clever visuals to change their minds gently without bruising their egos.

Imagine a scenario where the head of marketing was convinced a new email campaign was driving website traffic, but your analysis proved the emails had a terrible open rate. Instead of just telling him he was wrong bluntly, you built a side-by-side line chart comparing the metrics. By visually mapping the failure of the emails against the success of an organic social media post, you let the stakeholder arrive at the correct conclusion themselves.

Persuasion Step	Purpose	Outcome
Empathy	Acknowledge their perspective	Reduces defensiveness
Visual Evidence	Use clear charts instead of raw numbers	Makes the truth undeniable
Recommendation	Offer a pivot strategy	Turns a failure into an action plan

Fundamental Analytics Concepts

You absolutely cannot build a strong career without a rock-solid foundation in the basics. These questions test your grasp of the fundamental rules of working with digital information. Interviewers check if you know the exact differences between cleaning, wrangling, and structuring. If you mess up these basic definitions during the interview, they will immediately assume your technical skills are also incredibly weak.

6. What is your standard data analysis process?

This question specifically checks if you work systematically or if you just dive blindly into a database and start guessing. You need to outline a highly logical, step-by-step workflow that you follow every single time. Never tell the interviewer that you start by writing code or building charts. You must always start by talking to the business team to understand the actual problem they want solved.

Once you define the business question, you locate and extract the relevant data from the company servers. The next, and usually longest, step is cleaning and standardizing all that chaotic information. After that, you run your exploratory analysis to find hidden trends, and finally, you build visualizations to present the final story to the leadership team.

Process Step	Description	Importance
Define Scope	Understand the business question	Prevents wasted effort
Data Collection	Extract records from databases	Gathers the raw materials
Data Cleaning	Fix typos and missing values	Ensures absolute accuracy
Visualization	Build charts and present insights	Makes data digestible for leaders

7. Explain descriptive, predictive, and prescriptive analytics.

This is a classic textbook question that proves you understand the different levels of strategic value a data team provides to an organization. You need to keep your definitions incredibly simple and attach a real-world business scenario to each one for clarity. Descriptive analytics tells you exactly what happened in the past, like looking at a dashboard showing that last month’s user registrations dropped by five percent.

Predictive analytics uses those historical trends to guess what will happen next, forecasting that registrations will likely drop again next month due to upcoming seasonal holidays. Finally, prescriptive analytics goes a step further and tells the business exactly what to do about it, like suggesting an algorithm increase ad spend by ten percent next week to offset the predicted drop.

Analytics Type	What It Answers	Real-World Example
Descriptive	What happened?	Last month’s revenue report
Predictive	What will happen?	Forecasting next quarter’s sales
Prescriptive	What should we do?	Recommending optimal discount prices

8. Why is data cleaning so important?

Anyone who has worked in this field knows that real-world data is breathtakingly ugly. It is full of typos, missing fields, corrupted characters, and completely wrong entries. If you do not deeply care about data quality, your reports will single-handedly ruin company strategy. You must show the interviewer you respect the golden rule of analytics: garbage in, garbage out.

It absolutely does not matter how beautiful your final presentation is; if the underlying records are dirty, your insights are dangerously wrong. Cleaning removes hidden duplicates, fixes null values, and standardizes formats so everything aligns perfectly. If you accidentally include duplicate purchase records, the company might falsely think they hit their massive revenue targets when they actually missed them completely.

Common Issue	How to Fix It	Risk of Ignoring It
Duplicates	Run a distinct drop function	Inflates revenue or user counts
Null Values	Impute the median or drop rows	Breaks mathematical calculations
Bad Formats	Standardize dates and text strings	Prevents tables from joining

9. How do you define data wrangling?

People outside the industry often confuse cleaning with wrangling, but they are two very distinct processes. Cleaning is just about fixing small mistakes and typos in a dataset. Wrangling, however, is the heavy lifting of reshaping and molding the data so your specific software tools can actually read and process it.

It is the much broader process of taking raw, chaotic information and transforming it into a highly structured format. For example, you might take a messy JSON file from a web scraper, extract the exact text fields you need, join it with a clean SQL table using a unique key, and pivot the rows into columns. It is basically preparing all the raw ingredients perfectly before you actually start cooking the final meal.

