Introduction
Data is one of the most important resources for modern e-commerce businesses. Online marketplaces such as Jumia generate large amounts of product information, including prices, discounts, customer ratings, and reviews. However, collecting data is only the first step. The real value comes from transforming raw data into useful information that can support better business decisions.
In this project, I used Microsoft Excel to clean, analyze, and visualize a dataset of Jumia products. The goal was to develop an interactive dashboard that could help sellers and decision-makers understand how pricing, discounts, customer ratings, and customer engagement relate to product performance.
The project was designed not simply as a chart-building exercise, but as a complete data-analysis workflow. It involved preserving the original data, identifying data-quality problems, cleaning and transforming the dataset, creating calculated fields, performing statistical analysis, building PivotTables, and finally developing an interactive dashboard.
The analysis focused on several important business questions:
- Are larger discounts associated with more customer reviews?
- Do highly rated products attract stronger customer engagement?
- Is there a relationship between product price and rating?
- Which products perform best based on ratings and reviews?
- Which products may require a different pricing or marketing strategy?
One important limitation is that the dataset does not contain units sold or revenue. Therefore, customer reviews were treated as an engagement proxy, not as a direct measure of sales.
Project Objective
The main objective was to transform Jumia product data into an interactive Excel dashboard that could provide useful insights into pricing, promotions, ratings, and customer engagement.
The project followed a structured process:
- Import and preserve the raw data.
- Audit the dataset for quality issues.
- Clean prices, discounts, reviews, and ratings.
- Handle missing values and duplicates.
- Create additional analytical fields.
- Calculate descriptive statistics and correlations.
- Build PivotTables and PivotCharts.
- Create an interactive dashboard using slicers.
- Interpret the findings and develop evidence-based recommendations.
This approach helped demonstrate how Excel can be used as a practical data-analysis tool from raw data through to business recommendations.
Understanding the Dataset
The dataset contained information about Jumia products, including:
- Product – the product name.
- Current Price – the current selling price in Kenyan shillings.
- Old Price – the price before the advertised discount.
- Discount – the advertised percentage discount.
- Review – the number of customer reviews.
-
Rating – the average customer rating out of five,
as shown in the photo.
The project guide highlighted several data-quality issues that needed attention. The original rating column was named Ratingd, some review and rating values were blank, review counts appeared with negative signs, one product contained a price range, and the source contained repeated rows.
These issues demonstrated why data cleaning is an essential part of the analytical process.
Initial Data-Quality Audit
Before changing the data, I first investigated the quality of the source information.
The audit focused on:
- Missing values.
- Duplicate records.
- Nonnumeric prices.
- Nonnumeric discounts.
- Nonnumeric reviews.
- Nonnumeric ratings.
- Ratings outside the 0–5 range.
- Discounts outside the 0%–100% range.
- Negative review counts.
- Cases where current price was greater than old price.
- Price ranges and other ambiguous values.
This process ensured that cleaning decisions were based on evidence rather than assumptions.
The project instructions also emphasized that the original Raw_Data sheet should remain unchanged, while cleaning and transformations should be performed in a separate Cleaned_Data sheet.
Data Cleaning and Preparation
Cleaning Product Names
Product names were standardized using trimming and cleaning techniques to remove unnecessary spaces and unwanted characters while preserving meaningful punctuation.
Cleaning Prices
The price fields contained values such as:
KSh 2,880
These values needed to be converted into numeric values by removing KSh, commas, and unnecessary spaces.
A formula such as the following can be used for standard price values:
=VALUE(SUBSTITUTE(SUBSTITUTE(B2,"KSh",""),",",""))
This converts the text representation into a numeric value that Excel can use for calculations.
Handling Price Ranges
One of the challenges was a product whose price was represented as a range rather than a single number.
Instead of deleting the record, the project used a midpoint approach. The minimum and maximum values were identified and the midpoint was calculated.
For example:
Midpoint = (Minimum Price + Maximum Price) / 2
The original value was retained in the raw data so that the cleaning decision remained traceable.
Cleaning Discounts
Discount values were converted into numeric percentages.
For values containing %, the percentage symbol was removed and the result was divided by 100 where necessary.
For example:
=VALUE(SUBSTITUTE(D2,"%",""))/100
The resulting values could then be formatted as percentages in Excel.
Cleaning Reviews
Some review values contained negative signs. Since a review count cannot logically be negative, these values were treated as a likely scraping or formatting artifact.
The absolute value was therefore used:
=IF(E2="","",ABS(VALUE(E2)))
Importantly, genuinely blank review values were not automatically changed to zero. A missing value and zero reviews are not necessarily the same thing.
