Sprint Retrospectives

In this page, we will reflect on the team's work during each sprint, documenting what went well, what went wrong, what is still a problem, and a strategy for improvement after each iteration.

Sprint 1

What Went Well

• Jumpseller Connection: the connection itself went smoothly, however, we encountered issues when retrieving product reviews: the response data contained inconsistent content, making it incompatible with our database structure and preventing proper population, as a result, our database and populate had to be corrected and updated manually.

• Surprise ML Models: integrating the Surprise service was straightforward, however, because the Jumpseller database contains very few products, we were required to create mock data in order to provide the ML models with a sufficient dataset for training and evaluation. The job trigger was also created and verified that runs every night at 2am, ensuring fresh recommendations to users.

• Frontend: the CSS and overall frontend development progressed well, and we now have several ideas for future sprints to further enhance and optimize the interface; during the integration between the frontend and the API, we faced a few unexpected errors, but we managed to identify and resolve them successfully.

What Went Wrong

• Inter Group Communication: if the Surprise ML Models give no recommended products, then the recommendations fallback to the most popular products. Another group was working on this (Best-Selling Products), so we reached out to them to understand how they were implementing it and how we could integrate their service with our own. We found out that they weren't going to do the most popular/best-selling products overall, but rather the best-selling products by category, so we would have to work on it ourselves. However, later on, mid our implementation, they corrected themselves, telling us they would implement the overall best-selling products after all, but would only finish at the end of the sprint or later, making us unable to complete one of the issues we initially commited to.

What is Still a Problem

• Lack of Users/Authentication: the lack of an authentication service will provide some issues for the future, for example, we are still unsure on how to implement the follower/followee logic for the recommendation algorithm.

Strategy for Improvement

• More organization.
• More and better communication with other groups.

Project Board

Beginning of Sprint

End of Sprint

During the sprint, we had to refine the board and remove the issue #323, since we were dependent on the work of another team in order to finalize it. Besides #323, which we are waiting for the other team input, we were able to finish all the planned issues.

Sprint 2

What Went Well

• Jumpseller connection: the API was successfully implemented, with some alterations, due to third alterations.
• Surprise ML Models: all the models implemented with no difficulties.
• Pub/Sub connection: the connection was easily done. Already connected to the reviews, only missing the orders (waiting for others to create the topic).
• Frontend: recommendations section created in the landing page. A carousel that leads to the recommendations page when clicking "See more".

What Went Wrong

Since the group that was dealing with the most bought products implemented their part to the landing page, it would be redundant if we also add the most popular products at the landing page. Therefore, when no enough data for recommended products' generation, it will appear this message:

What is Still a Problem

In our perception, this sprint went flawlessly, implementing all the issues for this stage.

Strategy for Improvement

The strategy will be to maintain our rhythm and communication, stating all the alterations and communicating with the other groups.

Project Board

End of Sprint

This sprint we were able to sucessfully complete all the issues we proposed to at the initial sprint planning.