Q&A with MGM Resorts, Pinterest and Walmart on Boosting Brand and Customer Insights through Sentiment, Emotional and Behavioral Analytics.
Other than AI/Machine Learning, the two-other areas of greatest interest among researchers this year are related in better predicting behavior by understanding emotions and the subconscious. This is where some of the most exciting techniques, including my favorite, text analytics (including Sentiment and Emotional Analytics) and Behavioral Economics intersect.
Ultimately, it’s a quest to gain clarity from ever more complex data. The Insights Exchange Network has responded to this need by putting together a series of interesting events, I attended a great one on content attribution and modeling in NYC last month.
This July EIN is hosting a West Coast event specifically focused on Sentiment, Emotional and Behavioral Analytics (SEBA for Short). Ahead of the conference which will be held July 17-18 in San Francisco CA, I reached out to a couple of the client-side speakers to understand beforehand what this type of analytics means to them.
Today on the blog I welcome, Brian Johnson, Head of Knowledge at Pinterest, Harsh Gupta, Director Voice of Customer at Walmart, and Paul Her-Sturm, Director of Content at MGM Resorts. Below is a brief informal Q&A on this interesting and important topic. One that we’ll all be learning a lot more about s year and certainly at next month’s event.
Q. Tell us a little about your experience with Sentiment, Emotional and Behavioral Analytics? For purposes of brevity lets simply call it SEBA for now.
Harsh-Walmart: I am currently leading Voice of customer for Walmart.com. Prior to Walmart, I led VoC for eBay. That kind of makes me a practitioner in the area.
Brian-Pinterest: Behavioral analysis of user behavior is a core part of my job. Pinterest is all about recommendations and personalization. We couldn’t do this without understanding our users.
We analyze engagement with content over the past 12 months, with an exponential decay for recency. You can find slide examples here:
Paul-MGM: My first exposure to SEBA was during my studies at Northwestern University, while earning a master¹s degree in Integrated Marketing Communications. There was significant emphasis placed on the importance of understanding your audience¹s behaviors, in order to determine the relative impact of your marketing efforts. However, the efficient measurement of sentiment was not as well developed in the mid-2000s as it is now. The social networks were just taking off in 2007. Therefore, we were using data sets of prior purchases, as well as qualitative research, to determine purchase intent.
That was the extent of our behavioral and emotional analytics. With the relatively recent development of efficient text analytics and machine learning algorithms, there is increasing emphasis on sentiment analytics in my role for social media optimization and reputation management. It is no longer enough to know your engagement on a social post, we must now understand the ratio of positive to negative engagement, as well as the purchase intent as potentially revealed by said sentiment and emotion.
While our company is well aware of these needs, we are still working to implement the tools and report templates to effectively activate on the data now swirling around us on the internet.
Q. What does SEBA mean to you?
Harsh-Walmart: SEBA is an understanding of drivers behind the transactional data. Transactional interaction between the customer and the organization is defined by need. SEBA defines the desire to interact with the brand for that need.
Paul-MGM: I would define SEBA as the analysis and interpretation of signals customers display throughout the entire continuum of touch-points with an organization. From statements made on social sites to purchase behavior to post-purchase reviews that impact future potential customers, the complete analysis of these signals drives valuable marketing information and strategy for an organization.
Q. How is SEBA being used at your company?
Harsh-Walmart: SEBA is extensively used to understand how Walmart is relevant for customers both at retail and online front. We have regular cadence to understand Voice of Customer at an company level. SEBA is also closely monitored for every product launch and new strategic initiative.
Paul-MGM: SEBA, in its many forms, is being used throughout our organization, from direct marketing, where we use gaming and purchase data to inform predictive analytics on purchase intent and offer reinvestment to shift behaviors, to social media crisis communications and reputation management, where we use sentiment analysis of social data to determine reputation impact, in coordination with reservation pace for hotel rooms.
Q. What do you see as the major weaknesses in SEBA currently?
Brian-Pinterest: Pinterest has 175M users curating a graph of 100B pins on 2B boards. The kind of recommendations we can make based on this user activity are unbelievably awesome!
For an example, I like this one, slide 11:
Harsh-Walmart: Still relies too much on text analytics. In a world where not many people are willing to take a survey and write feedback, we need to figure out new ways of getting relevant feedback which can be used to analyze customer sentiment.
I am not saying text analytics is not effective. It is effective, you bet. My point is, in this world of multiple ways of interacting with brand, pics, speech, etc. analytics for sentiment need to move beyond text.
Paul-MGM: Sentiment is still not efficiently and consistently determined by the machine-based algorithms to accurately classify social data without the extensive assistance of humans. With regard to the Las Vegas market, the use of slang, sarcasm and alternative definitions means that our tool has an extremely difficult time determining positive versus negative sentiment, let alone a more complex sense of sentiment. An oft shared example of challenging language that we review is in regard to the word “fight.” Our tool is not able to determine whether the phrase references a dangerous fight on The Strip or the fight of the century occurring in the ring at MGM Grand Garden Arena. And what about a fight at the fight? Thus, a lot of our social sentiment requires human review.
Q. What do you see as the major strengths?
Harsh-Walmart: SEBA highlights themes which may be missed in transactional data.
Paul-MGM: The major strength of adding sentiment analysis to behavioral analysis allows marketers to provide a framework for why behaviors might be changing, or how they could be changing moving forward. The key is tying the two pieces together for deeper insights.
Q. How do you think SEBA will change in the future?
Brian-Pinterest: Everyday people tell us what they need. If your company is not listening you will be run over by someone else who is listening.
Harsh-Walmart: AI has a major role to play in future. AI will play a part in asking questions which are relevant as well
Paul-MGM: Automation will continue to evolve quite rapidly to enable more efficient and consistent reporting and analysis with higher confidence in the data being presented. In addition, marketers will find better ways to tie sentiment analysis to purchase data.
Q. Are there any tips you would give others who are just beginning to investigate this?
Brian-Pinterest: Listen to your users by watching what they do and do not do.
Harsh-Walmart: Well thought out strategy for Listen –> Analyze –> Act –> Measure –> Listen
Paul-MGM: Start with a targeted project and timeline. There is a lot to cover and it can quickly become overwhelming. You will also need to identify the correct tool(s) to gather and classify the data you anticipate analyzing. Each of the tools have their own strengths and weaknesses that must be reviewed carefully.
Q. If you could wave a magic wand and improve your current SEBA efforts in some way, what would you wish for?
Harsh-Walmart: Could use Facebook mind reader for every customer which comes to the site J
Paul-MGM: I would ask for two resources. The first would be a dedicated FTE to monitor, guide and analyze the social sentiment data that is sometimes overwhelming for a relatively small team with nearly 30 brands and 78 social channels to analyze. The second would be greater integration with our customer and purchase data, so that we can tie the sentiment analysis to business goals more closely.
A big thanks to Paul, Harsh and Brian for answering my questions and giving us a preview on their thoughts. You can check out the event agenda and other details here. For Next Gen Market Research readers interested in attending the event feel free to use code ODIN for a $200 discount.
As always, I look forward to your thoughts, comments or questions.
[DISCLAIMER: Everything (participation, content, views, opinions) are personal and do not represent Walmart in any form or matter. Content based on general learning.]