.png)
1. Team Formation
On the 29th of September, our team got to meet for the first time, having readily arranged a communication platform where each member could talk and get to know each other.
The team quickly found out each member's background and experience, verifying that everyone has a solid background on retail/marketing software development and different professional experiences and job positions, ranging from RPA to Data Engineering, which could prove useful in the future.
Furthermore, every member shares the same passion for artificial intelligence and believes that the area of Retail can benefit from its usage, in order to provide customers with a smarter and more personalized experience in their stores of choice. The aforementioned belief served as the basis for the team's mission and it should be reflected in every project developed throughout the course.
With this in mind, when the time came to choose the team's name, Smart Retail seemed like a very appropriate choice, as it reflects the team's goal to develop smarter retail solutions.
2. Brainstorming
After the team's formation, it was time to identify the first challenge's theme. After a brainstorming session, many ideas came up and were promptly discussed with the team's inspirator, professor Tiago Pinto.
In the end, the following 4 ideas were settled as possible project themes:
Customer Satisfaction Diagnosis / Churn Prediction;
Transaction Fraud Diagnosis;
Pricing Mechanism;
Product Recommendation System.
All of these ideas had good potential, but after some consideration, the team concluded that a Customer Satisfaction Diagnosis would benefit the most from an expert system implementation, as the thought process behind a customer satisfaction diagnosis could prove valuable to store managers that may want to an explanation behind every decision.
Nevertheless, the other ideas are not forgotten and could potentially serve as a future challenge's theme, benefiting from techniques such as Machine Learning and Deep Learning.
3. Solution Design
When designing the solution, the team discussed the possibility of implementing a two-phase expert system that, besides identifying the unsatisfied customers, would provide an action plan to build the trust from those customers, preventing them from switching to a competitor. After considering the suggested action plan, the store manager can then decide if he /she wants to apply the suggested actions.
In a real-world scenario, most store managers would benefit from a user friendly graphical user interface (GUI), where they could interact with the expert system, analyse the obtained conclusions/statistics and then decide which actions to apply.
As such, the team decided to design a web application, complete with:
a Single Page Application, that provides a user friendly dashboard to interact with the expert system, accessible in any device with a web browser;
a Data Management Module, that allows access to users, customers and other data stored in the system;
the Expert Systems developed in Prolog and Drools, each with its own REST API that allows other modules to interact with them.
(Click on the slideshow above to expand the figures)
Comments