Introduction: Turning Jumia Product Data into Business Insights
E-commerce platforms generate a lot of product data, but raw numbers become useful only when they can support better decisions. For Jumia sellers, prices, discounts, ratings and customer reviews can provide clues about product performance, customer engagement and possible pricing strategies.
For this project, I worked with a dataset of 112 Jumia products to explore these relationships using Microsoft Excel. I wanted to find out whether higher discounts are associated with more customer reviews, whether highly rated products receive stronger engagement, and whether product price is related to rating. I also wanted to identify the products performing best and those that may require a different pricing or marketing approach. I followed a complete data-analysis workflow:
Raw Data → Cleaning → Transformation → Analysis → Visualization → Insights → Recommendations
The project uses Excel Tables, Power Query, formulas and functions, PivotTables, PivotCharts, slicers and dashboard techniques. This article documents that process and shows how the raw Jumia data was transformed into an interactive dashboard and, ultimately, evidence-based business recommendations.
Understanding the Dataset and Its Initial Problems
Before cleaning the data, I first needed to understand what I was working with. The dataset contains 112 Jumia products and six main fields: Product, Current Price, Old Price, Discount, Review and Rating. Current Price and Old Price represent product pricing, Discount captures the promotional percentage, Review represents the number of customer reviews, while Rating records the average customer rating out of 5. I treated this stage as a data-quality audit rather than immediately changing anything. The purpose was to identify issues that could affect calculations and visualizations later.
The raw dataset contained formatting and consistency issues that needed attention, particularly around numerical fields. Prices could contain currency symbols or unnecessary characters, while ratings and review counts needed to be standardized into usable numerical values. I also needed to check systematically for missing values, duplicates and other inconsistencies before beginning the analysis.
This preliminary inspection was important because reliable analysis depends on reliable input data. Instead of correcting problems as I encountered them later, I first established what needed to be fixed and why.
Data Cleaning and Preparation
The initial dataset contained 115 products, but some records were duplicates. The Balloon Insert product appeared three times, while the 6 Layers Steel Pipe Shoe Cabinet appeared twice. I removed the duplicate records so that each product was represented only once, leaving the required 112 products for analysis.
The Review column also contained negative values such as -2, -14 and -55. Since review counts cannot be negative, I removed the minus signs and converted the values into positive numbers. At the same time, 58 products had missing review counts and ratings, so I kept these as missing rather than incorrectly treating them as zero.
The rating field was stored as text, for example 4.5 out of 5, and the column itself was labelled Ratingd. I renamed the column to Ratings and extracted the numerical rating. Price values also contained KSh and commas, so I removed these characters and converted the prices into numerical values. One product used a price range rather than a single price, which required special handling before numerical analysis.
Data Enrichment
I added calculated columns to support deeper analysis. These additional fields transformed the original variables into useful measures and categories that could later be used in PivotTables, charts, slicers and the dashboard. I created a Discount Amount column to measure the absolute difference between the Old Price and Current Price. I used the Excel formula*=C2-B2*
I also created a Rating Category column to classify products as Poor, Average or Excellent according to the project criteria. I used the logical IF and AND functions with the formula =IF(G2="","Not Rated",IF(G2<3,"Poor",IF(G2<=4,"Good","Excellent"))) .
A Discount Category column was added to group products into Low Discount, Medium Discount and High Discount based on the required percentage thresholds. The formula used was =IF(E2<20%,"Low Discount",IF(E2<=40%,"Medium Discount","High Discount")).
Finally, I added a Price Category column to segment products into Low, Medium and High price groups using the formula =IF(B2<=1275,"Low",IF(B2<=2512,"Medium","High")) for the price thresholds established for the analysis. These enriched variables provided a stronger foundation for comparing product behaviour and developing the later seller-performance analysis.
Excel Techniques and Descriptive Analysis
After completing the cleaning and enrichment stages, I created a separate Analysis sheet to calculate the key descriptive statistics for the cleaned dataset. I started by determining the total number of products using the formula =COUNTA('Cleaned Data'!A2:A113), which confirmed that the cleaned dataset contained 112 products.
I then calculated the total number of reviews using the formula =SUM('Cleaned Data'!H2:H113), which added the available review counts while leaving missing values as blanks rather than treating them as zero.
To understand the overall pricing level, I calculated the** average current price** using =AVERAGE('Cleaned Data'!B2:B113) and the average old price using =AVERAGE('Cleaned Data'!C2:C113).