Wrangling Task	Description	Tool Used
Extraction	Pulling data from a raw source	Python (Requests/BeautifulSoup)
Transformation	Changing the shape of the data	Pandas or SQL
Loading	Moving it to the final destination	Database Connectors

10. What is the difference between structured and unstructured data?

Modern companies store massive amounts of data in many completely different ways. You need to prove to the interviewer that you know how to handle a perfectly neat spreadsheet versus a messy folder full of random text files. Structured data is highly organized and behaves predictably. It fits perfectly into standard rows and columns, like a typical customer database containing names, phone numbers, and home addresses.

Unstructured data has absolutely no predefined format and is completely chaotic. It includes things like audio recordings, raw text from thousands of customer complaint emails, or marketing images. Structured data is incredibly easy to query using basic SQL, while unstructured data usually requires highly specialized tools to extract any real value.

Data Type	Characteristics	Examples
Structured	Highly organized, tabular	Excel sheets, SQL databases
Unstructured	No fixed format, messy	Emails, audio files, social media posts

Technical and Tool-Based Data Analyst Interview Questions

This is the phase of the interview where the conversation gets tough. You have to prove you actually know the software you confidently listed on your resume. Be ready to explain your favorite analytical functions, exactly how you write your queries, and why you choose certain tools over others for specific tasks. Do not exaggerate your skills here; they will easily catch you during the live technical test.

11. What tools do you prefer for your daily work?

Interviewers ask this because they want to know your core tech stack, but they also want to see if you are flexible enough to learn their company’s specific stack. You should be completely honest about what software you love, but you must emphasize your adaptability. You might explain that your absolute favorite stack is using SQL combined with Python.

You use SQL to pull exactly what you need from the massive warehouse, and then you load it into Python because it handles heavy datasets much faster than Excel ever could. For presenting to stakeholders, you prefer a visualization tool because the drag-and-drop interface lets you build great-looking charts rapidly. However, make sure you explicitly state that you are completely comfortable switching to their preferred tools immediately.

Preferred Tool	Primary Function	Why It Excels
SQL	Data Extraction	Handles massive relational databases
Python (Pandas)	Deep Manipulation	Processes huge files without crashing
Tableau / Power BI	Visualization	Creates interactive, shareable dashboards

12. Explain the difference between a database and a data warehouse.

If you pull millions of rows of data from the wrong server, you could accidentally crash the company’s live operational website. This specific question proves you understand basic data architecture and respect server safety protocols. A database is strictly designed to handle day-to-day operations in real time. When a customer buys a shirt on your website, that exact transaction is instantly recorded in the operational database.

A data warehouse, on the other hand, stores massive amounts of historical data pulled from many different databases across the entire company. As an analyst, you must always query the warehouse because it is specifically optimized for reading huge datasets. Running a massive analytical query on the live database could slow down the actual website and ruin the customer experience.

System	Primary Purpose	Analyst Interaction
Database (OLTP)	Real-time operations and transactions	Rarely queried directly by analysts
Data Warehouse (OLAP)	Storing historical, aggregated data	The primary source for running reports

13. How do you handle missing data?

Almost every single dataset you will ever touch has massive holes in it. You need to show the hiring manager that you possess a toolkit of various statistical methods to deal with blanks without ruining the integrity of the whole project. You must explain that your method depends entirely on why the data is actually missing and exactly how much is missing.

If only one percent of the rows have missing ages, you might just choose to drop those rows completely. But if thirty percent of the data is completely missing, dropping it would violently skew your sample size. In that severe case, you might impute the data by filling the blanks with the median or mean age of the overall group.

Missing Data Volume	Recommended Action	Why Use It
Less than 5%	Drop the rows	Quickest fix, negligible impact
10% to 30%	Impute with Mean/Median	Preserves the overall dataset size
Over 40%	Investigate root cause / flag it	Data is highly unreliable, needs engineering

14. What Python libraries do you use most often?

If the job description strictly requires coding, you absolutely must know the standard analytical ecosystem. Listing these libraries confidently shows the interviewer you have real, hands-on experience and aren’t just reciting theory from a boot camp. You should state that you practically live inside the Pandas library because it is absolutely essential for filtering, grouping, and merging giant data frames seamlessly.

