Skip to content

Day 05 Part 1 — End-to-End Analytics Project

The capstone session. This walks through a complete analytics project from raw data to stakeholder recommendation — integrating every skill from both weeks. Use this as the blueprint for your own portfolio project.

Learning Objectives

  • Execute a full analytics project end to end
  • Integrate SQL, Python, visualisation, and storytelling
  • Structure a project for a portfolio
  • Document your work for reproducibility and review

The End-to-End Workflow

1. UNDERSTAND   → What's the business question? Who's the stakeholder?
2. GET DATA     → Source it. Understand the schema.
3. CLEAN        → Handle missing values, duplicates, types, outliers.
4. EXPLORE      → EDA — distributions, relationships, trends.
5. ANALYSE      → Answer the specific business questions.
6. VISUALISE    → Build the dashboard / charts.
7. RECOMMEND    → Turn findings into actions.
8. COMMUNICATE  → Present to the stakeholder.
9. DOCUMENT     → Make it reproducible and portfolio-ready.

Each step maps to a course module — this session ties them together.

Worked Example — Structure of a Complete Project

Using the Customer Churn project as the model:

1. Understand

"TelcoMax's churn rose from 5% to 8%. The Head of Customer Success has £200k to spend on retention and needs to know who's churning, why, and who to target."

2-3. Get Data & Clean

Load the data, audit quality, fix the TotalCharges type, handle the blanks, engineer features (tenure buckets, add-on count). Document every cleaning decision. (See cleaning steps)

4-5. Explore & Analyse

EDA reveals overtime, contract type, and tenure as the drivers. SQL quantifies them: month-to-month customers churn at 42.7% vs 2.8% for two-year contracts. (See SQL)

6. Visualise

A two-page dashboard: executive overview + at-risk operations page. (See dashboard)

7. Recommend

Contract conversion campaign, 90-day onboarding, fix the fiber experience — each tied to a finding and quantified. (See recommendations)

8-9. Communicate & Document

A 5-slide deck for the stakeholder + a clean, reproducible notebook for the portfolio.

Building a Portfolio-Ready Project

Recruiters and hiring managers look at your portfolio. Make each project shine:

  • [ ] A clear README — business problem, your approach, key findings, tools used
  • [ ] A reproducible notebook — runs top to bottom, relative paths, clear sections
  • [ ] Documented decisions — why you cleaned the way you did, what you assumed
  • [ ] A visual artefact — a dashboard screenshot or chart that catches the eye
  • [ ] A written summary — findings and recommendations in plain English
  • [ ] Honest caveats — what the analysis can't tell you (shows maturity)

One great project beats five mediocre ones

Recruiters spend seconds per portfolio. A single deep, well-documented, well-presented project with a clear business story beats a pile of half-finished notebooks. Pick one course project and polish it to perfection.

Choosing Your Capstone

The [[01-sales-performance-dashboard|6 capstone scenarios]] are designed for this. Pick one that: - Uses tools you want to be hired for (SQL + Python + a BI tool) - Tells a clear business story - Has a dataset you can actually access - Genuinely interests you (it shows in the work)

Reproducibility Checklist

# Good notebook hygiene
# 1. All imports at the top
import pandas as pd
import matplotlib.pyplot as plt

# 2. Relative paths, never absolute
df = pd.read_csv("data/orders.csv")   # not C:\Users\...

# 3. Clear markdown sections explaining each step
# 4. Restart & Run All before sharing — confirm it works end to end
# 5. Set random seeds for reproducibility
import numpy as np
np.random.seed(42)

Practice Exercises

  1. Choose one course project or capstone scenario as your portfolio piece.
  2. Write its README: business problem, approach, key findings, tools.
  3. Audit your notebook against the reproducibility checklist.
  4. Produce one polished "hero" visualisation for the project.
  5. Write a one-page plain-English summary of findings and recommendations.

Project Self-Review Questions

Before calling a project "done," ask: - Can someone understand the business problem in 30 seconds from the README? - Does the notebook run top to bottom without errors? - Is every cleaning decision documented and justified? - Does each chart have an insight-driven title? - Is there a clear, specific, quantified recommendation? - Did I state the limitations honestly?

← Data Storytelling · Next: Mock Interview & Resume →