How companies use sentiment analysis to ensure strong brand management
“Sentiment” is a rather intriguing concept. It can mean attitude, feeling, bias, view, thought, and even something as deeply felt as emotion. Sentiment analysis that utilizes AI and machine learning has become a powerful tool for companies to understand how their customers and/or potential customers feel about their company. It can also be used for competitive analysis to take the temperature of a rival’s products or services. “When you can’t convince them with intellect, persuade them with sentiment,” is author Amit Kalantri’s recommendation. It’s a concept businesses should keep in mind when trying to ensure strong brand management, just not quite in the way he had intended it.
Online Marketplaces
In their article How Artificial Intelligence and Machine Learning Can Impact Market Design, Paul Milgrom and Steven Tadelis state that AI and natural language processing can be used in today’s online marketplaces to help foster more trustworthy environments, which should lead to better customer buying experiences. The writers believe today’s online marketplaces act in similar way to the institutions that emerged in the medieval trade fairs of Europe that fostered environments where customers felt comfortable doling out their hard-earned money to sellers who were, in many cases, complete strangers — sometimes ones who were just passing through town.
For Milgrom and Tadelis, online marketplaces like eBay, Taobao, and Airbnb have experienced exponential growth since their inception because they provide “businesses and individuals with previously unavailable opportunities to purchase or profit from online trading.” The remarkable successes of these online marketplaces came as a surprise to many because of the hazards of anonymous trade and asymmetric information, argue Milgrom and Tadelis. Specifically, how can strangers who have never transacted together before be willing to trust each other while exchanging a tangible product for money, ask Milgrom and Tadelis. There has to be confidence on both sides of the market for parties to agree to do business and for a marketplace to succeed.
eBay’s early success is often credited to its groundbreaking feedback and reputation system, which has since been replicated by practically every other marketplace that followed it. The problem with these markets is that buyer-generated feedback doesn’t accurately reflect the buyer’s actual performance. User-generated feedback systems are often biased, suffer from ‘grade inflation,’ and can easily be manipulated by bad faith sellers, say Milgrom and Tadelis. With a 99.4% average positive rating for sellers on eBay, it’s clear that this rating system is highly flawed, if not being outright manipulated by deceitful sellers.
Recognizing this, Milgrom and Tadelis wondered how an online marketplace could use its treasure trove of data to measure the quality of a sales transaction and predict which sellers would provide the best service to their buyers? After all, these online marketplaces aren’t short of data. Surely it was possible to leverage the millions of transactions, searches, and browses that occur on these sites every day to create a more trustworthy environment. Milgrom and Tadelis believe that AI, specifically Natural Language Processing (NLP), could be applied to these marketplaces to help create a more dependable and better consumer buying experience.
“One of the ways that online marketplaces help participants build trust is by letting them communicate through online messaging platforms,” explain Milgrom and Tadelis. On eBay, buyers question sellers about their products, while Airbnb allows potential renters to query hosts about property details that might not be in the original listing.
Using NLP, i.e., the attempt to extract information from the spoken and written word using algorithms, marketplaces can utilize sentiment analysis to predict the kind of features that customers might like, as well as get instant reactions to things that they don’t. However, Milgrom and Tadelis go one step further, arguing there may be subtler ways to apply AI to manage the quality of marketplaces. The messaging platforms are not only restricted to pre-transaction inquiries, but they also provide both parties with the ability to send messages to each other post-transaction, state Milgrom and Tadelis. The obvious question that emerges for Milgrom and Tadelis is, “How could a marketplace analyze the messages sent between buyers and sellers post the transaction to infer something about the quality of the transaction that feedback doesn’t seem to capture?”
The question is answered in Masterov et al.’s paper Canary in the e-commerce coal mine: Detecting and predicting poor experiences using buyer-to-seller messages: NLP was used on internal eBay data to identify transactions that went bad because a buyer indicated he or she was unhappy. These included claims that items received were not as expected, items were not as described, or the buyer left negative or even neutral feedback.
