"Machine learning offers marketers the opportunity to move beyond sentiment and instead harness intent data. Such information is incredibly powerful and allows marketers to answer crucial questions."
What people share with their friends and family on social and other forums is the ultimate litmus test for your brand. When people are swapping tales with their friends about their latest purchases or interactions with companies, they don’t tend to hold back, and that’s why this human-generated information is so valuable.
While filtering such data for trends has become increasingly popular, the truth is that the sentiment distilled from these social media posts has limited value. How do you correlate positive or negative sentiments with an outcome? The simple answer is you can’t because you’re left to draw your own conclusions. Or put simply, you just have to assume that assumptions don’t scale up as they should.
In a multi-channel, multi-screen and multi-device world you cannot drive loyalty and deliver a seamless consumer experience or being able to effectively engage your target audience based on assumptions. What is needed is detail. Are people raving or complaining? Is it specific to a product, brand or industry? Knowing that a group of people are simply neutral towards the launch of your latest product would not be considered by most to be the path to enlightenment.
Machine learning offers marketers the opportunity to move beyond sentiment and instead harness intent data. Such information is incredibly powerful and allows marketers to answer crucial questions. Which products are people considering buying? How satisfied are my customers? Are they thinking of leaving me? Which products or services prompt the best reviews?
These are all great questions, but marketers have to bear in mind the nuances of what they’re looking for. On the other hand, machine learning means that platforms can become self-learning and begin to understand subtleties such as sarcasm. Its ultimate success rests in a strong understanding of the audience and their relationship with your key words, topics and brand.
In all likelihood, how people talk about your product or service on social platforms is unlikely to correlate perfectly with your corporate messaging document. After all, this is human-generated data, so spelling mistakes, interpretations, inconsistencies and slang are all common. What is needed is an understanding of how your key words and topics translate in the real world.
Much of machine learning happens on what we would call ‘sanitised data’ sources such as Wikipedia. Often the language used in these sources is very academic and not real world applications of words. There is a big difference and nuances need to be understood if marketers are to firstly recognise signals from their target audiences, understand these signals and use the data to inform business decisions. Furthermore, understanding key word relationships empowers machine learnings to find signals that could be ‘hidden’ behind neutral sentiment, but could in fact infer other actionable findings.
Sentiment, while well understood, lacks associated action. For example, I could be really positive about a new gadget I’ve brought, but if I am past the purchase or customer service stage, what value does that sentiment deliver to a brand? It’s positive, which is great, but at the same time it’s pretty one dimensional. It’s a bit like measuring social engagement by the number of followers you have, rather than quality of interactions.
Marketers have to understand the terms that are being used in association with their brand and machine learning offers an effective and scalable way of doing this. If marketers fail to gain a more in-depth understanding of the language used in relation to their brand, they are missing a big, very important, part of the story.
By Jason Rose, SVP Marketing at DataSift