Building a Jumia Product Performance Dashboard in Excel.

An interactive Excel dashboard analyzing pricing, discounts, and customer engagement across 112 Jumia products — built to uncover whether promotions, ratings, and price actually drive customer engagement, and which products need a different strategy.

Project Objective

Turn a raw Jumia product export into a working analysis that answers:

1. Are larger discounts associated with more reviews?
2. Do highly rated products attract stronger engagement?
3. Do price and rating move together?
4. Which products perform best based on ratings and reviews?
5. Which products may need a different pricing or marketing strategy?

Data Source

Raw product data scraped from Jumia (Kenya's e-commerce platform), covering 115 products with product name, current price, old price, discount percentage, review count, and star rating. The raw export contained several data quality issues typical of scraped e-commerce data — see Cleaning Process below.

Tools Used

• Microsoft Excel — data cleaning, formulas, PivotTables, PivotCharts, slicers, interactive dashboard

Cleaning Process

The raw data had several issues that needed fixing before analysis:

• Prices stored as text (e.g. KSh 1,525) — stripped the currency prefix and thousands separators, converted to numeric with VALUE(SUBSTITUTE(...)).
• Discount stored as text percentages (e.g. 38%) — converted to a true decimal (0.38).
• Review counts were all negative (a scraping artifact) — corrected with ABS().
• Rating was stored as descriptive text ("4.5 out of 5") — extracted the numeric value with LEFT() and FIND().
• A column header typo (Ratingd) was corrected to Rating.
• 3 fully blank rows were removed.
• 55 of 112 products had no Review or Rating data — left blank rather than filled with 0.

Key Findings

1. Discount size does not reliably predict review count. Average reviews across discount bands bounce around with no consistent upward trend.

2. Higher ratings don't guarantee stronger engagement. Products rated 2.5–3.0 stars actually average more reviews (24.6) than products rated 4.5–5.0 stars (14.4).

3. Price and rating are essentially unrelated. Average rating stays flat (3.78–4.1) across every price bracket.

4. Top performers by review volume include a mix of ratings — the single most-reviewed product (69 reviews) is rated only 2.8/5.

5. Underperformers — 18 products rated below 3.5 stars — already carry an average discount of 43%, similar to the dataset average, meaning discounting alone isn't fixing their poor engagement.

Dashboard

The workbook includes an interactive Dashboard tab with 5 linked PivotCharts and slicers allowing live filtering.

Building a Jumia Product Performance Dashboard in Excel

When I started this project, I had a Jumia product export sitting in front of me — 115 rows of product names, prices, discounts, reviews, and ratings — and five specific questions I needed to answer:

• Are larger discounts associated with more reviews?
• Do highly rated products attract stronger engagement?
• Do price and rating move together?
• Which products perform best based on ratings and reviews?
• Which products may need a different pricing or marketing strategy?

Here's how I got from raw, messy scraped data to a working interactive dashboard that answers all five — and what the data actually said, which wasn't always what I expected.

Step 1: Cleaning the data

Scraped e-commerce data is never clean, and this dataset was no exception. A few issues stood out immediately:

Prices were stored as text, like KSh 1,525, not numbers. Excel couldn't do any math on that. I fixed it with:

\
=VALUE(SUBSTITUTE(SUBSTITUTE(B2,"KSh ",""),",",""))
\\

This strips the "KSh " prefix and the thousands-separator comma, then converts what's left into a real number.

Discount was also text, formatted like 38%. A simple =VALUE(D2) was actually enough here — Excel is smart enough to parse the % symbol itself and convert "38%" straight into 0.38.

Review counts were all negative — every single one, like -2, -14, -55. That's clearly a scraping glitch, not real data, so I corrected it with =ABS(E2).

Rating was buried in descriptive text, like "4.5 out of 5". I extracted just the numeric part with:

\
=IFERROR(VALUE(LEFT(H2,FIND(" ",H2)-1)),"")
\\

FIND(" ", H2) locates the first space, LEFT() grabs everything before it, VALUE() converts that to a number, and IFERROR() handles blanks gracefully.

55 of the 112 products had no Review or Rating data at all — I left those cells genuinely blank instead of filling them with 0, since 0 would misleadingly suggest an actual rating of zero.

I also caught a header typo (Ratingd → Rating) and removed 3 fully empty rows at the bottom of the sheet.

Step 2: Building the analysis

With clean numeric data, I built a Calculations layer with a few extra metrics:

• Discount Amount (Ksh) — =Old price - Current price
• Price-to-Rating ratio — =Current price / Rating
• Engagement Level — bucketing products into Low/Medium/High/No reviews

Then five PivotTables, one per objective question, each with its own PivotChart:

1. Discount vs. Average Review — discount grouped into ~10% bands

1. Rating vs. Average Review — rating grouped into 0.5-point bands

1. Price vs. Average Rating — price grouped into bands

1. Top Performing — top 10 products by review count

1. Underperformers — products rated below 3.5, sorted by lowest engagement

Step 3: What the data actually showed

Discount size doesn't predict reviews. Average reviews across discount bands were essentially random: 15, 8.8, 19.2, 12.3, 12.3, 8.4 — no trend at all.

Higher ratings don't mean more engagement, either. Products rated 2.5–3.0 stars average 24.6 reviews, more than products rated 4.5–5.0 stars (14.4).

Price and rating are basically unrelated. Average rating barely moves across price brackets — 3.78 to 4.1 to 3.85 to 4.0

My top performer by review count — a cordless vacuum cleaner with 69 reviews — is only rated 2.8 out of 5.

My underperformers list (18 products rated below 3.5) already carry an average discount of 43%, almost identical to the dataset's overall average. Discounting harder isn't fixing whatever's actually wrong with them.

Step 4: The dashboard

I pulled all 5 PivotCharts onto a single Dashboard tab with slicers (Discount, Current Price, Product) so the whole thing stays interactive.

Overall conclusion: none of price, discount, or star rating strongly predicts customer engagement on their own in this dataset. That's a real, defensible finding.

---

Full workbook, cleaned data, and dashboard are on GitHub: [https://github.com/kimutaikoros/-Jumia-product-performance-dashboard]