For companies that own many product lines and launch new products, understanding customers’ needs and getting quick feedback is core to their business and success. Such insights have historically been drawn from manual market research (e.g. hard-copy surveys and phone calls), but digital channels are becoming key sources of competitive intelligence and customer data. As of 2019, there were 4.5 billion internet users, nearly 2 billion digital buyers, and 3.8 billion active social-media users. This massive audience generates a significant amount of data, 80% of which is unstructured. Although organizations derive intelligence from the other 20%, the structured dataset, unstructured data can drive results if captured and processed properly.
This data can distill meaningful social sentiments at scale, which is extremely valuable when formulating business strategies. Leaders in many sectors have gained competitive advantage by successfully churning this kind of data into insights and actions such as launching new products, informing product development, and improving customer care.
For example, if a customer care center receives a complaint about the change of a product’s smell, it is difficult to assess how many total customers share the same sentiment. Similarly, in market research, a large proportion of customers who receive free samples might be overly complementary or critical of the product, while opinions from genuine online consumers might be completely different. Yet how can companies address the business impact of changes in customer sentiments if they’re unable to accurately interpret the data in a sentiment analysis?
Our Recommended Approach
To get comprehensive insights, an organization must gather all forms of customer feedback, including care-center interactions and new-product survey feedback from the company’s own portal, along with customer reviews on e-commerce sites like Amazon, user blogs, and video sites such as YouTube. This requires the company to engage in a thorough data discovery project involving data acquisition, cleansing, neuro-linguistic programming (NLP), sentiment analysis, business analysis, and insight presentation.
Organizations will also need an analytical platform to ingest these datasets and develop a brand- and product-level data model iteratively. Wipro’s Data Discovery Platform enables companies to gain insights based on raw data, accessing the platform step by step to deepen their understanding of the information and the potential responses. The process can leverage Python’s NLP data libraries as a start, and gradually build up a library of texts attuned to the brand’s features. Each step in this model feeds into the next, as shown below.
While customer sentiment can be analyzed through multiple NLP programs, customer ratings can also be a pseudo indicator. Pairing these sentiments with a reason can highlight areas of opportunity and risk for marketers.
For example, while users may seem to care most about smell, some might show favorable sentiments (“smells great”), while others show unfavorable sentiments (“strong smell”). The same can be said about features such as price, availability and other attributes. Without fully understanding the data’s nuance, organizations may make flawed business decisions.
Marketers can use a digital dashboard populated by analytically gathered insights to rank and prioritize each topic. Brand teams can also select specific features or topics to drill down further. With each selection, the proportion of positive/negative/neutral sentiment, and the corresponding list of original feedback, is pulled from the database – a process that, if done manually, would be too time consuming to be effective.
In order to analyze customer sentiment’s impact on business, organizations can compare year-to-year and quarter-to-quarter changes in key topics, sentiments, and ratings with sales. For example, brand managers who launch new products can immediately analyze the efficacy of their research programs by comparing customer sentiments from Amazon (filtering for verified purchases) against the market research sample provided by users, leading to more informed discussion about whether their product strategy needs adjusting.
New Product Insights
New product launches are stressful for brand managers, largely due to the lack of certainty about how the products will perform. To combat this uncertainty, organizations can run a dedicated sentiment analysis on new products launched over the past six months.
In our experience working with a cosmetics client, the uptake of product sales and the product ratings are consistent, while in the case of mature products (those with 1.5 to 2 years of online history) this relationship did not exist.
New product managers should carefully watch the volume of discussion in the marketplace and which features are discussed (and how) to quickly modify messages for the right type of customers. For example, if smell is the most discussed feature of the product, the brand manager can promote “new smell” as a new message. Very specific concerns and questions around how to use products are also often discovered this way, empowering brand managers to respond quickly to and gain customers’ trust.
Business Implications
Business teams can use these insights to drive internal strategies and business actions. And because they’re acquiring customer feedback from open-source datasets, they can also gain insights about key competitors. Capturing customer feedback about brands and products from digital channels can help businesses across industries make more-informed decisions regarding product R&D, supply chain, marketing and sales campaigns. With Internet use showing no signs of slowing down, it’s time for businesses to leverage 100% of the generated data, to become truly intelligent enterprises, and to empower their organization to both understand and act upon customer sentiment.
References
- Adam Hale Shapiro et al 2018: Measuring News Sentiment, Federal Research Bank of San Francisco
- Tracy Alloway 2019: JPMorgan Creates ‘Volfefe’ Index to Track Trump Tweet Impact, Bloomberg
- Teng (Diana) Ma 2014: the social impact on product launch, McCann WorldGroup
The author would like to extend special thanks to Diana Ma for her contributions to this piece. Diana has more than 15 years of experience in strategy, budget and resource allocation, planning, business operation, and performance optimization using data-driven decision science. Her work cuts across numerous industries including pharmaceuticals, consumer banking and finance, telecommunications, CPG, media, and branding.