An end-to-end data analytics project exploring M-Pesa adoption, financial inclusion trends, and unbanked population segments across Kenya's 47 counties — built using Python · SQL Server · Power BI.
Kenya is one of the world's most remarkable mobile money success stories. Since M-Pesa launched in 2007, the country has gone from 67.5% of its population being financially excluded to just 9.9% today. Yet millions of Kenyans — particularly in rural counties, older age groups, and lower income segments — still lack access to formal financial services.
This project analyses 10 years of CBK (Central Bank of Kenya) mobile money data and 7 waves of the FinAccess Household Survey (2006–2024) to answer four core questions:
- How has M-Pesa transaction value and adoption grown from 2015 to 2024?
- Which counties are the most and least financially included — and why?
- Which demographic groups (by age, gender, income, education) are most at risk of exclusion?
- Is there a measurable relationship between agent network density and financial inclusion?
- M-Pesa transaction value grew 7× in 10 years — from KES 1,238 Billion in 2015 to KES 8,700 Billion in 2024
- Mobile money now represents 57.6% of Kenya's GDP — among the highest ratios in the world
- Tana River County has the highest exclusion rate at 37.6% — compared to Nyandarua at just 2.6%
- People with no formal education are 4.6× more likely to be excluded than the national average
- Agent density and inclusion are positively correlated (r = 0.55) — counties with more agents per 1,000 people consistently show higher inclusion rates
- The 2020 COVID-19 pandemic accelerated digital adoption by 41% in a single year — the largest single-year jump in the dataset
Financial_Inclusion_Kenya/
│
├── Data/
│ ├── Raw/ # Original .xlsx files from CBK & FinAccess
│ └── Clean/ # 6 analysis-ready CSV files
│ ├── c_Mobile_Money_Monthly_Transactions_2015_2024.csv
│ ├── c_County_Financial_Inclusion_2024.csv
│ ├── c_FinAccess_Historical_Survey_2006_2024.csv
│ ├── c_Agent_Network_by_County.csv
│ ├── c_Demographic_Exclusion_Analysis.csv
│ └── c_Annual_KPIs_Summary.csv
│
├── Scripts/
│ ├── 01_clean.py # Data cleaning pipeline
│ ├── 02_eda.py # Exploratory data analysis + 8 charts
│ ├── 03_load_sql.py # Load data into SQL Server
│ └── fix_failed_tables.py # Targeted SQL load fix script
│
├── SQL/
│ └── create_tables.sql # All CREATE TABLE statements (star schema)
│
├── Dashboard/
│ └── Financial_Inclusion_Ke_Dashboard.pbix # Power BI report
│
└── README.md
| Tool | Purpose |
|---|---|
| Python (pandas, matplotlib, seaborn, scipy) | Data cleaning, EDA, and SQL loading |
| SQL Server + SSMS | Data warehouse — star schema with 4 dim tables and 5 fact tables |
| Power BI Desktop | Interactive 4-page dashboard |
| CBK Open Data | Primary data source — mobile money statistics |
| FinAccess Household Survey | Financial inclusion survey data (2006–2024) |
| File | Rows | What it contains |
|---|---|---|
| Monthly transactions | 120 | M-Pesa transaction volume, value, agents and accounts — every month 2015–2024 |
| County inclusion | 47 | Formal inclusion rate, exclusion rate, mobile money usage and agent density for all 47 counties |
| FinAccess historical | 7 | National inclusion figures from each of the 7 survey waves between 2006 and 2024 |
| Agent network | 47 | Active agent counts and density per 1,000 population for every county |
| Demographic exclusion | 20 | Exclusion rates broken down by gender, age group, location, income quintile, education level and disability |
| Annual KPIs | 10 | One row per year summarising national transaction totals, account registrations and GDP share |
The raw data came from CBK annual reports and FinAccess survey publications as Excel files. The cleaning script (01_clean.py) does the following for each file:
- Standardises all column names to lowercase with underscores (no spaces or special characters)
- Validates that all 47 counties are present and correctly spelled
- Checks that percentage columns are in the 0–1 decimal range, not 0–100
- Removes duplicate rows
- Adds a
relative_exclusion_indexto the demographics file — showing how many times worse each demographic segment is compared to the national average - Outputs a cleaning report to
data/clean data/cleaning_report.txtdocumenting every check and fix
The EDA script (02_eda.py) generates 8 charts answering the core project questions:
- M-Pesa transaction value growth 2015–2024 (bar + line, dual axis)
- Monthly transaction volume heatmap — seasonal patterns by year and month
- Top 10 most excluded counties (horizontal bar, colour-coded by region)
- Financial inclusion evolution 2006–2024 (stacked area across 7 survey waves)
- Agent density vs inclusion rate by county (scatter plot)
- Mobile money as % of GDP over time (dual-axis line)
- County inclusion distribution by region (box)
- Correlation matrix of county-level variables
The clean CSV files were loaded into SQL Server using a star schema — a design pattern commonly used in business intelligence.
