How osamah-anmar-sami/retail-analysis works
1. Repository type
- This is a data analysis / BI portfolio project, not an application service.
- The architecture is centered on raw retail data → curated dataset → exploratory analysis → dashboards/reporting.
2. Primary data source
- The project uses a multi-country retail transactional dataset from Kaggle.
- The data includes customer, product, purchase, and transaction information across countries such as the USA, UK, Canada, Germany, and Australia.
3. Data curation layer
- DataCuration.ipynb is the first processing stage.
- It performs the main cleanup and shaping work: handling missing values, creating new features, converting data types, renaming columns, and removing duplicates.
- This notebook is the boundary between the raw external dataset and the analysis-ready dataset.
4. Exploratory analysis layer
- EDA.ipynb consumes the curated data and performs univariate, bivariate, and multivariate analysis.
- Its role is to identify patterns, relationships, and insights rather than to transform the data further.
- This is the analytical core of the repository.
5. Visualization / dashboard outputs
- Analysis Multiple Countries Retail Data.twbx is a Tableau packaged workbook used to build the published dashboard.
- The README links to a Tableau Public dashboard, so Tableau is the main external visualization surface for sharing results.
- Retail Analysis.pbix is a Power BI report artifact, indicating a second BI visualization deliverable in the repo.
6. Reporting layer
- Final Report.pdf is the static presentation layer that summarizes findings from the curation and EDA work.
- This file is downstream of the notebooks and dashboards: it packages the insights into a consumable deliverable for review or portfolio presentation.
7. Technology stack
- The analytical pipeline is built in Python notebooks with:
- pandas for data manipulation
- numpy for numeric operations
- matplotlib, seaborn, and colorcet for plotting
- The project also uses Tableau and Power BI for dashboarding.
8. Execution model
- The repo is notebook-driven, with the main entry points being:
- DataCuration.ipynb
- EDA.ipynb
- The README explicitly instructs running them via Jupyter Notebook, so there is no separate application server or API layer.
9. Data flow
- External Kaggle dataset feeds into the curation notebook.
- Curated output flows into EDA.
- EDA findings feed into Tableau / Power BI dashboards and the final PDF report.
- The flow is essentially one-way and batch-oriented, not interactive or transactional.
10. Architectural boundaries
- The key boundary is between raw data ingestion/cleanup and analytical consumption.
- Another boundary exists between analysis artifacts (notebooks) and presentation artifacts (dashboard workbooks and PDF report).
- The repository does not expose a service boundary, database boundary, or deployment boundary.
11. External dependencies and distribution
- Data acquisition depends on the Kaggle dataset URL in the README.
- Dashboard distribution relies on Tableau Public hosting for the published visualization.
- The remaining artifacts are local files intended for reproducible analysis and portfolio sharing.
12. What to draw in an architecture diagram
- Show:
- Kaggle retail dataset as the source
- DataCuration.ipynb as the transformation stage
- EDA.ipynb as the analysis stage
- Tableau workbook / Tableau Public and Power BI report as visualization outputs
- Final Report.pdf as the final narrative deliverable
- The diagram should reflect a single analytical pipeline with multiple presentation outputs.
Main entry points: EDA.ipynb EDA · README.md README · DataCuration.ipynb Data curation · Analysis Multiple Countries Retail Data.twbx Tableau workbook
Components 8
Source data
External retail dataset feeding the analysis pipeline.
- Kaggle data · Multi-country retail transactional source data from Kaggle.
Analysis layer
Notebook-driven curation and exploratory analysis.
- Data curation · Cleans, renames, deduplicates, and feature-engineers the raw dataset.
DataCuration.ipynb - EDA · Performs univariate, bivariate, and multivariate exploration on curated data.
EDA.ipynb - README · Explains how to run the notebooks and access the published dashboard.
README.md
Presentation layer
Packaged dashboards and report outputs for sharing findings.
- Tableau workbook · Packaged Tableau workbook for interactive retail visualization.
Analysis Multiple Countries Retail Data.twbx - Tableau Public · External publishing surface for the dashboard experience.
- Power BI report · Secondary BI report deliverable for retail analysis.
Retail Analysis.pbix - Final report · Static narrative summary of findings and insights.
Final Report.pdf
Connections 9
- Kaggle data → Data curation: raw data · External retail data is ingested and cleaned.
- Data curation → EDA: curated data · Analysis uses the cleaned, analysis-ready dataset.
- EDA → Tableau workbook: insights · Exploratory findings inform dashboard design and content.
- Tableau workbook → Tableau Public: publish · Workbook is distributed through Tableau Public.
- EDA → Power BI report: reporting input · EDA findings also feed the Power BI deliverable.
- EDA → Final report: summary · Notebook outputs are distilled into the final PDF report.
- README → Data curation: run · README identifies the curation notebook as a main entry point.
- README → EDA: run · README identifies the EDA notebook as a main entry point.
- README → Tableau Public: view · README points to the published Tableau dashboard.