Cleaning Ratings
Ratings were recorded in formats such as:
4.5 out of 5
The unnecessary text was removed and the remaining value was converted into a number.
=IF(F2="","",VALUE(SUBSTITUTE(F2," out of 5","")))
Blank ratings were left missing rather than replacing them with an average value. The project instructions specifically recommend keeping missing values visible because automatically imputing them can distort the analysis.
Removing Duplicates
Duplicate records were investigated rather than removing every product with the same name.
The project required duplicates to be identified using the relevant source fields. Products with the same name but different prices or feedback could represent legitimate separate listings.
After exact duplicate removal, the completed analysis contained 112 products.
Creating Calculated Fields
Once the data had been cleaned, additional columns were created to support the analysis.
Discount Amount
The actual monetary discount was calculated as:
=[@[Old Price]]-[@[Current Price]]
This provides the difference between the original price and the current selling price.
Rating Category
Products were grouped into rating categories.
The working classification used in the project was:
- Poor: below 3
- Average: 3 to 4.5
- Excellent: above 4.5
The Excel formula was:
=IF([@Rating]="","Missing",IF([@Rating]<3,"Poor",IF([@Rating]<=4.5,"Average","Excellent")))
This categorization was documented because the original project brief left a gap between some rating boundaries.
Discount Category
Discounts were grouped into:
- Low Discount: below 20%
- Medium Discount: 20%–40%
- High Discount: above 40%
The formula used was:
=IF([@Discount]="","Missing",IF([@Discount]<20%,"Low Discount",IF([@Discount]<=40%,"Medium Discount","High Discount")))
Price Category
Rather than choosing arbitrary price boundaries, quartiles were used.
The first quartile was calculated with:
=QUARTILE.INC(tblProducts[Current Price],1)
The third quartile was calculated with:
=QUARTILE.INC(tblProducts[Current Price],3)
Products were then classified as low, medium, or high price based on these thresholds.
This approach makes the categories reproducible and data-driven.
Engagement Flag
Customer reviews were used as a proxy for engagement because the dataset did not contain sales information.
The 75th percentile of reviews was used as the engagement threshold. In the completed workbook, this threshold was 14 reviews. Products meeting or exceeding that level were considered to have strong engagement.
Descriptive Analysis
After cleaning and enrichment, the dataset was analyzed using descriptive statistics.
The completed analysis showed:
| Metric | Result |
|---|---|
| Total products | 112 |
| Average rating | 3.89 / 5 |
| Total valid reviews | 723 |
| Average advertised discount | 36.78% |
| 75th percentile review threshold | 14 |
Here is a clear photo to support above to statement
These metrics provide a high-level overview of the dataset.
An average rating of approximately 3.89 out of 5 suggests that the products generally received reasonably positive customer feedback, although there was still variation between products.
The average advertised discount was approximately 36.78%, indicating that promotional pricing was an important feature of the dataset.
The total of 723 valid reviews also shows that customer feedback provides a useful engagement measure, although it should not be interpreted as a measure of sales.
Correlation Analysis
Correlation analysis was used to investigate relationships between important variables.
The analysis produced three key correlation values:
- Discount vs Reviews: -0.137
- Rating vs Reviews: 0.057
- Price vs Rating: -0.597
Discount and Reviews
The correlation between discount and reviews was approximately -0.137.
This indicates a weak negative relationship between the advertised discount and number of reviews in this dataset.
In practical terms, larger discounts were not associated with substantially higher review counts. This suggests that simply increasing a discount may not automatically result in stronger customer engagement.
Rating and Reviews
The correlation between rating and reviews was approximately 0.057.
This is a very weak positive relationship.
It suggests that products with higher ratings did not necessarily have dramatically more reviews. A product can therefore have an excellent rating without having a large amount of customer engagement.
This distinction is important because rating quality and engagement volume represent different dimensions of product performance.
Price and Rating
The strongest relationship observed was between price and rating, with a correlation of approximately -0.597.
This represents a moderate-to-strong negative relationship in this dataset. In other words, higher-priced products tended to be associated with lower ratings.
However, correlation does not prove causation. Other factors, such as product type, quality expectations, brand, listing quality, or customer demographics, could influence this relationship.
PivotTables and Dashboard Development
After completing the calculations, PivotTables were created to summarize the cleaned data.
PivotTables made it possible to analyze products by categories such as:
- Price category.
- Discount category.
- Rating category.
- Engagement level.
The summarized data was then used to create charts and dashboard components.
The final dashboard was designed to provide a quick overview of product performance while allowing users to interact with the data.