Dimension tables (lookup tables):
dim_county— 47 rows, one per countydim_date— 120 rows, one per year-monthdim_demographic— 20 rows, one per demographic segmentdim_region— 8 rows, one per region
Fact tables (the measurable data):
fact_mobile_transactions— monthly M-Pesa datafact_county_inclusion— county-level inclusion metricsfact_demographic_exclusion— segment-level exclusion datafact_annual_kpis— national annual KPIsfact_finaccess_trends— historical survey wave data
The Python script (03_load_sql.py) connects to SQL Server using SQLAlchemy and pyodbc, builds the tables as DataFrames, and loads them with df.to_sql(). A final verification step runs a SELECT COUNT(*) on every table to confirm row counts match.
The Power BI report connects directly to SQL Server and contains 4 pages:
Page 1 — National Overview
Four KPI cards (total transaction value, active agents, national inclusion rate, excluded population), a bar chart showing annual transaction value growth, a line chart of registered vs active accounts, a donut chart showing included vs excluded split, and a table of the top 5 most excluded counties.
Page 2 — County Map
A filled map of all 47 Kenyan counties coloured by inclusion tier, a scatter plot showing agent density vs inclusion rate with region colour coding and population as bubble size, a region slicer that filters all visuals simultaneously, and a full county data table with conditional formatting on agent density tier.
Page 3 — Transaction Trends
A full-width monthly line chart with a year range slicer, a quarterly volume matrix showing seasonal patterns, and a clustered bar chart comparing registered vs active accounts with a utilisation rate line.
Page 4 — Demographics
A horizontal bar chart of exclusion rates for all 20 demographic segments with a national average reference line, a 100% stacked bar showing inclusion vs exclusion by category, a segment detail table with colour-coded relative exclusion index, and a category slicer that filters all visuals by Gender, Age, Location, Income, Education or PWD.
| Page | Description |
|---|---|
| National Overview | KPIs, transaction growth, account utilisation |
| County Map | Filled map + scatter + county data table |
| Transaction Trends | Monthly line chart, quarterly matrix, year slicer |
| Demographics | Exclusion by segment, inclusion evolution |
pip install pandas openpyxl sqlalchemy pyodbc matplotlib seaborn scipyYou also need:
- SQL Server (Express edition is free) with SSMS installed
- ODBC Driver 17 for SQL Server (download here)
- Power BI Desktop (free from Microsoft)
# 1. Clone the repo
git clone https://github.com/peter-ngamau/financial-inclusion-in-kenya.git
cd financial-inclusion-kenya
# 2. Run the cleaning script
# (place your raw xlsx files in data/raw/ first)
python scripts/01_clean.py
# 3. Run EDA and generate charts
python scripts/02_eda.py
# 4. Create the SQL Server tables
# (run SQL/create_tables.sql in SSMS first)
# 5. Load data into SQL Server
# (update SERVER name in the script first)
python scripts/03_load_sql.py
# 6. Open the Power BI file
# Dashboard/Financial_Inclusion_Ke_Dashboard.pbix
# Update the SQL Server connection in Transform Data → Data source settings| Source | Data |
|---|---|
| Central Bank of Kenya | Mobile money statistics, agent network data, annual reports |
| FinAccess Household Survey | Financial inclusion survey data — CBK, KNBS & FSD Kenya |
| Kenya National Bureau of Statistics | County population estimates 2024 |
[Peter Ngamau] Data Analyst | Python · SQL · Power BI
- 📧 [ptahmwangi@gmail.com]
- 💼 [LinkedIn profile]
- 🐙 [GitHub profile]
This project is open source under the MIT License.