A five-phase end-to-end analytics project examining where 2,000 newly onboarded customers dropped off across XYZ Bank's digital platform in FY 2024, and what Ksh 1.58 billion in revenue at risk means for the business.
Built for boardroom presentation. Four KPI cards, a customer journey funnel, activation rates by channel and segment, retention breakdown, and revenue impact — all in one interactive view.
XYZ Bank's digital banking platform recorded strong onboarding growth in FY 2024, with 2,000 new customers registered across Mobile App, USSD, Agent, and Branch channels. But monthly active user numbers were not keeping pace — raising urgent questions at senior management level about where customers were dropping off and why.
This project was commissioned to answer five core business questions:
Where exactly in the digital journey are customers dropping off — before or after activation?
Which customer segments are most at risk of churn, and how different are their behaviours?
Is this a channel problem, or a platform-wide retention problem affecting all onboarding methods?
What is the financial value at risk from customers who churned or went dormant?
Which channels activate the highest proportion of registered customers?
What interventions would have the highest impact on reversing the retention decline?
The analysis uncovered a systemic retention failure spanning every segment and every channel — with a single overriding conclusion: this is not a channel problem, it is a platform-level engagement problem.
| # | Finding | Detail | Severity |
|---|---|---|---|
| 1 | Activation Gap | 26.4% of registered customers (528) never made a single transaction after signing up. | High |
| 2 | Retention Crisis | Only 19.6% of all registered customers were active by year-end — a critical engagement failure. | Critical |
| 3 | Churn Dominates | 45% of activated customers churned — the single biggest drop-off in the entire journey. | Critical |
| 4 | High-Risk Segments | SME had the highest churn rate (48.9%), followed by Mass Affluent (47.1%). | Critical |
| 5 | Platform-Wide Problem | Churn rates were consistent across all channels (43–47%) — confirming this is not a channel issue. | High |
| 6 | Revenue at Risk | Ksh 1.19B at risk from churn + Ksh 394M from dormant customers = Ksh 1.58B total exposure. | Critical |
Three analytical lenses across the same 2,000-customer dataset — each revealing a different dimension of the retention failure.
Churn rates of 43–47% across all four channels — Mobile App, Branch, Agent, and USSD — is the clearest possible signal that the problem is not in how customers onboard. It is in what the platform offers them after they do.
→ This is a product engagement problem, not a channel problem.
This project was built from raw data generation through to boardroom-ready reporting — replicating the full lifecycle of a real banking analytics engagement.
Designed a realistic 2,000-row digital banking dataset with intentional data quality issues to simulate real-world conditions — including duplicate rows, impossible dates, negative values, casing inconsistencies, blank strings, and ETL sync failures.
Connected Python to MySQL via SQLAlchemy and PyMySQL. Conducted a structured audit across all columns before applying any fixes. Each step followed a preview → apply → verify pattern. Nine distinct issue types resolved across 2,012 raw rows.
Connected directly to MySQL via SQLAlchemy. Analysis conducted in Pandas with visualisations in Matplotlib and Seaborn. Five analytical areas: activation, funnel, segment retention, channel retention, and revenue impact.
Built an executive-ready interactive dashboard with 4 views and 4 KPI cards. Published to Tableau Public. Covers the customer journey funnel, activation rates by channel and segment, retention breakdown, and revenue at risk.
Produced a one-page boardroom-ready executive brief covering business context, key findings, recommendations, and next steps. Formatted to professional banking reporting standards.
Every issue was documented, previewed, applied, and verified before moving on — following a structured audit methodology across all 14 columns.
| # | Issue | Records | Action Taken |
|---|---|---|---|
| 1 | Duplicate rows | 12 | Removed using ROW_ID deduplication |
| 2 | Impossible dates (first transaction before registration) | 25 | Set to NULL |
| 3 | Negative total_transactions | 8 | Corrected using ABS() |
| 4 | Segment casing inconsistencies | 37 | Standardised using TRIM() + UPPER/LOWER |
| 5 | Blank status values | 10 | Converted to NULL |
| 6 | Blank region values | 15 | Converted to NULL |
| 7 | Blank age_band values | 20 | Relabelled as 'Unknown' |
| 8 | Value/transaction mismatch (ETL failure) | 40 | Flagged as 'REVIEW' for Finance |
| 9 | VARCHAR date columns | 2 cols | Converted to DATE type |
Final dataset: 2,000 rows × 14 columns. Zero rows lost to cleaning.
Each recommendation is tied to a specific finding and includes a measurable target to enable post-implementation tracking.
Implement a 72-hour post-registration follow-up workflow for Branch-onboarded customers. Branch has the lowest activation rate (56.4%) — a structured nudge at the right moment can close that gap.
Target: Activation 56.4% → 70%Launch tailored digital banking features addressing bulk payments, payroll, and supplier management. SME churns at 48.9% — the highest of any segment — yet likely has the highest revenue potential per customer.
Target: Reduce SME churn below 40%Target 418 dormant activated customers with personalised push notifications and transaction fee waivers. These customers proved they can activate — they need a compelling reason to return.
Revenue at risk recovered: Ksh 394MCommission a product review focused on post-activation engagement. Consistent churn across all channels (43–47%) is the clearest possible signal of a platform-level problem, not a channel-level one.
Platform-level retention programme