I recently completed a data analysis project using a Jumia product dataset, and honestly, it looked pretty simple when I first started.
The goal was to take the raw product data, clean it, find some useful patterns and then turn everything into an interactive Excel dashboard.
What I learned pretty quickly is that the dashboard is actually one of the last things you should worry about. There is quite a bit that needs to happen before you get there.
Starting With the Data
The dataset had 115 product records with information such as product name, current price, old price, discount, reviews and ratings.
Before doing any analysis, I had to clean the data.
There were duplicate records, inconsistent rating formats, missing ratings and reviews, negative review values, spelling issues, extra spaces and even a product where the price was given as a range instead of one value.
I kept the original data untouched and worked on a separate cleaned dataset. I also recorded the cleaning decisions because I wanted to be able to explain what I changed and why.
One thing I was careful about was missing values. A product with a missing number of reviews or rating does not mean it has zero reviews or ratings, so I did not just replace those values with zero but rather noted that it had missing reviews/ratings.
After cleaning the data and removing duplicates, I had 112 products for the main analysis.
Revelation?
One of the questions I wanted to answer was whether bigger discounts resulted in more customer engagement.
The answer was not what I expected.
The correlation between discount percentage and reviews was approximately negative 0.17. That is a weak negative relationship.
Medium discount products averaged about 15.9 reviews, while high discount products averaged about 10.8 reviews.
So based on this dataset, bigger discounts did not seem to result in more reviews.
I also looked at product price and ratings. High price products had an average rating of about 4.08, compared with 3.64 for low price products.
However, the correlation between price and rating was only about 0.11. So although there was a difference in the average ratings, the overall relationship was still weak.
This was a good reminder for me that seeing two things move together does not automatically mean there is a strong relationship between them.
Building the Dashboard
Once the cleaning and analysis was done, I used Excel PivotTables, PivotCharts and slicers to build the dashboard.
The dashboard brings together product ratings, reviews, discounts, prices, engagement and product rankings.
I also added slicers for rating category, discount category and price category so the data could be explored from different angles.
The main point was not just making something that looked nice. I wanted the dashboard to make it easier to look at the data and actually understand what was going on.
What I Learned
This project taught me that data analysis is not just about making charts.
A lot of the real work happens before you even get to the dashboard.
Cleaning the data, checking formulas, deciding how to handle missing values and questioning the results all turned out to be important.
I also became more careful about correlation. A correlation can tell you that two variables have some kind of relationship, but it does not mean that one caused the other.
Final Thoughts
I started with a messy spreadsheet and ended up with a cleaned dataset, analysis and an interactive Excel dashboard.
It was a really useful practical exercise and another step in my journey into data science.
I am still learning, and I know there is a lot more for me to learn, but projects like this are helping me understand how everything connects.
You clean the data, ask questions, analyse the results and then try to communicate what the data is actually saying.
And sometimes the hardest part of a data project is not building the dashboard.
It is making sure you can trust the data behind it.
The Project
You can view the complete project and documentation, on my GitHub repository Jumia Product Performance Dashboard
Top comments (0)