Introduction
E-business platforms generate large amounts of product data, including prices, discounts, ratings, and customer reviews. When properly analyzed, this information can help sellers and businesses understand product performance and identify areas that may require attention.
For this project, I built an interactive Excel dashboard for Jumia product analysis. The project focused on transforming raw product data into a structured dataset, performing exploratory analysis, and presenting the results through an interactive dashboard.
The main objective was to investigate relationships between price, discount, rating, and customer reviews, while identifying products with strong or weak performance.
1. Project Objective
The main objective was to build an interactive Excel dashboard that could answer questions about Jumia product performance.
Specifically, I wanted to understand:
• Whether larger discounts were associated with more customer reviews.
• Whether highly rated products attracted stronger review engagement.
• Whether product price and rating appeared to move together.
• Which products had the highest and lowest ratings.
• Which products had the highest discounts.
• Which products generated the most reviews.
• Which products might require additional pricing, marketing, or product-quality attention.
An important consideration throughout the project was that reviews were treated as a proxy for customer engagement, not as a measure of sales or revenue.
2. Dataset and Business Questions
The dataset contained information about 112 Jumia products.
The analysis focused on product-level variables such as:
• Product
• Current Price
• Old Price
• Discount
• Rating
• Reviews
The business questions guiding the analysis were:
- Does a higher discount correspond to more reviews?
- Do highly rated products receive more reviews?
- Is there a relationship between current price and rating?
- Which products have the highest ratings?
- Which products have the lowest ratings?
- Which products have the highest discounts?
- Which products have the highest number of reviews?
- Are there products with high discounts but relatively weak ratings or engagement?
These questions helped determine which calculations, PivotTables, and visualizations were needed.
3. Initial Data-Quality Audit
Before making changes to the dataset, I performed a data-quality audit to understand the condition of the raw data.
The audit focused on:
• Number of rows and columns=116
• Blank or missing values =55 each for two columns
• Duplicate records =2
• Price ranges =2
• Misspelled column header=1
• Improperly written product names=1
One of the most important findings from the audit was the presence of missing ratings.
Out of the 112 products, 55 products had no recorded rating and review, meaning that approximately 49.1% of the products lacked rating and review information.
This was important because rating-based and review-based analysis could only be performed reliably on the products with available ratings.
The original dataset was preserved rather than overwritten so that the cleaning process remained traceable.
4. Cleaning and Preparation Decisions
The raw data was kept separate from the cleaned dataset. This ensured that the original data remained available for comparison and verification.
The cleaned dataset was then prepared for analysis by:
• Checking and correcting data types.
• Handling missing values where appropriate.
• Ensuring prices were treated as numerical values.
• Ensuring discount values were represented consistently as percentages.
• Checking rating and review fields.
• Creating additional analytical fields.
• Converting the cleaned range into an Excel Table for easier reference and analysis.
The cleaned table was named cleaned
5. Excel Formulas and Enrichment Fields
After cleaning the dataset, I created additional fields to support the analysis. Some of these tables were:
Discount Amount
The discount was calculated from the difference between the old price and current price.
For example:
= [Old Price]- [Current Price]]
Rating Categories
Products were also classified according to their rating to make it easier to analyze rating quality.
The categories were designed to distinguish products with poor, average, and excellent ratings.
For example:
=IF([Rating]="","Missing”, IF([@Rating] <3,"Poor”, IF([@Rating] <=4.5,"Average","Excellent")))
This classification made it easier to summarize rating performance using PivotTables.
These enrichment fields transformed the original variables into categories that could be used for business-oriented analysis.
6. PivotTable and Analysis Workflow
After preparing the cleaned dataset, I created PivotTables to summarize the data. This was vital because some of these tables would bring about slicers and graphs that would be used to represent data on the dashboard.
The PivotTables were used to investigate:
• Distribution Rating.
• Discount distribution.
• Price versus rating.
• Engagement by discount category.
• Top products by rating.
• Bottom products by rating.
• Top products by discount.
• Top products by reviews.
7. Relationship Analysis
To investigate relationships between numerical variables, I created scatter plots with trendlines.
The three relationships analyzed were:
Discount vs Reviews
The calculated correlation for this was 0.0440, which indicated a very weak relationship.
The analysis therefore did not suggest that discount percentage strongly explained differences in review engagement.
Rating vs Reviews
The calculated correlation was 0.0572, indicating a very weak positive relationship.
This suggested that rating alone did not strongly explain differences in review engagement.
Current Price vs Rating
The calculated correlation was 0.1101, indicating a very weak relationship.
The analysis therefore provided limited evidence of a meaningful relationship between current price and rating.
These analyses however, reinforced an important lesson, that a single variable should not automatically be treated as the main driver of product performance.
8. Dashboard Design and Slicer Connections
The final stage of the project was to transform the analysis into an interactive dashboard. At the topmost of the Dashboard was the heading/header written in bold formart and a large font size. Beneath it was 5 key performance indicators commonly known as KPI's. They included
• Total Products: 112
• Average Current Price: Ksh 1,186.89
• Average Discount: 0.365089286
• Average Rating: 3.889474
• Total Reviews: 723
Beneath this KPI's were three slicers that allowed the dashboard to be filtered interactively. To do these it was paramount to Report connection by right clicking while your cursor is on a slicer and choosing "Report connection" , The slicers used were:
Price Category-it categorized data into High price, Low price and Medium Price.