I also calculated the average discount using =AVERAGE('Cleaned Data'!E2:E113) and the average rating using =AVERAGE('Cleaned Data'!G2:G113).
To identify the most expensive product, I first used =MAX('Cleaned Data'!B2:B113) to find the highest Current Price, then used** =INDEX('Cleaned Data'!A2:A113,MATCH(MAX('Cleaned Data'!B2:B113),'Cleaned Data'!B2:B113,0))** to return the product associated with that price.
I followed the same approach for the least expensive product, using =MIN('Cleaned Data'!B2:B113) to identify the lowest Current Price and =INDEX('Cleaned Data'!A2:A113,MATCH(MIN('Cleaned Data'!B2:B113),'Cleaned Data'!B2:B113,0)) to return the corresponding product name.
These calculations gave me the initial numerical summary of the dataset and established a foundation for investigating relationships between pricing, discounts, customer reviews and ratings.
Correlation Analysis
To investigate whether pricing, discounts and customer engagement were related, I used correlation analysis. Correlation measures the direction and strength of a linear relationship between two variables, with values ranging from -1 to +1. A value close to +1 indicates a strong positive relationship, a value close to -1 indicates a strong negative relationship, while a value close to 0 indicates little or no linear relationship. I conducted three correlation tests using the Excel **CORREL **function.
Discount vs. Reviews
The first test examined whether products offering higher discounts received more customer reviews. I used the Excel formula =CORREL('Cleaned Data'!E2:E113,'Cleaned Data'!H2:H113). The resulting correlation coefficient was -0.14, indicating a very weak negative relationship. In the Jumia dataset, this means that products with higher discounts tended to have slightly fewer reviews, but the relationship was too weak to indicate a meaningful pattern. Therefore, the data does not support the assumption that increasing a product's discount automatically leads to greater customer engagement through reviews.
Rating vs. Reviews
The second test investigated whether products with higher customer ratings also attracted more reviews. I used the formula =CORREL('Cleaned Data'!I2:I113,'Cleaned Data'!H2:H113). The result was 0.06, which is very close to zero and indicates an almost nonexistent linear relationship. Within this dataset, highly rated products were therefore not necessarily the products with the highest number of reviews. This suggests that customer satisfaction, represented by rating, and customer engagement, represented by review volume, were largely independent of each other among the products analysed.
Current Price vs. Rating
The final test examined whether the price of a product was associated with its customer rating. I used the formula =CORREL('Cleaned Data'!B2:B113,'Cleaned Data'!I2:I113). The resulting coefficient was 0.11, showing a very weak positive relationship. This indicates that more expensive products tended to have slightly higher ratings, but the relationship was not strong enough to suggest that price was an important predictor of customer satisfaction. In this dataset, a higher-priced product was therefore not necessarily better rated than a lower-priced product.
Top 5 Highest and Lowest Rated Products
To identify the products with the strongest and weakest customer ratings, I used Excel's LARGE and SMALL functions. These functions were appropriate because they allowed me to rank the numerical ratings and identify the highest and lowest values directly from the cleaned dataset.
For the top 5 highest-rated products, I first used =LARGE('Cleaned Data'!G2:G113,1) to identify the highest rating, which was 5. I then used the FILTER function to retrieve the products associated with this rating. Five products achieved the maximum rating of 5, making them the highest-rated products in the dataset and potential indicators of strong customer satisfaction.
For the top 5 lowest-rated products, I used the SMALL function to rank the ratings from lowest upward. The five lowest ratings was 2.0 . I then used FILTER to identify the products corresponding to these ratings. These products represent the weakest-rated products in the dataset and may require further investigation into product quality, customer expectations or seller performance.
Top 10 Products with the Highest Discounts
To identify the products receiving the highest discounts, I ranked the discount percentages from highest to lowest using Excel's LARGE function. I used =LARGE('Cleaned Data'!D2:D113,1) to identify the highest discount and adjusted the ranking number to retrieve the top 10 values. I then used FILTER to identify the products corresponding to these discounts. This showed which products were being promoted most heavily and provided a basis for comparing high discounts with reviews and ratings.