You also rely heavily on NumPy whenever you need to execute heavy mathematical operations on complex arrays. When it comes time to build visuals directly inside your Jupyter notebooks, you generally start with Matplotlib to get the basic structural charts down, and then you layer on Seaborn when you need the final graphics to look incredibly polished and professional.

Python Library	Main Use Case	Typical Scenario
Pandas	Data Manipulation	Merging two CSV files together
NumPy	Mathematical Operations	Performing complex array calculations
Seaborn	Advanced Visualization	Building a visually appealing heatmap

15. Describe Pandas Series versus DataFrames.

This is a hyper-specific technical check designed to weed out people who just memorized buzzwords. If you claim to know Python on your resume, you absolutely must know the fundamental difference between these two basic architectural structures. You can explain that a Series is strictly a one-dimensional array. You can easily picture it as a single, vertical column of data inside a standard Excel sheet, where each entry has an index tag.

A DataFrame, however, is a fully two-dimensional data structure. It is essentially a collection of multiple different Series neatly put together to form a complete table, complete with horizontal rows and multiple named vertical columns. You use a Series when looking closely at a single variable and a DataFrame for the entire dataset.

Data Structure	Dimensions	Visual Comparison
Series	1-Dimensional	A single column in a spreadsheet
DataFrame	2-Dimensional	An entire spreadsheet table

16. What is SQL, and what functions do you use most?

SQL is widely considered the absolute backbone of the entire analytics industry. You must express extreme confidence in your personal ability to join massive tables and aggregate numbers efficiently. You start by defining that SQL stands for Structured Query Language, and it is the exact language we use to communicate directly with relational databases to extract targeted information.

You emphasize that you use it literally every single day. Detail how your most used commands are various JOINs to properly connect customer demographic profiles with their historical transaction records. Furthermore, you heavily rely on aggregate functions like SUM, AVERAGE, and COUNT tightly combined with a GROUP BY clause to summarize millions of rows into clean, readable metrics.

SQL Function	What It Does	Why Analysts Use It
JOIN (Inner/Left)	Connects different tables	Merging customer names with their orders
GROUP BY	Groups rows sharing a property	Finding total sales per city
CASE WHEN	Creates conditional logic	Tagging customers as “High” or “Low” spenders

Statistics and Mathematical Questions

You certainly do not need a Ph.D. in applied mathematics to be a successful analyst, but you absolutely do need to understand how numbers naturally behave. Managers will heavily test your ability to find real correlations, accurately measure spread, and correctly validate your sample sizes. If you choose to ignore the underlying math, you will confidently draw false conclusions that could seriously hurt the company’s revenue.

17. What is the difference between variance and covariance?

Understanding data spread and directional relationships is critical for finding actual, usable patterns. You must keep your definitions highly straightforward and actively avoid overcomplicating the underlying math during the interview. Variance strictly looks at one single variable. It tells you exactly how far spread out the numbers are from their overall average.

If we closely analyze customer ages, a high variance means we have both very young teenagers and older retirees shopping with us, while a low variance means almost every customer is around the exact same age. Covariance, however, looks closely at two entirely different variables simultaneously. It tells you the directional relationship between them, proving whether they move together or aggressively push apart.

Statistical Term	What It Measures	Example
Variance	Spread of a single variable	How wide the age range of users is
Covariance	Direction of two variables	As marketing spend rises, does revenue rise?

18. Explain univariate, bivariate, and multivariate analysis.

This specific question shows the hiring manager exactly how you approach exploring a brand-new dataset, starting from simple sanity checks and getting progressively more complex. Univariate analysis looks strictly at just one single variable at a time, like calculating the overall average salary of all employees in a company spreadsheet.

Bivariate analysis takes the next step and compares exactly two variables to see if they are somehow connected, like mapping years of professional experience directly against salary to see if older workers reliably make more money. Finally, multivariate analysis looks deeply at three or more variables all at once. This complex method helps analysts understand real-world situations, like how salary is impacted by experience, education, and location combined.