Discount Category-it categorized data into High discount, Low discount, and Medium discount
Rating Category- that categorized data into average, excellent, poor and missing for missing values.
Charts were located beneath these slicers and used to communicate patterns in the data.
The dashboard was designed to allow users to move from an overall view of product performance to more specific product-level analysis.
Besides the chart I created a recommendation table that would inform the Jumia sellers.
It is also important to note that when building a dashboard, Presentability and neatness are key, the colors used in the graphs, dashboard background and headings must be consistent.
9. Key Findings
High-Engagement Products
The top 10 products by reviews accounted for 388 of the 723 total reviews, representing approximately 53.7% of all reviews in the dataset.
The 120W Cordless Vacuum Cleaner had the highest review count with 69 reviews, followed by the 137 Pieces Cake Decorating Tool Set with 55 reviews.
This shows that review engagement was concentrated among a relatively small group of products.
Top-Rated Products
The top 10 products by rating had an average rating of 4.94/5.
Seven of these products had a rating of 5.0, while the remaining three had ratings of 4.8.
These products could provide useful benchmarks for understanding characteristics associated with positive customer experiences.
Low-Rated Products
The bottom five products by rating had an average rating of 2.12/5, compared with the overall dataset average of 3.89/5.
The lowest-rated product was the Wall-Mounted Sticker Punch-Free Plug Fixer, with a rating of 2.0/5.
These products may require investigation before additional marketing or promotional resources are allocated to them.
Highly Discounted Products
The top 10 products by discount had an average discount of 54%.
The highest individual discount was 61%, recorded for the Pen Grips for Kids Pen Grip Posture Correction Tool for Kids.
However, the relationship analysis showed that discount percentage had only a very weak relationship with reviews. Therefore, larger discounts should not automatically be assumed to produce stronger engagement.
Screenshot of original data, cleaned data, pivot tables, and Dashboard.
To to attach a face to our long article I have uploded the four most vital images that were generated in this analysis but most of all a view of the dashboard that was created.
a) The original data
b) The Cleaned Data
c) Pivot tables
d) Dashboard
10. Business Recommendations
Based on the analysis, several recommendations can be made.
1. Prioritize High-Engagement Products
Since the top 10 products accounted for approximately 53.7% of all reviews, sellers could identify what these products have in common and prioritize similar products for visibility, promotion, and inventory planning.
2. Test Targeted Discounts
Medium-discount products recorded the highest total reviews at 348. However, the correlation between discount and reviews was very weak.
Rather than relying on large discounts across all products, sellers should test targeted promotional strategies and evaluate their results using multiple performance indicators.
3. Investigate Low-Rated Products
Products with consistently low ratings should be investigated before receiving additional promotional investment.
Possible areas for investigation include product quality, product descriptions, customer expectations, and fulfillment experience.
4. Use Highly Rated Products as BenchmarksThe highest-rated products can be examined to identify practices that may contribute to positive customer experiences.
These could include product quality, accurate descriptions, competitive pricing, and reliable fulfillment.
5. Use Multiple Metrics for Decision-Making
Because the relationships between price, discount, rating, and reviews were weak, sellers should avoid relying on a single metric when making business decisions.
A broader analysis incorporating sales, revenue, listing age, customer feedback, and other product-level factors would provide stronger evidence.
11. Limitations
This analysis has several limitations.
- No sales or revenue data: The dataset does not contain actual sales, revenue, profit, or conversion data. Reviews were therefore used only as a proxy for customer engagement.
- Missing ratings: 55 of the 112 products, or 49.1%, had no recorded rating. This limits rating-based comparisons.
- No listing-age information: The dataset does not indicate how long each product had been available. Older listings may naturally have more reviews.
- Correlation does not imply causation: The correlation analysis identifies relationships between variables but cannot establish that discounts, prices, or ratings directly cause changes in review engagement.
- Limited customer information: The dataset does not contain detailed review comments, product-quality measures, seller information, or delivery performance.
- Dataset size: The analysis covers 112 products and therefore should not automatically be generalized to every product on Jumia or the wider e-commerce market.
Conclusion
The Jumia Product Performance Dashboard transformed raw e-commerce product data into an interactive analytical tool covering 112 products.
The analysis showed that discount, rating, and current price had only weak relationships with the available review data. At the product level, however, engagement was concentrated among a smaller group of products, while a separate group of products had notably low ratings.
At the same time, the limitations of the dataset highlighted the importance of combining reviews with additional business information such as sales, revenue, listing age, and customer feedback. Additionally, the importance of having a good dashboard in a business was seen.
Project Resources
The complete project files are available in my GitHub repository and include original data, cleaned dataset, analysis, PivotTables, dashboard, and supporting documentation.
GitHub Repository link
https://github.com/gee-999/JUMIA-PRODUCT-PERFORMANCE-PROJECT
Top comments (0)