Top 10 Products by Customer Reviews
To identify the products with the strongest customer engagement, I ranked the review counts from highest to lowest using Excel's LARGE function. I used =LARGE('Cleaned Data'!H2:H113,1) to find the highest review count and then ranked the remaining values from 2 to 10. I used XLOOKUP to match each review count to its corresponding product with =XLOOKUP(F61,'Cleaned Data'!H2:H113,'Cleaned Data'!A2:A113).
The highest review count was 69, and the Top 10 results identified the products receiving the most customer reviews. These products represent the strongest customer engagement in the dataset and can later be compared with their ratings, prices and discount levels to understand what may be associated with stronger customer interaction.
High Discounts with Low Ratings
I also investigated products that were receiving high discounts but poor customer ratings. I used the FILTER function with two conditions: a discount above 40% and a rating below 3. The formula was** =FILTER('Cleaned Data'!A2:A113,('Cleaned Data'!F2:F113>40%)('Cleaned Data'!I2:I113<3),"No matching products")*. This was useful because it highlighted products where heavy price reductions were not translating into strong customer satisfaction, providing potential areas for sellers to review product quality, pricing or marketing strategies.
Products Requiring Better Pricing or Marketing Strategies
I then identified products that may require a different pricing or marketing strategy by combining three indicators: poor ratings, low review activity and discounts above 20%. I used the FILTER function with the formula =FILTER('Cleaned Data'!A2:A113,('Cleaned Data'!I2:I113<3)*('Cleaned Data'!H2:H113<5)*('Cleaned Data'!F2:F113>20%),"No matching products"). I chose these conditions because they highlight products that have low customer satisfaction, limited engagement and are already being discounted. The analysis returned several products, suggesting that discounting alone may not be sufficient for this group and that sellers may need to review their pricing, product quality, positioning or marketing approach.
Key Insights
The Jumia dataset reveals that customer response is more complicated than simply offering a cheaper price. Despite discounts reaching as high as 64%, the relationship between discounts and reviews was almost nonexistent (-0.14). This suggests that aggressive promotions may attract attention, but they do not automatically create lasting customer engagement.
A particularly interesting pattern is the separation between popularity and satisfaction. With a rating-review correlation of only 0.06, products generating reviews are not necessarily the products customers rate highly. This means a seller could have a product that attracts considerable attention while still failing to deliver a satisfying experience.
The data also shows that price is a poor substitute for perceived value. The price-rating correlation of 0.11 indicates that paying more does not guarantee a better customer experience. At the same time, the presence of products rated 5.0 alongside products rated as low as 2.0 suggests that product-level factors can create much larger differences in customer satisfaction than price alone.
Business Recommendations
Turn discounts into a targeted tool, not a rescue strategy. Sellers should use large discounts strategically on products that already demonstrate good customer response, rather than repeatedly discounting poorly rated products. A product with a 64% discount and weak customer feedback should trigger investigation before another promotion.
Build around the products customers already love. The five products achieving 5.0 ratings provide potential benchmarks. Sellers should examine what makes these products successful such as product quality, presentation, value or reliability and replicate those strengths across weaker products.
Treat poor ratings as a product problem first. Products falling around 2.0–2.2 should be reviewed for quality issues, misleading descriptions, poor fulfilment or unmet customer expectations. Reducing their price further may increase sales temporarily but could worsen customer dissatisfaction if the underlying problem remains.
Use different strategies for different products. The absence of products with more than 40 reviews and average ratings of 3–4 shows that highly engaged products in this dataset tend to fall outside that specific middle-rating segment. Sellers should therefore segment products by both engagement and satisfaction instead of applying one marketing strategy to the entire catalogue.
Conclusion
Working on the Jumia dataset was more interesting than I expected. At the beginning, I thought I would simply clean the data, apply a few Excel formulas, create some charts and be done. But as I worked through the dataset, I started noticing things that made me question my own assumptions. I expected bigger discounts to attract more customers and expensive products to have better ratings, but the data did not always support that. That was probably the biggest surprise for me.
I also realised how much the quality of the data matters. Removing duplicates, organising the information and creating useful categories may seem like small steps, but they made a big difference in how confidently I could interpret the results. I started seeing Excel differently too. It was no longer just about knowing which formula to use; it was about knowing what question I was trying to answer and what the result was actually telling me.
The part I found most interesting was watching ordinary rows of data turn into something that could tell a business story. A rating represents a customer's experience, a review shows engagement, a discount reflects a seller's strategy, and the price gives another piece of the picture. Putting all these pieces together made the dataset feel much more real.
Top comments (0)