Akanbi Abdullahi

Cyclistic Bike Sharing Analysis

May 2024

Overview

Cyclistic is a fictional bike-share company. The objective was to understand behavioural differences between casual riders and annual members in order to support a marketing strategy aimed at converting casual riders into subscribers. I took the project from raw data through to a polished executive presentation.

What I Did

Collected and cleaned 12 months of trip data, resolving null values, duplicate records and inconsistent column formats. Performed exploratory analysis to surface ride duration trends, station popularity, and seasonal usage patterns segmented by rider type. Built a series of visualizations that clearly showed where and when casual riders diverge from members — arming the marketing team with specific, actionable targeting opportunities.

Key Outcomes

• Identified that casual riders average 2× longer trips than members, concentrated on weekends
• Pinpointed the top 10 stations used almost exclusively by casual riders — direct marketing targets
• Delivered a fully annotated presentation ready for stakeholder review

Customer Analysis Dashboard

Sep 2025

Overview

Built an interactive Tableau dashboard to surface customer behaviour patterns, purchase trends and segment performance for a retail dataset. The goal was to give business stakeholders a single view they could self-serve — filtering by region, product category, and time period — without needing to request ad-hoc reports from the data team.

What I Did

Modelled the underlying data in SQL to produce a clean, performant dataset for Tableau to consume. Designed the dashboard layout iteratively — starting with stakeholder interviews to establish which metrics mattered most — then building calculated fields, parameters and actions that made the dashboard genuinely interactive rather than static. Refined the visual hierarchy to draw attention to the most business-critical KPIs immediately on load.

Key Outcomes

• Replaced a weekly manual Excel report that took 3 hours to produce
• Surfaced a previously unnoticed drop-off in repeat purchases in Q3 — flagged for campaign action
• Delivered a reusable dashboard template now applied across multiple business units

Data Warehouse Project

Oct 2025

Overview

Designed and built a SQL-based data warehouse from scratch to consolidate data from multiple source systems into a single, query-optimised analytical store. The project followed a layered architecture — bronze (raw ingestion), silver (cleaned and conformed), gold (aggregated, business-ready) — to ensure clean separation of concerns and maintainability.

What I Did

Defined the warehouse schema using a star schema model with clearly separated fact and dimension tables. Wrote ETL scripts to extract, transform and load data from source CSVs, handling type coercion, null management and deduplication at the silver layer. Created documentation covering table definitions, lineage diagrams, and loading procedures so the warehouse could be maintained and extended by others.

Key Outcomes

• Reduced average analytical query time by structuring data for read-optimised joins
• Produced a fully documented schema enabling team onboarding without tribal knowledge
• Established a repeatable ETL pattern reusable across future data ingestion projects

End-to-End Azure Data Engineering Pipeline

2025

Overview

Designed and built a complete cloud-based data engineering pipeline on Microsoft Azure — from raw HTTP data ingestion through to a Power BI reporting layer. The project follows a medallion architecture (bronze → silver → gold), using Azure-native services at every stage to ensure scalability, security and maintainability.

Architecture

Data is ingested from an HTTP source via Azure Data Factory into a Data Lake Gen2 raw store (bronze layer). Azure Databricks handles all transformation logic, cleaning and conforming data into the silver layer before writing to a second Data Lake Gen2 store. The transformed data is then served via Azure Synapse Analytics (gold layer), which feeds directly into Power BI for reporting. Access control is managed through Microsoft Entra ID.

What I Did

Configured the end-to-end pipeline from scratch — provisioning all Azure services, designing the Data Factory pipelines, writing Databricks notebooks for transformation logic, and setting up Synapse as the serving layer. Implemented role-based access control via Entra ID across all components and documented the full architecture for handover and reproducibility.

Key Outcomes

• Delivered a fully automated, end-to-end pipeline requiring no manual data movement
• Medallion architecture ensures clean separation between raw, clean and business-ready data
• Power BI report layer connects directly to Synapse for real-time, query-on-demand reporting
• Full documentation produced covering architecture decisions, setup steps and access model

Customer Segmentation — Regression Model in R

Oct 2024

Overview

Built a regression-based customer segmentation model in R to group customers by purchasing behaviour and predict their likely spend category. The model was designed to feed directly into marketing targeting — giving the team a data-backed basis for personalising campaigns rather than relying on broad demographic assumptions.

What I Did

Performed exploratory data analysis to understand the shape and distribution of customer data, handling outliers and missing values before modelling. Trained and evaluated multiple regression models, comparing performance using RMSE and R² metrics. Produced clear, annotated visualisations of model outputs — segment membership, predicted spend bands, and feature importance — to make the results accessible to non-technical stakeholders.

Key Outcomes

• Identified 4 distinct customer segments with significantly different spending profiles
• Achieved a model accuracy sufficient for reliable campaign targeting decisions
• Produced a reproducible R script and commentary enabling the analysis to be re-run with fresh data

SQL Data Cleaning Project

Sep 2024

Overview

Designed and implemented a comprehensive SQL-based data cleaning pipeline to transform a raw, messy dataset into a reliable, analysis-ready form. The project established a set of reusable data quality checks and transformation patterns that could be applied across future datasets — not just a one-off fix, but a scalable framework.

What I Did

Audited the raw dataset systematically to catalogue data quality issues: nulls, duplicates, inconsistent formatting, out-of-range values and orphaned records. Wrote modular SQL scripts addressing each category of issue, with inline comments explaining the rationale for each transformation. Documented the full process in a data quality log that records what was changed, why, and what the source data looked like before and after.

Key Outcomes

• Reduced null and duplicate records from ~18% of rows to under 0.5%
• Standardised date and string formats across 14 columns for consistent downstream querying
• Produced a documented, repeatable cleaning framework reusable on future projects