The simple NLP approach Milgrom and Tadelis used created a ‘poor-experience’ indicator as the target while the messages’ content was considered the independent variable. A standard list of negative words — ‘annoyed,’ ‘dissatisfied,’ ‘damaged,’ or ‘negative feedback’ — was used to identify a message as negative, note Milgrom and Tadelis. Messages lacking these terms were considered neutral. The researchers grouped transactions into three distinct types: “(1) No post-transaction messages from buyer to seller; (2) One or more negative messages; or (3) One or more neutral messages with no negative messages,” say Milgrom and Tadelis.
Masterov et al. then constructed a novel measure of seller quality based on the idea that sellers who receive a higher frequency of negative messages are probably bad sellers. According to Masterov et al., a measure is “calculated for every seller at any point in time using aggregated negative messages from past sales, and the likelihood that a current transaction will result in a poor experience.” This simple use of NLP can be a powerful indicator of sellers who are destined to create poor experiences, which ultimately hurts the marketplace brand. It’s also a better indicator than one inferred from the highly inaccurate and wildly inflated feedback customer data, which is often manipulated by sellers agreeing to refund money on the condition the buyer doesn’t leave a bad review. “The key is that there is information in communication between market participants, and past communication can help identify and predict the sellers or products that will cause buyers poor experiences and negatively impact the overall trust in the marketplace,” conclude Milgrom and Tadelis.
Creating a market for feedback
Another problem with marketplace customer feedback forums is their utter lack of use. Few buyers ever bother leaving feedback and the ones who do aren’t always motivated to do so for the right reasons. In fact, Milgrom and Tadelis argue this should be expected because leaving feedback “is a selfless act that requires time, and it creates a classic free-rider problem.” There’s also a ‘cold start’ problem here, the writers contend because new sellers with no feedback face a high barrier-of-entry because potential buyers will naturally choose sellers who have good ratings.
Lingang Li and others address this problem in their paper Buying Reputation as a Signal of Quality: Evidence from an Online Marketplace by showing how a unique and novel implementation approach is being used by the huge Chinese marketplace Taobao. Here, sellers are allowed to pay buyers for feedback. Milgrom and Tadelis acknowledge it might be concerning to let sellers pay for feedback as sellers will, naturally, choose to pay for good feedback while also trying to suppress bad feedback, which does little to promote trust. However, Taobao implemented a clever use of NLP to negate this problem, say Milgrom and Tadelis; the NLP, not the seller who pays for the feedback, decides whether the comment is relevant. The reward to the buyer for leaving feedback is handled by the marketplace, and was paid out for informative feedback rather than just for positive feedback, note Milgrom and Tadelis. Taobao NLP’s algorithm measured the quality of the feedback, ensuring key product features were described, explains Milgrom and Tadelis.
Li et al. analyzed data from the period where the Rebate-for-feedback (RFF) system was featured, “and confirmed that first, as expected, more feedback was left in response to the incentives provided by the RFF feature.” Li et al. also discovered that “the additional feedback did not exhibit any biases, suggesting that the NLP algorithms used were able to create the kind of screening needed to select informative feedback.” In conclusion, Li et al. argue the novel market for feedback solved both the free-rider and cold-start problem, two issues that so often hamper growth in online marketplaces.
Conclusion
Convincing with NLP sentiment rather than intellect might not have been exactly what Kalantri meant in his quote but it does make a lot of sense when looked at from a business standpoint. Customers often don’t want to spell things out in easy-to-understand language therefore a lot of looking between the lines is required to glean the meaning of a comment made on a social media platform.
Milgrom and Tadelis are correct, there has to be confidence between both buyers and sellers for a marketplace to succeed. Digging a little deeper into the comments of both buyers and sellers can prove quite fruitful in understanding which sellers are good and which are bad. This information should filter through to a seller’s ranking positions, the good and honest sellers should always be promoted over the bad ones, and NLP allows a marketplace to utilize information that cannot be manipulated, it’s the very lifeblood of sales. Understanding sentiment can boost the trust in the overall marketplace. However, we should always remain vigilant and not forget, as Traci Chee warns us, that “sentiment makes fools of us all.”
