It’s one thing for a brand to get customers talking with word of mouth marketing. It’s another thing for a brand to understand the tone and attitudinal undertones of the conversations people are having about the brand.
As VP of Product at Scout Labs, Margaret Francis is responsible for designing online listening tools to help brands better understand consumer sentiment in online word of mouth conversations.
For the WOMMA Measurement & Metrics Guidebook, Margaret contributed an article about “mining social media for consumer opinion” to develop “business intelligence” and “measure marketing impact.” This article is a deep dive into the budding field of Sentiment Analysis.
I asked Margaret a few questions to help us better understand her perspective on harvesting, analyzing, and measuring consumer sentiment about brands.
Explain the importance for why a brand should do Sentiment Analysis.
Margaret Francis (MF): “A brand like Nike needs to know if people like them or not, but on the whole they probably know that many people associate them with favorable qualities that support their brand promise. I’m sure they have a brand tracker that measures overall Nike brand perception in much the same way as geologists study glaciers, looking for long term evidence of movement. Traditional sentiment analysis falls into this genre of study. It’s interesting data, but not for every day, and mainly intended for the very few people at Nike that have custody of ‘the brand.’
What Nike doesn’t have nearly as good sentiment analysis data about is consumer opinion about the specific products, sub brands, campaigns, and verticals that the marketing organization works on day-to-day. For instance, the people in Basketball might have makes, models, athletes, and campaigns whose social media presence they need to be tracking. This is the information that marketers need to make practical decisions- what video to feature on the splash page, which campaign to spend more money on, via what channel, which shoe to ask an athlete to wear at an appearance. Quickly and affordably getting below the mega-brand level- from ‘Nike’ to ‘LeBron IV’- is a big leap forward.
Also, it isn’t just brands that need sentiment analysis. How do consumers feel about use of sustainable materials, Chinese knock-offs, personalization, marathons, Spandex, the color pink, or a sponsored athlete’s extramarital affairs? These are not strictly “brand” inquiries but they are very important questions for marketing and other functional areas that deal with consumer preferences- areas like product development and customer service.”
What are some absolute “No No’s” businesses must consider when creating and using a Sentiment Analysis measurement process?
(MF): “There are known problems with having humans decide what is and is not positive/ negative sentiment:
Consistency: Humans don’t agree with each other more than about 85% of the time. The same person might, in different sessions, rate the same document a different way depending on interpretation. Is ‘Merrill Lynch rated Google a hold’ good, because Google used to be a ’sell,’ or bad, because it used to be a ‘buy’? Is ‘hold’ always bad? Or is this even an opinion? Many would call it neutral because the author is not expressing an opinion. Humans try to overcome this with custom guidelines.
Interpretation: The humans involved don’t always know what they need to know to make a good call on sentiment. Is ‘I love my iPhone, I unlocked it this weekend’ good for Apple? Or bad, because people ought not to unlock their iPhones? Humans try to overcome this with educated resources familiar with brand strategy.
Cost: It is very expensive to have a person, or better yet 3-10 people, read the same document/ watch the same video/ track the same tweet, and apply a rating to it. Humans try to overcome this by finding cheaper humans.
Problem is, humans don’t scale in speed or expense. We need machine analysis of sentiment not because it is better than humans but because with the advent of social media, the pace and volume of the data far outstrips the human resources available to analyze it.
So the absolute no-no in creating a sentiment analysis program is thinking that a machine is going to be better than people. Faster and cheaper, sure. Better, probably not. Sentiment data that is produced using machine analysis should be used for all the same reasons, and with all the same caveats, that human scored sentiment data is.
That said, there’s a reason traditional PR firms have stuck to the same methodology for a long time: They do sentiment analysis only on blue chip media pieces, using only in-house resources with strict guidelines, so analysis is limited in scope and expense and absolutely defensible. This is understandable but hardly forward looking. The rich nuggets of market intelligence derived from social media are a source of competitive advantage that no one can afford to ignore.”
Talk about a specific brand that has benefitted from doing Sentiment Analysis.
(MF): “It’s hard for me to give you some of the more compelling examples I know of, because the brands concerned view those examples as confidential business intelligence. Let me give you a hypothetical example. It isn’t a mystery that there is a furor right now over the Toyota brand due to the ongoing recalls on multiple makes and models. Any user of Scout Labs could plug in a search for the brand “Toyota” and see that there’s been a shift from positive to negative opinion over the past 6 months, and read consumer comments about the reasons why. Notice the shift from overall positive to overall negative sentiment for Toyota:
But only Toyota knows what their strategic response has been. Perhaps they pulled all marketing spend on affected models and redirected it to products that have been in market for ages and have a great safety record. Has anyone else been seeing a lot of Toyota Sienna commercials lately? Perhaps they shifted attention to non-Toyota branded products, like Lexus. More importantly, what should Toyota do in the longer term to rebuild consumer confidence? What do consumers think? Some recent online sentiment include:
“…The crashworthiness of Toyota vehicles is still superior to many other manufacturers.”
“And as soon as the clouds pass, Toyota should talk about the amazing deals that they have.”
“Toyota should however complement the effort with a theme campaign that simply states that Toyota loves you.”
Platforms like Scout Labs exist to inform brand strategists, not replace them. In the hands of the right decision makers, the kind of real time sentiment analysis Scout Labs can provide on any word, brand, product or phenomenon is very powerful marketing intelligence.”
*** Learn more about WOMMA’s MEASUREMENT & METRICS GUIDEBOOK ***
