Client Story
Automating Feedback Analysis with Artificial Intelligence
Project Overview
Our approach involved developing a universal language model tailored to process inputs in five languages: Estonian, Latvian, Lithuanian, Russian, and English. Following this, we devised a sophisticated classification algorithm. This algorithm efficiently sorted the incoming feedback into relevant categories, ensuring that it reached the appropriate store and personnel promptly. Remarkably, this method conserved approximately 1000 man-hours.
Our efforts in accelerating the feedback analysis process for Rimi Baltics were multi-faceted. By implementing artificial intelligence, we enhanced the speed and simplicity of handling feedback. The result? Timely and efficient management of responses, which saved an impressive amount of labor time.
AI Helps Save Hundreds of Man-Hours and Quickly Identifies Problem Areas
Rimi Baltics, a major player in the retail sector across the Baltic nations, with 291 stores and a workforce exceeding 12,000, faces a considerable challenge: managing vast volumes of daily customer feedback, both positive and negative. While positive feedback affirms well-executed services, requiring minimal direct response, addressing negative feedback promptly is paramount to maintaining high standards of customer service.
The challenge for a business of Rimi Baltics' scale is the sheer volume of feedback, which historically took up to 1000 man-hours annually to process. This volume could delay timely responses to negative feedback, impacting customer experience.
Recognizing the need for an innovative solution, Rimi Baltics opted for an AI-driven approach. They sought expertise from Intelex Insight, a leader in the field, to develop a bespoke algorithm that would revolutionize their feedback management process.
The Algorithm Developed from 25,000 Feedback Forms
Central to machine learning solutions is the quality of training data. Faced with the challenge of operating in a multilingual market, Rimi Baltics needed a versatile language model that performed consistently across five languages: Estonian, Latvian, Lithuanian, English, and Russian. To this end, we leveraged 25,000 customer feedback forms in these languages as our foundational dataset.
Our development process was meticulously planned. Firstly, we computed semantic vectors from the text of the 25,000 feedback forms, which were instrumental in creating a robust universal language model. Following this, we rigorously analyzed a range of classification algorithms, selecting the one that most effectively met Rimi Baltics' specific requirements. The next step involved training and validating this algorithm with our dataset, ensuring its accuracy and reliability. Upon achieving satisfactory results, we seamlessly integrated the algorithm into Rimi Baltics' business processes.
The Algorithm's Development Involved 5 Key Steps:
- Calculating semantic vectors from texts in five different languages and crafting a universal language model.
- Analyzing and identifying the optimal classification algorithm.
- Diligently training and validating the algorithm on the training data.
- Transferring the algorithm to the client for implementation within their operational framework.
As a result, Rimi Baltics acquired a sophisticated machine learning algorithm capable of categorizing feedback in five languages into 34 distinct categories. This system ensures that feedback in each category is directed to the responsible personnel, facilitating prompt and relevant responses.
The Impact: Enhanced Efficiency and Customer Satisfaction
The introduction of the algorithm marked a significant shift from the previous manual processing of feedback forms. This automation not only eliminated the need for labor-intensive categorization into 34 groups but also liberated Rimi Baltics' employees from spending about a thousand hours each year on this mundane task.
The most significant gain, however, was the acceleration in the feedback analysis process. The automated system ensures that critical feedback reaches the respective department much faster than before. This efficiency in identifying and addressing issues has led to a notable improvement in customer satisfaction.
In choosing to collaborate with Intelex Insight for an AI-driven solution to analyze incoming feedback, Rimi Baltics made a forward-thinking decision.
A Smooth and Collaborative Journey to Success
The entire project was characterized by a seamless and productive collaboration between Intelex Insight and Rimi Baltics. Rimi Baltics supplied the necessary feedback data, including specifics such as store locations, a comprehensive product group catalog, desired categories, and access to previously manually classified feedback.
At Intelex Insight, our approach was thorough. We computed over 500 semantic language vectors and pinpointed the most effective classification algorithm. This algorithm underwent a rigorous cycle of training, testing, and validation. Along with the model, we provided Rimi Baltics with detailed documentation, supporting them in integrating the new system into their existing business processes.
Upon completion, Rimi Baltics had in their hands a powerful tool that has been saving them nearly a thousand man-hours annually since 2022. The tool has not only expedited the identification of issues and the dissemination of information to relevant personnel but also played a pivotal role in elevating customer satisfaction levels.
Every year, Rimi Baltics processes over 20,000 feedback forms. The introduction of the algorithm has significantly reduced the time and manpower required for this task. The investment in the algorithm's development and implementation paid off within its first year of operation, as the automated system took over the task efficiently. This has allowed employees to redirect their efforts towards more strategic, value-adding activities, further benefiting the company.
