Lesson 15 of 16 · Level 4: Expert

Portfolio Project: Analyze a Business End to End

  • ⏱ 14 min read
  • ❓ 4 review questions
  • ✍️ Exercise with solution
  • 📥 Downloadable template
In this lesson
  1. The brief
  2. The project plan
  3. Step 1: Clean and document
  4. Step 2: Analyze, question by question
  5. Step 3: Recommendations
  6. Step 4: Package the deliverables
  7. Step 5: Publish it
  8. Make it yours
  9. Self-check rubric
  10. Summary
What you'll learn
  • How to scope a portfolio project around business questions, not tools.
  • How to run the full analysis process on the practice data, from raw file to recommendations.
  • Which deliverables to produce: cleaning log, queries or notebook, dashboard, summary deck and README.
  • How to publish the project on GitHub, a public dashboard link and LinkedIn.
  • A rubric to check your project before you show it to employers.
📥 Practice this lesson

This project starts from the raw data, just like a real job. Use the clean file only to check your work at the end.

A portfolio project is proof that you can do the job: take a messy dataset and a vague business need and turn them into clear findings and recommendations. Hiring managers can't see your course certificates in action, but they can open a project and see how you think. In this lesson you'll complete an end-to-end year-end review of the practice retailer, using every skill from the course, and package it so it's ready to share.

$42,514gross profit on $89,735 of revenue: a 47.4% margin
68.5%of revenue from wholesale orders, the project's central finding
+41.9%growth in consumer sales from the first to the second half of 2025

The brief

Write the brief as if a manager had sent it. That framing turns a dataset into a business problem, which is what employers want to see:

Project brief

"You're the new data analyst at a U.S. retailer that sells electronics, home, office and fitness products online, in stores and wholesale. Leadership wants a 2025 year-end review before planning 2026: how did we do, where does our money come from, what are the risks, and what should we do differently next year?"

Break it into five questions you can actually answer with the data:

  1. How did the business perform in 2025 (revenue, orders, order value, gross profit)?
  2. Where does revenue come from by region, channel, category and product?
  3. How concentrated is revenue, and how risky is that?
  4. How did sales move through the year?
  5. What should the company do in 2026?

The project plan

  1. AskBrief and questionsThe five questions above, written at the top of your README.
  2. Prepare and processClean the raw fileProfile it, fix duplicates, blanks, state names and the text date, and keep a cleaning log (Lesson 5).
  3. AnalyzeAnswer each questionPivot tables, SQL queries or pandas code; join product costs for gross profit (Lessons 6–9 and 14).
  4. ShareDashboard and summaryA one-page dashboard and a five-slide summary with headline titles (Lessons 10–12).
  5. ActRecommendationsThree to five concrete actions, each tied to a finding and a metric to track.

Step 1: Clean and document

Start from sales-practice-data.csv. Your cleaning log should end with a reconciliation that anyone can verify:

StepRowsRevenue
Raw file as received203$90,231 (4 blanks ignored)
Remove 3 exact duplicate rows200−$853
Recalculate 4 blank revenues (units × price)200+$357
Fix 7 state names and 1 text date200$0
Clean dataset200$89,735

Then compare with the Sales Clean sheet. If your total matches, your cleaning is right.

Step 2: Analyze, question by question

Q1. How did we do?

Revenue was $89,735 from 200 orders. The average order was $448.68, but the median was only $138, so "typical" orders are small. After joining product costs, gross profit was $42,514, a 47.4% gross margin.

Q2. Where does revenue come from?

2025 revenue by sales channel Wholesale $61,479 (68.5%), online $15,861 (17.7%) and retail $12,395 (13.8%). $89,735 2025 revenue
  • Wholesale (68.5%) $61,47937 orders, averaging $1,661.59.
  • Online (17.7%) $15,86192 orders, averaging $172.40.
  • Retail (13.8%) $12,39571 orders, averaging $174.58.
Three channels, two very different businesses: a high-volume consumer business and a low-volume, high-value wholesale business.

