Going out with a clipboard and asking people how they feel about President Trump before deciding whether to launch quarter-pounder Trump burgers with stars and stripes on the packaging is a primitive pre-digital-age form of sentiment analysis that will probably be inaccurate.
Technological developments are increasingly allowing product developers to unlock vast amounts of data and trace sentiment trends through a process of computationally identifying and categorising opinions expressed in a piece of text. The widespread – and frequently unfiltered – use of social media enables sentiment analysts to identify how groups of consumers think and feel in relation to brands, topics and products.
Tech increasing accuracy
Gary Barter, founder and chief executive of Hertzian, which uses machine learning and AI for sentiment analysis for its clients, believes sentiment analysis, put simply, is “the task of identifying parts of text that express positive and negative feelings towards something”.
Sentiment systems can now not only identify the context of specific statements in milliseconds, but they can also understand sophisticated or unexpected sentence structure, such as sarcasm or double negatives – something previous methods of sentiment analysis lacked, he explained.
“The increased accuracy from these new systems, coupled with the high usage of social media, have resulted in many industries using sentiment analysis to help provide a new way of understanding customer demands and reaction.”
‘Understanding consumer insights is key’
Frederico Pascual, chief operating office and co-founder of machine learning platform MonkeyLearn, believes sentiment analysis’ adoption among companies has exploded in recent years because of the advance of AI and machine learning.
“Understanding customer insights is key for introducing a new product in the food industry. At the end of the day, the customers will ultimately decide whether a new product is a hit or miss.”
Sentiment analysis allows consumers to be factored into an earlier point of the development stage, he continued “If you only listen to them after they try them on a new meal or recipe, it might be too late….by incorporating sentiment analysis, food companies are able to work faster and smarter towards more useful ends in product development and other key areas.”
AI and ‘human creativity’ provide depth
Netherlands based MetrixLab, which has worked with Nestlé, General Mills and Driscoll’s, revealed gauging sentiment on its own is insufficient. “The only thing that adds value is the reason behind being positive or negative or neutral,” said Jolique Weelink, global brand engagement strategist who leads the social insights team at MetrixLab.
Many use “scraping” tools, she says, to gauge sentiment online for PR, sales monitoring and to optimise sales campaigns.
“It’s tricky based on how the scraping tools are working to convert that into real marketing insight and to fuel product development, marketing strategies or activation strategies.”
It is necessary for the digital tools used for sentiment analysis not just to scrape the likes of Facebook, Twitter and YouTube but to be far more comprehensive, she noted.
“If you are a company working in berries and you are scraping for ‘blackberry’, you need to find a way to delete the mobile Blackberry…you need to be able to enrich the data, contextualise it…to localise it”, which is something an AI tool cannot do in itself.
“For now that will be done by a hybrid combination of the AI tools available and human creativity,” Weelink explained.
Weelink does not believe there will be a single tool long term that will provide all the sentiment analysis across all the platforms that companies need.
Currently sentiment analysis focuses on text, but the next will be visual, then voice, she predicted.
“After that, we do not know yet…Understanding sentiment is the most crucial fuel for product development because…it is written or posted by people in the moment. It is what people are doing, feeling, experiencing and dreaming of. It’s more truthful.
“You need the tool, you need the analysis and you need the brains to turn it into insight.”
Matching the ‘sentiment of our time’: Face-to-face has a place
Leatherhead Food Research head of insight Emma Gubisch, points out the food and drink industry is already using some of the latest market and consumer research tools.
“When thinking about new product development in food and beverage, it is about understanding where a company has permission to go from a consumer perspective – which products fit with the company brand and persona and what feels to consumers as though it jars or doesn’t work.
“Then it’s about the footprint of what that product should look like from a sensory, ingredient, pack and marketing point of view and how this fits with the prevailing consumer mood or zeitgeist and what consumers are looking for, [such as] healthy, nutritious, with few ingredients but still tasting great. It’s about creating the product to match the consumer sentiment of our time,” Gubish explained.
She added, however, that it is face-to-face discussions or qualitative research with consumers that are vital to really discuss a topic in-depth or for consumers to get their hands on an early-stage prototype.
“It is in this space that you can pick up on the deeper truths or insight of what consumers are saying and thinking.”