Building an Interactive Excel Dashboard for E-commerce Product Analysis

A Case Study of Jumia Products

When managing an e-commerce retail operation, offering product discounts may seem like the easiest way to attract customer attention. But does a larger discount actually translate into greater customer engagement?

To explore this question, I built an end-to-end Excel analytics pipeline using a dataset of 109 Jumia products.

The project took me from messy raw data through cleaning, data enrichment, PivotTables, PivotCharts, KPIs, and finally an interactive dashboard with connected slicers.

Along the way, I uncovered an interesting discount-and-engagement pattern and, more importantly, encountered a subtle logic trap in Excel's IFS function.

The Data Pipeline

The raw Jumia data contained several structural and formatting problems.

Some price fields contained text such as KSh, ratings required parsing, duplicate records were present, and some review values contained negative signs.

I used Excel's Text to Columns, Find and Replace, conditional formatting, and the built-in Remove Duplicates functionality to clean the dataset.

The documented cleaning process included:

• Removing 6 duplicate records
• Removing KSh text from price fields
• Cleaning the rating field
• Correcting column data types
• Addressing negative review values
• Cleaning price-range values
• Creating calculated analytical fields

The cleaning process transformed the raw dataset into a structured table for further analysis.

The IFS Logic Trap

Data enrichment introduced one of the most interesting problems in the project.

I initially used the following formula to classify product ratings:

=IFS(
    Table1[@rating]<=3,"Poor",
    Table1[@rating]<=4,"Average",
    Table1[@rating]>4,"Excellent"
)

At first glance, the formula looked reasonable.

But consider a product with a rating of 4.2.

It does not satisfy:

rating <= 3

and it also does not satisfy:

rating <= 4

It therefore reaches:

rating > 4

and is classified as Excellent.

The result was a classification boundary that did not match the intended rating categories.

Fixing the logic

This was a useful reminder that formulas can be syntactically correct while still producing analytically misleading results.

I revised the rating categorization logic so that the boundaries were explicitly defined and the categories covered the intended rating ranges without unintended gaps.

This was particularly important because the rating categories were later used in the dashboard's slicers and PivotTable analysis.

Calculating Discount Performance

I also created additional fields to make the analysis more reliable.

Discount Amount

=old_price-current_price

Calculated Discount Percentage

=discount_amount/old_price

The calculated discount percentage was used because the original discount field contained discrepancies.

I then created discount categories:

• Low: below 20%
• Medium: 20%–40%
• High: above 40%

I also created price categories:

• Inexpensive: up to KSh 500
• Moderate: KSh 501–1,500
• Expensive: above KSh 1,500

Does a Bigger Discount Mean More Engagement?

This was one of the questions I wanted the dashboard to help answer.

The correlation between the calculated discount percentage and review count was approximately

-0.026

That's extremely close to zero.

In this dataset, there is therefore little evidence of a linear relationship between discount percentage and review count.

In other words, simply offering a larger percentage discount was not strongly associated with having more customer reviews.

The Mid-Discount Engagement Pattern

The PivotTables revealed another interesting descriptive pattern.

Products in the Medium discount category (20%–40%) recorded the highest average review count at approximately 12.9 reviews per product.

The High discount category, representing discounts above 40%, averaged approximately 9.1 reviews per product.

This creates an interesting pattern:

Discount category	Average reviews
Low	—
Medium	12.9
High	9.1

The pattern suggests that the relationship between discount percentage and review engagement is not simply:

Bigger discount → more engagement.

Instead, the medium-discount group showed the highest average review count in this dataset.

An Interesting Product-Level Finding

The product with the highest review count was:

120W Cordless Vacuum Cleaners Handheld Electric Vacuum Cleaner

It recorded:

• 69 reviews
• 2.8 rating

This combination is particularly interesting because high review volume does not necessarily correspond to high product ratings.

Building the Dashboard

After completing the cleaning and analysis, I consolidated the results into an interactive Excel dashboard.

The dashboard combines:

• KPI cards
• Product-review analysis
• Price-category analysis
• Discount-category analysis
• Rating categories
• PivotCharts
• Interactive slicers

The slicers allow users to filter the analysis dynamically by categories such as:

• Price
• Discount
• Rating

This transformed the workbook from a collection of calculations into an interactive analytical tool.

Insights

Based on this dataset:

1. Larger discounts were not strongly associated with higher review counts.
2. The medium-discount category recorded the highest average review count.
3. High-review products can still have relatively low ratings, making review volume alone an incomplete measure of product performance.
4. Price and discount segmentation can provide useful ways to explore product behavior.
5. Review count should be treated as an engagement indicator rather than a direct measure of sales demand.