Slicers were incorporated so users could filter the analysis by relevant categories without manually changing the underlying PivotTables. this photo shows
The project requirements emphasize that slicers should control the intended PivotCharts and KPIs and that charts should have clear titles, readable labels, appropriate units, and suitable formats.
Dashboard Design
The dashboard brings the analysis together in one place.
Important KPI elements include:
- Total products.
- Average current price.
- Average old price.
- Average discount.
- Average rating.
- Total reviews.
- Lowest price.
Charts can then be used to communicate patterns across pricing, discounts, ratings, and engagement.
The purpose of the dashboard is not simply to make the workbook visually attractive. Instead, it provides a way for users to move from raw numbers to business questions and decisions.
For example, a seller can use the dashboard to identify products with high engagement, compare discount categories, and investigate products that have strong ratings but relatively low review counts.
Key Findings
Several important findings emerged from the analysis.
1. Discounts did not strongly predict engagement
The weak negative discount-review correlation of approximately -0.137 suggests that higher discounts were not strongly associated with higher review counts.
This means sellers should be careful about assuming that larger discounts automatically produce greater customer engagement.
2. High ratings did not necessarily mean high engagement
The rating-review correlation was approximately 0.057, indicating almost no linear relationship between rating and number of reviews.
A product can therefore have a high rating but relatively few reviews.
3. Price and rating showed the strongest relationship
The price-rating correlation of approximately -0.597 was the strongest relationship identified.
Higher-priced products tended to have lower ratings in this dataset. This is an important signal that could justify further investigation into product quality, customer expectations, value for money, or product categories.
4. Reviews are useful as an engagement proxy
Because sales and revenue were unavailable, reviews provided a useful way of comparing customer engagement.
However, reviews should not be described as sales. The project instructions specifically warn against treating reviews as units sold or claiming that relationships in the data prove causation.
Business Recommendations
Based on the analysis, several recommendations can be made.
Test discount strategies rather than simply increasing discounts
Since the discount-review relationship was weak, sellers should test different discount levels rather than assuming that larger discounts will automatically generate more engagement.
Different products may respond differently to promotions.
Investigate highly priced, poorly rated products
The negative price-rating relationship deserves further investigation.
Products with relatively high prices and low ratings could be reviewed for:
- Product quality.
- Customer expectations.
- Product descriptions.
- Value for money.
- Customer complaints.
Promote products combining strong ratings and engagement
Products that combine excellent ratings with strong review counts may represent particularly attractive listings.
These products could be considered for additional promotional visibility, subject to business objectives and profitability.
Improve listings with weak engagement
Products with reasonable ratings but low review counts could benefit from improvements to product descriptions, images, visibility, and promotional strategies.
However, these actions should be tested rather than assumed to produce a guaranteed increase in sales.
Lessons Learned
This project demonstrated that data analysis involves much more than creating charts.
One of the most important lessons was the importance of data quality. If prices, ratings, discounts, or reviews are stored as text, calculations and PivotTables can produce misleading results.
I also learned the importance of preserving raw data. Keeping the original dataset unchanged makes the cleaning process traceable and allows results to be checked against the source.
Another important lesson was the value of calculated fields. Fields such as Discount Amount, Rating Category, Discount Category, Price Category, and Engagement Flag made it easier to transform raw information into meaningful business measures.
Finally, the project demonstrated the importance of communicating analytical findings responsibly. A correlation can identify a relationship, but it does not automatically explain why that relationship exists.
Conclusion
Building the Jumia Product Performance Dashboard provided a practical example of how Microsoft Excel can be used for an end-to-end data-analysis project.
The process began with raw e-commerce data and progressed through data-quality auditing, cleaning, transformation, statistical analysis, PivotTables, and dashboard development.
The analysis found a weak relationship between discounts and reviews, almost no linear relationship between ratings and reviews, and a considerably stronger negative relationship between price and rating.
These findings demonstrate why data-driven decision-making is valuable in e-commerce. Instead of relying solely on assumptions such as "bigger discounts always attract more customers," sellers can use evidence to identify patterns and design better strategies.
The completed dashboard provides an interactive way to explore these relationships and turn the results into actionable business questions.
Most importantly, the project reinforced a fundamental principle of data analytics: good decisions depend on good data, careful analysis, and responsible interpretation.
Project Structure
The completed project follows a structure that separates the raw data, cleaned data, analysis, PivotTables, dashboard, and documentation. The recommended GitHub structure is:
jumia-product-performance-dashboard/
├── README.md
├── data/
│ └── Excel_jumia_dataset.csv
├── dashboard/
│ └── jumia_product_dashboard.xlsx
└── images/
├── raw-data.png
├── cleaned-data.png
├── pivot-tables.png
└── dashboard.png
Top comments (0)