By region, the West leads with $36,626 (40.8%), and 76.8% of that is wholesale; California ($15,467) and Arizona ($9,776) are the top states. By category, Electronics brings in 49.1% of revenue. The single biggest product, the Smart Watch ($23,482, 26.2% of revenue), depends heavily on wholesale: 6 wholesale orders produced $18,308, or 78.0% of its revenue.

Q3. How concentrated is revenue?

Very. Wholesale is 18.5% of orders but 68.5% of revenue. The 5 largest orders produced $21,454 (23.9% of the year) and the 10 largest produced $33,307 (37.1%). Losing a handful of business buyers would hit revenue far harder than losing hundreds of consumers.

Q4. How did sales move through the year?

Q1 was the weakest quarter ($9,087, 10.1% of revenue) and Q4 the strongest ($29,459, 32.8%), with Q2 close behind ($28,927). The quarterly swings follow wholesale orders. Consumer sales (online plus retail) were steadier and grew from $11,679 in the first half to $16,577 in the second half, an increase of 41.9%.

Q5. Anything else worth flagging?

Gross margins are similar across channels (about 47%), so wholesale buyers aren't getting discounts that erode profit. Fitness has the lowest margin (43.9%), driven by adjustable dumbbells (41.1%). And the Desk Lamp, which is in the catalog, sold zero units all year: a LEFT JOIN found it.

1. Which finding is the strongest candidate for the headline of the project?

  • "The data has 10 columns and 200 rows."
  • "Wholesale is 18.5% of orders but 68.5% of revenue, so revenue depends on a few business buyers."
  • "Yoga mats are the lowest-revenue product."

The headline should be the finding with the biggest business consequence. Revenue concentration affects strategy, risk and next year's plan. The others are true but either trivial or minor.

Step 3: Recommendations

Every recommendation should point back to a finding and include a way to measure it:

RecommendationBased onMetric to track
Create a wholesale account program for the top business buyers (starting in California and Arizona)68.5% of revenue from 37 wholesale ordersWholesale orders per month; repeat wholesale buyers
Keep growing consumer channels, testing marketing changes with A/B testsConsumer sales +41.9% from H1 to H2, steadier than wholesaleMonthly online + retail revenue; conversion rate
Review fitness pricing and supplier costsLowest category margin (43.9%)Fitness gross margin
Promote or discontinue the Desk LampZero sales in 2025Units sold after a promotion test
Add a customer ID and acquisition source to the order dataCan't measure repeat customers todayData available by Q2 2026

Also write down the limitations: one year of data, so no proof of seasonality; no customer IDs; no marketing spend; and the data is fictional (say so clearly in your README).

2. Which recommendation is best written?

  • "Improve wholesale."
  • "Wholesale is very important and we should think about it."
  • "Assign an account manager to the top wholesale buyers in California and Arizona in Q1 and track wholesale orders per month."

A strong recommendation says what to do, where, when and how success will be measured. The first two are directions, not actions anyone can execute.

Step 4: Package the deliverables

README

The front page of your project: brief, questions, key findings, recommendations, tools, limitations.

Tip: many reviewers read only this. Put the findings in the first screen.

Cleaning log and code

Your SQL file or pandas notebook, plus the cleaning log table.

Tip: comment the code so a reviewer can follow each step.

Dashboard

A one-page Power BI or Looker Studio dashboard with KPIs, trend, channel mix and filters.

Tip: include a screenshot in the README in case the link breaks.

Five-slide summary

The storyline from Lesson 12, exported as a PDF.

Tip: headline titles only; details in an appendix.

