Businesses generate vast amounts of data every single day, mostly so-called dark data or information that go unused. Long before the current pandemic-driven e-commerce boom, a search technology firm estimated that 7.5 septillion gigabytes of data were generated by businesses worldwide each day.
That number has reached stratospheric levels in today’s environment. In the retail space, customer feedback data has been growing in leaps and bounds over the past decade as newer digital tools are deployed for people to socially interact — think Twitter, Snapchat and Instagram, to name a few. More consumers are using apps and various digital tools to communicate with customer representatives, often to express their satisfaction with products they like or displeasure with those that fail to meet their expectations.
The massive data flow is a constant source of worry for cybersecurity experts, but their business counterparts also have other concerns: how to mine the vast information for valuable insights (such as consumer behavioral patterns) that can be used to enhance the organization’s digital transformation journey and overall customer experience. Increasingly, gathering customers’ digital sentiments has become essential for retailers when they are making operational decisions around market expansion, product development, brand enhancements, risk mitigation, fraud detection and inventory management.
“Retailers are increasingly relying on customers’ digital sentiments to make operational decisions around market expansion, product development, brand enhancements, risk mitigation, fraud detection and inventory management.”
The client needed to understand the nature and magnitude of customer sentiments received daily and formalize a process to address them. Improving customer satisfaction was the ultimate goal. Additionally, the team had to work fast and efficiently.
Recently, a luxury apparel company engaged Protiviti on a customer issue analysis project that involved analyzing emails and Instagram chats. The primary task for the apparel company was to understand the nature and magnitude of customer sentiments received daily. Second, it needed to devise a process of categorizing these issues so they can be easily addressed. The ultimate goal was to improve customer satisfaction.
Urgency was a factor in the project timeline. Specifically, because the company generates about 50% of its revenues in the fourth quarter, traditionally its peak season, the company wanted the project to be completed during the low season and before the rush of business activity at the end of the year. In other words, the engagement team had to work fast and efficiently.
However, before the team could start on the project, it needed to build proprietary intelligent tools to sort and analyze the data, a highly technical process that requires testing and calibration to fit the organization’s specific needs — which, if performed incorrectly, could result in lost time and money.
The team built an artificial intelligence and machine-learning (ML) tool or algorithm for the project. Here’s how it worked: The algorithm was pretrained on millions of English words, with the capability to understand the general meaning of texts in many different contexts. It was also designed to group similar concerns into various category levels to allow the company to identify the root causes of persistent or common issues.
With this powerful tool in hand, the engagement team began the process of extracting and analyzing more than 100,000 Instagram chats sent by the customers. The team immediately identified and flagged a number of key issues related to shipping, tracking, open payments, loyalty programs, and website challenges. Then, the issues were subsequently sorted across 34 different categories.
The apparel company used insights from the analysis to enhance critical processes and improve inventory planning to ensure proper delivery, reduced costs and more personalized services for customers.
Take this example of an email text message from a customer: “The leggings I ordered were damaged and had some torn snags on them.” In this case, the module picked up the keyword “damaged” during its initial scan, prompting the message to be correctly labeled as a product-quality issue and then subsequently recategorized as a return-or-exchange issue so it could be routed to the appropriate department for a quick resolution.
After categorization, a root cause analysis was conducted to understand the frequency of specific issues. The root cause issues were determined to fit into three broad categories: website features, inventory management and delivery and logistics.
Using the insights from the analysis, the apparel company was able to begin critical process enhancements, including rebuilding certain website features to be user friendly, updating the return portal to eliminate frequent glitches, improving the inventory planning schedule to align with the website to reflect current product availability, and reviewing third-party shipping vendors and shipment processes to ensure proper delivery and reduced costs.
Now, the company clearly has a better understanding of customers and their views on its products. The knowledge has allowed the company to develop more personalized services for customers and forge deeper connections with them. Finally, going through the process of analyzing customer feedback data has improved the algorithms used, which means they can be recalibrated to perform more effectively during the next go-around.