Analysis Type	Number of Variables	Core Objective
Univariate	One	Describe the data and find basic patterns
Bivariate	Two	Determine if a relationship exists between two things
Multivariate	Three or more	Understand complex, overlapping interactions

19. What is data normalization?

When you try to mix numbers that are measured on totally different scales, predictive algorithms get incredibly confused. You have to definitively prove you know how to safely level the playing field before running advanced models. Normalization is the critical process of systematically adjusting values measured on different scales to one common scale, usually shrinking them to fit tightly between zero and one.

You use this heavily if you are manually preparing clean data for a complex machine learning model. If you feed an algorithm a customer’s age, which is around thirty, and their annual income, which is over one hundred thousand, the algorithm might falsely think income is vastly more important just because the raw number is larger.

Problem	Solution	Benefit
Different numeric scales	Apply Min-Max Normalization	Prevents large numbers from dominating the model
Extreme Outliers	Apply Z-Score Standardization	Handles extreme values effectively

20. Can you explain the central limit theorem?

This is widely known as the most famous foundational theory in all of statistics. You absolutely must know it to prove you actually understand basic probability theory and proper sampling techniques. The central limit theorem simply states that if you take enough random, independent samples from absolutely any population, the averages of those specific samples will eventually form a perfect normal distribution, which looks exactly like a smooth bell curve.

The truly amazing part of this theorem is that this phenomenon happens even if the original population data is heavily skewed and completely chaotic. This mathematical theorem is the entire reason professional analysts can confidently take a small sample of user data and make safe predictions about millions of users.

Concept	Explanation	Why It Matters for Analysts
Sampling	Taking small chunks of data	Allows analysis without needing the entire database
Bell Curve	The shape the sample means form	Provides a predictable, mathematical baseline
Inference	Making guesses based on samples	Lets us predict user behavior confidently

21. What is time series analysis?

Nearly all vital business metrics are tracked extensively over periods of time. Knowing exactly how to handle dates, tracking changes, and manipulating time stamps is a strict daily requirement for this job. Time series analysis specifically involves looking very closely at data points that are collected at highly regular time intervals, like meticulously tracking the exact number of active mobile app users every single hour.

The main strategic goal here is to find deep, underlying patterns as time progresses. When you look at time series data, you are actively hunting for two specific things. First, you look for the long-term trend to see if user growth is generally rising or shrinking. Second, you hunt for predictable seasonality spikes.

Time Series Component	Definition	Business Example
Trend	The long-term direction of the data	Overall company revenue increasing over 5 years
Seasonality	Predictable, repeating short-term cycles	Retail sales spiking massively every December
Noise	Random, unpredictable data jumps	A one-day sales spike due to a celebrity tweet

Problem-Solving and Scenario-Based Questions

Hiring managers use these tough scenario questions to actively watch your brain work in real time. They will intentionally give you a terribly vague problem and demand that you fix it on the spot. Do not jump straight to the final answer immediately. Talk completely out loud and carefully show them your logical, step-by-step troubleshooting process.

22. How would you deal with messy data from multiple sources?

In corporate reality, data is rarely stored beautifully in one single, perfect place. You have to prove you know exactly how to safely merge completely different tables together without accidentally creating thousands of duplicates or dropping vital records. First, you would carefully inspect the separate raw files to deeply understand how they are uniquely structured.

Your main goal is to hunt down a unique identifier, like an exact email address or a specific user ID, that reliably exists in all the disparate sources. Once you secure that key, you use Python or SQL to aggressively standardize the formatting. You make absolutely sure all the dates are in the exact same syntax. Only then do you securely execute the final join command.

Step	Action	Critical Focus
Audit	Review all incoming datasets	Identify structural differences
Match	Find a common primary key	Ensures a successful, clean join
Verify	Count rows before and after merging	Confirms no data was accidentally deleted

23. You notice a sudden twenty percent drop in sales. How do you investigate?

This scenario rigorously tests your core business intuition and critical thinking. Good analysts absolutely never panic; they systematically dissect the business problem layer by layer until they uncover the true root cause. You must clarify that you never assume the underlying data is immediately correct.

Your very first step is always to check quietly with the data engineering team to make perfectly sure a tracking pixel didn’t magically break or a database server didn’t silently crash overnight. If the data pipeline is totally healthy and the severe drop is actually real, you start aggressively segmenting the numbers. You slice the sales data by specific device type, traffic source, and exact geographic region to isolate the bleeding.

