Team project (Ironhack). My role: Exploratory Data Analysis (EDA), hypothesis testing (two-proportion z-test + confidence intervals), and business recommendations.
We tested a new UI for Vanguard to see if it actually improves customer experience.
KPIs: Completion Rate, Error Rate, Time per Step.
Business bar: +5% completion uplift.
Vanguard is one of the world’s most respected investment management companies.
Vanguard considered redesigning its onboarding UI to make the process smoother and encourage higher client conversion.
Key question: Is the redesign worth the investment?
We collected four datasets, cleaned and merged them into one.
Steps included: renaming columns, removing duplicates/invalid values, and dropping nulls in key variables.
Control and Test groups had similar demographics.
- Age ~48 years on average
- Tenure mostly 5–15 years
- Gender distribution balanced
This means the A/B test is valid and representative.
✅ Completion improved (+3.7 pp).
❌ Error rate also increased (+3.4 pp).
Users finished more often, but with more mistakes.
| Metric | Control | Test | Δ (Test – Control) |
|---|---|---|---|
| Completion % | 65.6 | 69.3 | +3.7 pp |
| Error rate % | 34.4 | 37.8 | +3.4 pp |
| Avg. time | Mixed | Mixed | Faster at Step 2 & 3, slower at Confirm |
- Null (H0): No improvement in completion.
- Alt (H1): Test has higher completion.
- Result: Statistically significant (p < 0.001).
- But: Business threshold of +5% uplift was not reached.
- Do not roll out the redesign fully yet.
- Optimize Step 1 and Confirm page (too slow, more hesitation).
- Keep the improvements from Step 2 and Step 3.
- Re-run the experiment after fixes.
- Python + Jupyter — cleaning, EDA, z-test
- Tableau — visual storytelling
- Excel — preprocessing
- Trello & Slack — teamwork & communication
- Notebook →
project5_egbe.ipynb - Images →
/images/(EDA & results) - Slides → Google Slides
👩🏽💻 Egbe Grace — Data Analyst
GitHub: Egbe34