Testing

Testing Strategy

For the FinSync project, our testing approach is meticulously designed to ensure thorough coverage and robustness of the application. We have achieved a code coverage of over 95%, reflecting our commitment to quality and our confidence in the product delivered to our clients.

Fully-automated:

Unit testing: We meticulously test each component to ensure individual functionalities operate as intended.
Integration testing: This ensures that various components of the application work together seamlessly, validating the entire workflow from end to end.
Performance testing: We assess the application's performance to ensure it meets the required speed and efficiency under varying loads.

Semi-automated:

User Acceptance testing: Conducted with selected end-users to ensure the application meets the real-world scenarios and fulfills user requirements.

Unit Testing

Unit testing was a foundational aspect of our quality assurance process. We tested each module in isolation to ensure that all individual functions and components behave as expected.

Backend Testing (Python)

We used pytest to write and run unit tests for all backend Azure Functions. Each core file, including the Excel processing pipeline, Blob triggers, and ingestion logic, was individually tested. The final test coverage for the backend reached 95%, with multiple files achieving full 100% coverage.

📸 Backend coverage screenshot:

Frontend Testing (TypeScript)

On the frontend, we used Jest to create unit tests for our key logic and components. Tests were written for our Excel parser and upload logic, and we ensured that all functions behaved correctly even with edge cases.

The frontend code achieved 92.52% statement coverage and 100% function coverage, with only a few uncovered branches due to conditional rendering that was less critical for unit testing.

📸 Frontend coverage screenshot:

Integration Testing

To validate the end-to-end functionality of our data processing pipeline, we developed a Python-based integration test script instead of relying on manual uploads. This script simulates a complete run of the system—starting from a document upload, through all intermediate steps, and finally verifying the generated outputs.

1. Programmatic Upload to Blob Storage

The script uploads Chanel UK Billed.xlsx directly to the subcontractor-documents container using the Azure SDK. This initiates the automated pipeline.

2. Automatic Conversion to CSV

The Azure Function reads the uploaded Excel file and converts it into CSV format, which is stored in the csv-conversion container.

3. CSV Triggers Ingestion Function

Once the CSV is uploaded, a second Azure Function is triggered. It:

Parses the CSV with pandas
Ingests the data into Azure Data Explorer (ADX)
Generates a formatted Excel report using openpyxl
Uploads the report to the summary container

4. Excel Output Verification

The test script downloads FormattedAnnualBudget.xlsx and validates:

✅ Presence of sheets: F&B Monthly, FSH&EW Monthly, W&FJ Monthly
✅ Data population and formatting
✅ Totals and visual styling correctness

🛠️ Logging & Monitoring

Azure Application Insights and Function logs were used to:

Trace the full execution pipeline
Confirm authentication and token acquisition
Detect issues like ingestion failures or corrupted uploads

🧪 Error Scenarios Tested

We tested the following failure conditions:

❌ Duplicate uploads: Re-uploads using overwrite=True were handled gracefully
❌ Missing intermediate files: The script uses wait_for_file() to detect timeouts and raise errors
❌ ADX token failures: Manually revoked tokens led to proper logging and halted ingestion
❌ Invalid Excel formatting: Our validate_excel() method checks for expected sheet names and formatting

✅ Results

All stages of the pipeline were successfully validated:

✅ Input files were correctly transformed
✅ CSV and Excel files reached their correct destinations
✅ Excel reports were accurately generated and styled
✅ All errors were logged and handled without data loss

This Python-based testing approach ensures repeatability, automation, and confidence in future updates.

Performance Testing

For performance testing, we focused on evaluating the execution speed of our Azure Functions responsible for processing and ingesting Excel files. Specifically, we benchmarked:

🔁 Excel to CSV conversion
📊 CSV ingestion into Azure Data Explorer (ADX)

We utilized pytest’s benchmark fixture to measure function performance across different file sizes. The chosen sizes—100 KB, 115 KB, and 125 KB—were selected to represent a realistic range based on our actual sample files (86 KB and 106 KB).

📈 Benchmark Results

The raw performance benchmarks captured from pytest are shown below:

To visualize the performance trends more intuitively, we also generated charts showing the time taken for each function across varying file sizes.

🧾 Process Billed File – Performance Chart

The billed processing function had the slowest performance, likely due to the number of sheets processed and the significant data cleaning involved. While larger files show a slight increase in processing speed, the performance variation remains minimal.

📘 Process Budget Tracker – Performance Chart

The budget tracker function performed faster overall, with negligible performance differences between the various file sizes.

🧪 Ingest to ADX – Performance Chart

The blob ingestion to ADX function was the most performant. Interestingly, the 115 KB file took slightly longer to ingest compared to larger files, suggesting possible optimization or network variation influencing the results.

These performance tests ensure our pipeline remains efficient under real-world workloads and help us detect potential bottlenecks for larger deployments.

User Acceptance Testing

Validation of the application's functionality and usability in real-world scenarios by our target user base.

🧑‍💻 Simulated Testers

To replicate real-world usage and gather early usability insights, we conducted simulated testing sessions with students acting as stand-ins for the actual Chanel Media users. These testers represented varying levels of technical proficiency.

📊 Media Executive: Uploads budget and billed reports, uses dashboard for reconciliation.
💼 Finance Analyst: Verifies monthly expenditures and investigates data anomalies.
🧪 New User: Explores the platform without training to assess onboarding clarity.

Testers were observed while completing tasks and their feedback helped improve user experience and functionality.

🧪 Test Cases

We defined a series of acceptance test cases aligned with the core FinSync workflows:

Test Case	Objective	Steps	Expected Result
Upload Billed Excel	Verify upload & conversion pipeline	Upload sample billed file to platform	File is converted to CSV and processed
View Dashboard	Ensure correct data visualization	Navigate to dashboard post-ingestion	Graphs display accurate, updated data
Download Final Excel	Confirm Excel output generation	Download Excel report from summary container	Report contains formatted sheets with totals
Responsive UI	Test mobile & tablet compatibility	Open app on different devices	Layout adapts properly to screen size
Error Handling	Validate system response to bad input	Upload a corrupted file	App rejects file and shows error message
Navigation & Usability	Test onboarding and intuitive UI	Use app with no prior instructions	User completes tasks and rates experience

📬 Feedback from Testers and Project Partners

To collect comprehensive feedback, we shared a Microsoft Form with the Chanel Media team and other external testers. This included access to our platform, clear instructions, and example files containing dummy data to protect confidentiality.

The feedback form—titled FinSync User Feedback—was divided into four evaluation areas:

Tool Evaluation — assessing ease of use, feature effectiveness, and reliability
UI Evaluation — reviewing the intuitiveness, design quality, and responsiveness of the interface
Automation & Integration — evaluating how well the system automates data flow
Reporting & Visualization — analyzing dashboard effectiveness and data filtering capabilities

Testers provided qualitative and quantitative feedback on how well FinSync meets operational needs and expectations. These insights were used to improve functionality and ensure the platform was ready for deployment.

The full form can be accessed here.