Evaluation
Reflecting on our project’s impact and future growth
Achievements
MoSCoW Requirements:
| ID | Requirement | Priority | State | Contributors |
|---|---|---|---|---|
| 1 | Automated Parsing of Spreadsheets for Budget Report Generation | Must Have | ✓ | Anika, Ines |
| 2 | Export Functionality for Processed Budget Report | Must Have | ✓ | Ines |
| 3 | Invoice Uploading with Automated Data Extraction | Should Have | ✓ | Anika, Ines, Daren |
| 4 | Real-Time Expense Reconciliation from Multiple Data Sources | Should Have | ✓ | Anika, Ines |
| 5 | Interactive Dashboard for Data Visualization | Could Have | ✓ | Anika, Ines, Daren |
| 6 | AI-Powered Chatbot for User Assistance and Engagement | Could Have | ✓ | Daren |
Non-Functional Requirements:
| ID | Requirement | Priority | State | Contributors |
|---|---|---|---|---|
| 1 | High performance | Must Have | ✓ | Anika, Ines, Daren |
| 2 | Robust security measures | Must Have | ✓ | Anika |
| 3 | Reliable data processing | Must Have | ✓ | Anika, Ines |
| 4 | Integrity checks | Must Have | ✓ | Anika, Ines |
| 5 | Scalability to accommodate growing data volumes and user base | Could Have | ✓ | Anika, Ines, Daren |
Known Bugs
| ID | Bug Description | Priority |
|---|---|---|
| 1 | Time taken to generate spreadsheet increases as more data is uploaded | Medium |
| 2 | Pie chart appears too large for card | Low |
| 2 | Pie chart text overlaps if there are many categories | Low |
Individual Contribution
System Artefacts:
| Work Packages | Anika | Ines | Daren | Aziret |
|---|---|---|---|---|
| Research and Experiments | 20% | 20% | 60% | 0% |
| UI Design | 33.3% | 33.3% | 33.3% | 0% |
| Coding | 40% | 40% | 20% | 0% |
| Testing | 33.3% | 33.3% | 33.3% | 0% |
| Overall Contribution | 33.3% | 33.3% | 33.3% | 0% |
Website Report:
| Work Packages | Anika | Ines | Daren | Aziret |
|---|---|---|---|---|
| Website Template and Setups | 0% | 100% | 0% | 0% |
| Home | 0% | 100% | 0% | 0% |
| Video | 42.5% | 42.5% | 15% | 0% |
| Requirements | 15% | 85% | 0% | 0% |
| Research | 100% | 0% | 0% | 0% |
| UI Design | 0% | 100% | 0% | 0% |
| System Design | 0% | 0% | 100% | 0% |
| Implementation | 0% | 0% | 100% | 0% |
| Testing | 25% | 75% | 0% | 0% |
| Evaluation and Future Work | 95% | 0% | 5% | 0% |
| User and Deployment Manuals | 0% | 0% | 0% | 100% |
| Legal Issues | 0% | 0% | 10% | 90% |
| Blog and Monthly Video | 42.5% | 42.5% | 0% | 15% |
| Overall Contribution | 35% | 35% | 29% | 1% |
Critical Evaluation
User interface and experience
Our goal was to create a clean, sleek looking user interface matching Chanel?s black and white minimalist aesthetic. We ensured that upload instructions were clear, and toast notifications informed users when data processing was occurring. General feedback from testers and clients was positive, with users stating that the website flow was intuitive and provided an overall enjoyable experience.
Overall, a rating of Good was given to our UI and UX.
Functionality
The data matching part of the project was tested with different sets of data to check for accuracy, ensuring that it was reliable. The function apps were also extensively tested to ensure that the triggers worked as expected consistently. There were no reported issues regarding functionality with our web app.
We rate our functionality to be Very Good.
Stability
As a team, we completed extensive unit testing, achieving over 90% test coverage for both the backend and the frontend. In addition to this, we wrote integration tests to check the interactions between the different components of the web app, and wrote performance tests to check how different loads are handled by the backend.
Overall, our Stability is Very Good.
Efficiency
Our performance tests found that data ingestion into ADX remains fast as file size increases, and the speed of conversion from Excel to CSV does not change much as file size increases. However, the excel generation function slows down a non-negligible amount as the amount of data in ADX increases.
We give a rating of Adequate to our efficiency.
Compatibility
The web app can be run locally, as well as deployed on a public static web page. It is compatible with most browsers, allowing for users with different devices to access the page.
We rate our compatibility to be Good.
Maintainability
Each feature is separately housed in an Azure function, allowing for bugs to be easily identified and fixed without interrupting other parts of the project. The project is also fairly well-documented, allowing our clients to understand how the app works.
Overall, a rating of Good was given to our Maintainability.
Project management
As a group, we used a variety of tools to manage the progress of our project. We communicated on various platforms: Teams, Whatsapp and email in order to keep track of what tasks had been completed and what there was left to do. We also had weekly meetings to update each other of our progress and assign new tasks. Lastly, we also used a Gantt chart to more accurately track the timeline of our tasks.
Overall, our project management was Very Good.
Future Work
Potential improvements:
Data Matching
One major improvement that could be made is to expand the accepted Excel sheet structures to allow users to insert more data to be summarized. This would involve customizing the cleaning and handling of the new spreadsheet type and finding a way to merge the pre-existing sheets with the new sheet. In addition to this, more accepted file types, such as PDFs could be added, which could be done by using OCR to extract the relevant text for processing.
Query page
- Improving the display of thoughts, action inputs and outputs with a better user interface in the dropdown menu.
- Implementing proper multi-turn conversation by passing a list of messages rather than a query to the Agent
- Give users the option to connect to their own inference server rather than Azure OpenAI Service (e.g Ollama)
- Displaying a relevant DataFrame preview alongside the query results