Leveraging Sentiment Analysis to stay nimble

Leveraging Sentiment Analysis to stay nimble

Recent market research based on a large of group online shoppers revealed that more than 90 percent of them use reviews and ratings as one of the key drivers for their buying decisions, and 84 percent trust online reviews as much as a personal recommendation. More interestingly, more than 40 percent of them view reviews and rating as the most important factor to purchase a product online. With the continuous high-speed penetration of online shopping, for the year of 2018, about 30% of the overall U.S. apparel sales happened online. These numbers tell us that the ability to collect, interpreting and quickly reacting reviews will significantly affect the success of business nowadays, and will be increasingly critical in the future.


Percentage on U.S. apparel sales made online from 2015-2017. Source: Internet Retailer.

Sentiment analysis
In the machine learning world, natural language processing (NLP) is a trending technology for the past a few years, and widely used in various industries. Sentiment analysis is one of the major applications for NLP, and is able to interpret customers’ words into a standard form of sentiment score, as well as to tell the keywords the customers are talking about based on a large set of product reviews, feeds or comments from social media, voice records from call center etc. By implementing sentiment analysis, apparel retailers can stay nimble for the product design and marketing strategy, and quickly identify any positive or negative signs of their brands.

Brand Sentiment
In the apparel industry, for lots of companies, brand accounts for the largest percentage of their product price. Also, a brand is the most important factor for a company to maintain a sustainable business in long terms. Most companies fail on their brand strategy rather than a product, due to the nature of the business that there is rarely a barrier to manufacture a product. More specifically, these companies are either not quick enough to capture the red flags on their brand or in lack of awareness about their competitors.

A comprehensive brand sentiment analysis can help a company stay nimble to immediately adjust branding and product strategy. By holistically collect the data from social media, customer services, and other sources, and then feed into the sentiment analysis AI model, one can immediately predict the overall sentiment score of the brand and look at the trending line based on the historical results, also compare the competitors. In addition, based on the aggregated data, retailers quantify the overall customer engagement and create a word cloud to identify what is most positive and negative things customer are talking about.


A sample of historical sentiment score.

A sample of retailer words cloud.

Product Sentiment
Research shows that most apparel retailers accept more than 70 percent of their apparel products as failed products. Fortunately, the rest of 30 could keep them in business and stay profitable. As was mentioned at the beginning, reviews and ratings are the top key drivers of customers’ purchase decision, being able to perform sentiment analysis from a review data lake and tokenize the key drivers of a product will tremendously help the retailers quantify sales volume, manage inventory and speed up the design process.

Applications of Sentiment Analysis in Apparel Retailers
Very few companies have the bandwidth to build sentiment analysis models as it involves a large amount of “heavy lifting” work including data collection, preprocessing, feature engineering and model constructions. Also, it’s costly to hire data scientists to support the work. At Neural Trend, we have our own data lake collected or bought from social media and data vendors. We also provide on-boarding services to help the client organize and pre-process data. In addition, our flexible model inventory can build bespoke sentiment analysis model that particularly fits different clients’ needs.