What is sentiment analysis and how can it help your business?
What is Sentiment analysis and how can it help your business?
What is the first thing you think of when you hear the word “data”? Numbers. While we love numbers at Museo Analytics, they aren’t the only basis for data-driven analysis. There is a vast ocean of another type of data that many businesses leave unexplored: textual data. This can include every piece of text that you might find online, be it a tweet, blog, news article, or a customer review. In the digital age, analysing data for your business is more than just figuring out smart ways to read the numbers; we need smart ways to read what people are saying.
That is exactly where sentiment analysis comes in. Sentiment analysis is generally the classification and interpretation of emotion or intention from text. From a marketing perspective, sentiment analysis usually involves automated tools that can file through huge sets of textual data and derive insights as to the positivity, negativity, or neutrality of what is being said. Whether you’re a burger joint that wants to know how many of your customers love the new recipe, or a large-scale commercial operation that wants to know how news media portrays your brand, sentiment analysis is a crucial tool for finding out how you stand in the eyes of the market.
How it works
Sentiment analysis at its most simple relies on “rules” that you might give a system to categorise text. For example, when you plug the phrase “This product is terrible” into a basic system, the system will need to be told to assign the word “terrible” into the negative category. You could do that with as many words and rules that you are willing to type up. The more rules you provide a system, the more accurately and widely it will be able to process and analyse.
As you could imagine, however, providing these rules manually will only get more time-consuming and fidgety as the amount of data you need analysed grows. Therefore, machine learning processes have been designed around the world to work through text automatically. Data scientists can “train” the machine learning model to identify and categorise words by feeding it examples with associated tags or categories. After the model has worked through many examples, it will improve and optimise itself relative to the rules or criteria that it has been given. Once it is fully trained, you can automatically feed text from social media, blogs and news articles into the model with various publicly available programs. So while basic rules-based analysis can provide the initial examples for the model to learn from, eventually a model can take over and process millions of words at the drop of a hat, giving you clean, insightful data about what people are saying online.
How it can be used
One of the most common applications of sentiment analysis is for businesses to analyse and quantify the opinions of their customers. Bain & Co. research found that generally positive customer experiences can grow 4-8% revenue over competition, through higher retention of customers for a longer time.
Sentiment analysis can be used to monitor the success of new product designs, marketing moves, or even political campaigns. It can also help a business to monitor the general perception of their brand over time, letting them figure out when brand sentiment increases or decreases in positivity, and what they can do to control those spikes and movements. The ability to measure that sort of effect on your brand has plenty of commercial repercussions - providing the opportunity to forecast, manage, and get ahead of the curve.
One prominent example that is indicative of the power of sentiment analysis is from Expedia Canada in 2014. After rolling out an ad involving a screeching violin, the company received some real-time social media sentiment data that was markedly negative and pointed towards the violin as the culprit. Expedia used the instant data at their disposal to pivot extremely quickly and roll out new ads satirising the violin from their first ad. The new ads were very well received and sentiment turned positive in a relatively short space of time.
A growing industry
As AI technology develops, sentiment analysis will only get stronger and more accurate. Many large and small businesses already couple it with other forms of analysis, and the best day to consider it for your business is today. There are many third-party applications already offering information and programs around sentiment analysis, such as MonkeyLearn, Lexalytics, and even Stanford’s open-source NLP software. IBM estimates that 80% of the world’s data is unstructured. Sentiment analysis gives businesses a chance to structure that wealth of information - to organise it, label it, and learn from it. The industry around sentiment analysis is only growing, and without implementing this sort of insight, businesses only risk waking up at the start line while everyone else is in the race.
Published by Will Eddowes