A simple README structure (GitHub displays a file named README.md automatically):

# 2025 Sales Review: U.S. Retailer (fictional data)

## Business question
How did the business perform in 2025, where does revenue come from,
and what should change in 2026?

## Key findings
- Revenue: $89,735 from 200 orders; gross margin 47.4%.
- Wholesale = 18.5% of orders but 68.5% of revenue.
- Top 5 orders = 23.9% of annual revenue (concentration risk).
- Consumer sales grew 41.9% from H1 to H2.

## Recommendations
1. Wholesale account program (CA and AZ first).
2. Grow consumer channels with A/B-tested marketing.
3. Review fitness pricing and costs (43.9% margin).

## Process and tools
Excel (cleaning log), SQL (SQLite), Power BI dashboard, Python (pandas).

## Limitations
One year of fictional data; no customer IDs or marketing spend.

Step 5: Publish it

  • GitHub: create a free account and a public repository with the README, your SQL file or notebook, the cleaned CSV and screenshots. It's the standard place employers look for code.
  • Dashboard link: Looker Studio reports can be shared with a public link. Power BI's "Publish to web" also creates a public link if your account allows it; only use it for non-confidential data like this fictional set.
  • LinkedIn: add the project to the Featured section with one sentence on the finding, plus a short post explaining what you learned.
Never publish real company data

If you analyze data from a current or former job, don't post it. Recreate the analysis with public or fictional data instead. This course's dataset is fictional, so it's safe to share; say so in your README so nobody thinks you leaked an employer's numbers.

3. A reviewer has 60 seconds for your project. What should they see first?

  • The full SQL code
  • A list of every tool you know
  • The README's business question and three key findings with numbers

Recruiters and hiring managers skim. A clear question and quantified findings at the top show business thinking immediately; the code and details are there for anyone who wants to dig deeper.

Make it yours

Many learners complete the same course project, so add something original. Ideas: analyze gross margin by state, compare average price per unit by channel, build a "what if" scenario (what happens to revenue if the 5 largest orders don't repeat?), or repeat the whole workflow on a public dataset from a government open-data portal such as data.gov. A second project on a topic you care about (sports, real estate, public health) makes your portfolio memorable.

Self-check rubric

4. Your cleaned total comes out to $90,231. What does that tell you?

  • The analysis is fine; small differences don't matter
  • You haven't cleaned yet: that's the raw total, with duplicates included and blanks ignored
  • The products file is missing

$90,231 is the raw sum: it includes $853 from 3 duplicate rows and ignores $357 of blank revenues. The clean total is $89,735. Reconciling to a known total is how you catch this before a reviewer does.

Practice exercise

Write the executive summary for this project: three bullet points with numbers and one recommendation, in no more than 80 words. It should work as the first section of your README and the first slide of your deck.

See solution

One possible version (66 words):

  • 2025 revenue was $89,735 from 200 orders, with a 47.4% gross margin ($42,514).
  • Wholesale was 18.5% of orders but 68.5% of revenue; the 5 largest orders alone brought in 23.9%.
  • Consumer sales were steadier and grew 41.9% from the first half to the second half.

Recommendation: launch a wholesale account program in Q1 2026, starting with California and Arizona, while growing consumer sales with tested marketing.

Check: every number traces back to the analysis above, and the recommendation follows directly from the biggest finding.

Summary

  • Frame the project as a business brief with a few answerable questions.
  • Clean from the raw file and reconcile: 203 rows and $90,231 become 200 orders and $89,735.
  • Answer each question with numbers: 47.4% margin, 68.5% of revenue from wholesale, +41.9% consumer growth.
  • Make recommendations specific and measurable, and state limitations.
  • Package a README, code or queries, a dashboard and a five-slide summary; publish on GitHub and LinkedIn.
  • Add an original twist so your project stands out.

Next lesson: How to Get Your First Data Analyst Job. You'll turn this project into a résumé, a LinkedIn profile and strong interview answers.

🎓 Save your progress and get a free certificate

Create your account in 30 seconds, mark your lessons and download your certificate at the end.

Enroll for free