In partnership with
The most common question I get from people building their data portfolio: what kind of project should I actually build? And my answer is ‘Something that looks like real work’, so when hiring managers are looking through it, they can mentally apply it to their business.
This week I recorded a full video walking through 5 of them in detail — the business context and why each one stands out to a hiring manager. Link at the end.
But first, here are the project ideas and what to focus on.
Before we proceed - a small ad. Your clicks on the ads help to cover newsletter hosting fees and make me $1. Thank you!
200+ Claude Prompts Top Professionals Actually Use at Work
Claude can be your analyst, editor, and strategist.
But most professionals are using it to fix grammar.
These 200+ Claude prompts take it from grammar tool to your most powerful AI work assistant.
Sign up for Superhuman AI and get:
200+ ready-to-use Claude prompts to get real work done in minutes — researched, tested, and used by professionals at Google, Microsoft, and NASA
Superhuman AI newsletter (4 min daily) so you keep learning new AI tools and skills to stay ahead in your career — the prompts are just the beginning
1. Churn Prediction with a Revenue Number Attached
Almost every industry has a churn problem — telecoms, banks, subscription businesses. But most churn projects stop at "here is a list of customers who might leave."
The version that matters for the business goes one step further: how much revenue is at risk, and which customers should retention actually call first? That second number is what changes a data project into a business decision.
What to focus on: precision vs recall trade-off, ranking customers by churn probability multiplied by their value, and writing up what you would recommend the business actually do.
2. Customer Segmentation That Actually Segments
Most segmentation projects run K-means on whatever numeric columns are available, get five clusters, and stop. A segment is only useful if someone in marketing can look at it and immediately know what to do with it.
It is important to name each cluster in plain English — not "Cluster 2" but "high value, at risk of lapsing" — and explains what the business should do differently for each group.
What to focus on: RFM analysis before touching any algorithm, elbow method to justify your number of clusters, and interpreting your segments in business language, not model language.
3. A/B Test Analysis
People run a test, see one number is bigger than the other, and call it done. No significance check, no thought about whether the difference is real or just noise.
A/B test is the project that proves you understand statistics, not just machine learning.
What to focus on: writing your null hypothesis before you look at the data, reporting the p-value, and translating the result into a plain recommendation — roll it out (I mean the feature or marketing campaign), do not roll it out, or run it longer because you do not have enough data yet.
4. Demand Forecasting
Forecasting is different from everything else — it is about understanding time, trend, and seasonality. Many data science portfolios consist of only classification projects. One forecasting project signals you can handle more than one type of problem.
What to focus on: plotting the raw data and naming the patterns before choosing a model, starting with a simple moving average baseline before reaching for Prophet or ARIMA, and communicating uncertainty clearly rather than presenting a single number as if it is guaranteed.
5. Fraud Detection
This is the project most people get wrong in the same way. They build a model, it comes back 99.5% accurate, and they think they have nailed it. Then someone asks how many actual fraud cases it caught. The answer is usually almost none.
What to focus on: explaining the class imbalance problem out loud, using precision, recall, and a confusion matrix instead of accuracy, and making the case for why a fraud team should trust the model enough to act on it.
6. Credit Risk Scoring
Predicting loan default risk using applicant data — income, employment history, loan amount. Directly relevant for finance, banking and fintech roles.
What to focus on: handling missing data thoughtfully (why someone didn't fill in a field is often itself a signal), calibrating probability scores rather than just predicting a binary outcome, and explaining the model decisions — in real credit scoring, you need to be able to explain why someone was declined, not just that they were.
7. Sales Forecasting Dashboard
Combining Python analysis with a visualisation layer in Power BI or Tableau. Shows the full pipeline from raw data to executive-ready output, which most pure Python projects do not demonstrate.
What to focus on: building the forecast in Python, exporting clean outputs, and designing a dashboard a non-technical stakeholder can actually use and update themselves. This one is particularly strong for analyst roles that sit between data and business teams.
8. Sentiment Analysis on Customer Reviews
Classifying whether customer feedback is positive, negative or neutral. Relevant for any consumer-facing business — retail, hospitality, SaaS, e-commerce.
What to focus on: going beyond running a pre-built model. Analyse which topics appear most in negative reviews, and turn the output into a recommendation about what the business should actually fix first. That last step is what separates a data project from a business insight.
A few more worth considering
If you want more options to choose from:
Price elasticity analysis — how does demand change when price changes, very common in retail and FMCG
Employee attrition — same as churn but for HR analytics, a growing space
Where to find data
You do not need to invent a dataset. These are the sources I use:
Kaggle — kaggle.com/datasets — thousands of cleaned and real datasets across every industry
UCI Machine Learning Repository — archive.ics.uci.edu — academic datasets, well documented
Google Dataset Search — datasetsearch.research.google.com — finds datasets across the web
Our World in Data — ourworldindata.org — well structured global datasets, great for trend analysis
Pick a dataset in a domain you find genuinely interesting.
One last thing
Code is only half the project. The other half is the writeup — what you built, the business question behind it, what you found, and what you would recommend. That is the part many people skip.
This week's YouTube video goes deeper on five of these.
Watch it here → link
Keep pushing 💪,
Karina
Just starting with Python? Wondering if programming is for you?
Master key data analysis tasks like cleaning, filtering, pivot and grouping data using Pandas, and learn how to present your insights visually with Matplotlib with ‘Data Analysis with Python’ masterclass.
Already know the basics and want something more hands-on?
Take the Python Challenge.
You'll work through a real business problem, complete a portfolio-ready project, and practise the kind of analysis employers expect from junior analysts.
👉 Start with the Masterclass if you're a complete beginner.
👉 Choose the Python Challenge if you're comfortable with the fundamentals and want to apply them to a real project.
Data Analyst & Data Scientist
