Broadly Applied Sentiment Analysis May Lead to Unrequited Love
Enterprise text is similar to social media data in that in most cases it represents a conversation between your organization and a customer. Customer service transcripts, email threads, video transcripts, and surveys all represent the captured opinions, preferences and intentions of your consumer.
Embedded in this data are also consumer biases and attitudes towards your brand, their intentions to purchase from you, their dislike of a price increase or disappointment in the quality of a product or customer service. But within each of these examples, is context – an established qualifier to put some boundaries around a consumer’s expression. Only after applying some context does it make sense to apply an extra level of analysis in the form of sentiment to divine if the consumer expressions around quality are positive or negative. Without a backdrop sentiment does little to extend an organization’s understanding of consumer motivation.
Relying on Sentiment Alone Can Be Misleading
Everyone likes to be liked but if you are wanting actionable insights so that you can engage with your consumer with the right message then you need to reset your analytics approach. You need to implement an approach that helps you get to “The Why” behind your consumer’s positive or negative reaction. Use your organization’s key performance indicators to identify the conversation drivers important to track (i.e. consumer expressions around purchasing language, satisfaction, or pricing). This initial analysis isolates those consumers, who are communicating a certain type of engagement with your brand.
Changing How You Like Your Business
It’s difficult. Tracking sentiment provides such an immediate rush. If it’s good, there are high-fives all ’round. If it’s negative, there’s little to get excited about. But if you don’t know what’s driving the sentiment, there’s little detail for your organization to use, to improve or amplify. The high-five’s and hang-dog reactions to sentiment without context are both a little misguided. You need to get to “the why” behind the conversation, even when you don’t exactly know what might be driving the feedback. Enterprise text is almost a bit easier to analyze because the content is generally associated with a particular effort or topic.
For example, we working with a hospitality company, who knew they were having a problem with guest satisfaction based on completed survey questionnaires. But they didn’t know “the why” and didn’t know how to go about looking for it. If they had stopped at simply recording the number of negative expressions, then they wouldn’t have known that the reason guests were dissatisfied with their stay. The number of negative expressions would have been a value of little merit in a dashboard. Instead we filtered the conversation to surface “the why” behind guest’s sentiment. It turned out to be the pool towels. Equipped with this additional level of detail, the company was able to correct the issue. High-fives all ’round!
