Data Sources Analyzed: Net promoter score surveys, emails, claims, product reviews, store reviews, Google Business Profile reviews, web site reviews
Type of Analysis: Omni-channel customer experience analysis, product quality monitoring, innovation watch, automatic moderation and alerts
Volume: Millions of documents analyzed continuously every year
Languages: French and Spanish (and some in Chinese and English)
4 data analysts
1,000 readers (primarily product owners, call center managers and store managers)
What is your role?
I’m a product owner on the Data Science/Artificial Intelligence team. In particular, I manage all the projects and use cases related to the offer, the design, and the creation of the products. We are currently working on several subjects such as the semantic analysis of product reviews, store reviews, contacts into the Customer Service center and more. Some projects are in the proof of concept phase and others are in production, for example related to images processing, price elasticity, and graph theory.
Why did you put into place a solution to analyze customer feedback?
At the beginning, the Customer Service center needed a solution for quantifying the broad themes that generate dissatisfaction. Indeed, every Customer Service representative was able to determine the main aggravations felt by our customers, but no one could quantify them or track their evolution over time. So we rapidly decided the product leaders needed a solution. In fact, they were reading and answering customer reviews, but did not have a tool to analyze them quickly and simply. We noticed that the reviews are a very valuable raw material, more qualitative, and useful for designing new products and driving innovation.
Why did you choose NETVIBES Proxem Studio solutions for your projects? Could you describe your thoughts about the implementation?
One of the advantages of NETVIBES Proxem Studio was its capacity to natively analyze a large number of languages. We submitted a dataset to measure the benefit of opting for an external company rather than developing a solution internally. On this dataset, the Proxem Studio team managed to go further in the analysis than we did, in just a few days (it took several months internally). Furthermore, we didn’t have the right profiles to complete the project internally.
The collaboration with the Proxem Studio team has been positive and effective. All of my requests for evolutions (mostly about the reporting) are always taken into account with priority. I meet and speak regularly with the team and that reinforces our relationship even further.
Could you describe the results of our solutions?
Some managers use the solution to identify action plans in order to improve consumer satisfaction. The semi-automatic moderation of the reviews project allows the company to save money. Otherwise, the increase in the number of reviews would have forced us to hire extra people to manually moderate the reviews. This semi-automatic process helps us to absorb the massive increase in reviews without recruiting.
70% of reviews moderated automatically
What are the next steps? What would you improve?
We plan to start working on SpeechToText and therefore to analyze, in addition to the written reviews of our customers, their phone conversations. We also plan to analyze more types of products based on customer reviews.