Investigation Phase	Question to Answer	Potential Finding
Pipeline Check	Is the data recording correctly?	A broken API is failing to log sales.
Segmentation	Where exactly did the drop happen?	Only iOS mobile users stopped buying.
Root Cause	What changed right before the drop?	A new app update broke the checkout button.

24. How do you evaluate if your data model is actually good?

Building a basic statistical model is incredibly easy. Proving it actually works accurately on totally future data is extremely hard. You must explicitly explain exactly how you test your hard work before handing it over to the business leaders. You explain that you absolutely never test a new model on the exact same data you used to train it initially.

You always rigidly split your entire dataset, using roughly eighty percent to actively train the algorithm and securely hiding the remaining twenty percent away. Once the model is fully built, you force it to make fresh predictions on that heavily guarded hidden twenty percent. Since you already know the real answers, you can perfectly compare the model’s blind guesses against reality.

Evaluation Metric	Model Type Used For	What It Tells You
Root Mean Square Error	Regression Models	How far off your numeric predictions are
Accuracy	Classification Models	Total percentage of correct guesses
Confusion Matrix	Classification Models	Breaks down false positives vs. false negatives

25. How do you present highly technical findings to non-technical people?

If the CEO or VP of Sales doesn’t understand your incredibly detailed chart, your entire analysis is completely useless. You must vigorously prove you can strip away all the heavy math and focus strictly on the immediate money and the actionable strategy. You completely remove all the heavy technical jargon from your vocabulary.

The marketing director absolutely does not care what specific Python library you used or how gracefully you handled the null values in the database. They strictly only care about what the numbers mean for their specific upcoming campaign. You always start your final presentation with the major conclusion right there on the very first slide, using extremely clean, simple bar charts and clear, readable line graphs.

Presentation Element	Do	Don’t
Visuals	Use simple bar charts and clear colors	Use messy 3D pie charts or complex scatter plots
Language	Focus strictly on business revenue	Explain the mathematical formulas you used
Conclusion	Start with the answer right away	Make them wait until the very last slide

Final Thoughts

Successfully landing a highly lucrative, great job requires vastly more than just knowing exactly how to quickly write a standard SQL query. By vigorously practicing these top data analyst interview questions, you actively train your brain to think incredibly clearly and logically under intense pressure. Remember to always focus heavily on the direct, measurable business impact of your specific work.

Executive managers fiercely hire analysts who can solve real, painful problems and boldly communicate those smart solutions effectively. Review your entire history of past projects, polish your complex technical explanations, and walk confidently into that interview room fully ready to show them exactly how you can turn their messy, chaotic data into serious, undeniable corporate value.

Frequently Asked Questions (FAQs) on Data Analyst Interview Questions

Preparing thoroughly for the standard data analyst interview questions is great, but anxious candidates often deeply worry about the unspoken, confusing rules of the hiring process. Here are a few uncommon, highly specific concerns you might uniquely face during your job hunt.

Do I need a portfolio if I already have a related university degree?

Yes, absolutely. A university degree heavily proves you sat down and read the theoretical textbooks, but a robust portfolio proves you can actually sit down and execute the real work. Hiring managers desperately want to see your actual raw code. Having a fully public GitHub repository loaded with two or three very clean, meticulously documented projects actively exploring real-world datasets puts you incredibly far ahead of lazy candidates who only submit a standard PDF resume.

How do I handle a technical question when I truly do not know the answer?

You must never, ever try to fake your way through it. Experienced interviewers can easily spot a lie almost immediately. The absolute best approach is to comfortably admit you haven’t used that exact, specific function or abstract concept recently in your work. However, you must immediately follow up by explaining clearly and exactly how you would figure it out. Say something highly professional like, “I am not intimately familiar with that specific Pandas function right now, but I would immediately read the official documentation or thoroughly search Stack Overflow to safely implement it.”

What should I expect from a take-home data assignment?

Tech companies very often give you a highly dirty, chaotic dataset and ask you to return a perfectly polished report in forty-eight hours. Do not attempt to build insanely overly complex machine learning models simply to show off unless they specifically ask for it in the prompt. They are almost always just strictly testing your basic data cleaning skills, your unique ability to write highly clean code, and your ability to confidently pull out one or two very solid, actionable business insights.