No quirky title from me today. Rather, I'm going to be straight with you: text analytics can be hard. There are too many different ways to handle unstructured/textual data. People don't use language in clear, concise, consistent ways. Sentiment analysis is often inaccurate. Even with automated tools, human intervention and guidance are still required.
Really, this list could go on and on.
Creating a robust, sustainable, and accurate text analytics program is a challenge – but do we always have to look for that all-encompassing, diamond-in-the-rough-finding program? Instead, could we maybe find a quick, simple, and straightforward method to use our unstructured data? This idea has been floating around at Walker recently, as I'm sure it's been debated in many other companies interested in what their customers have to say. What can we do to simplify our analysis, while still providing valuable customer experience insights?
One possibility is to look at the most common nouns that are appearing in your data set, and then mapping those nouns to a quantitative metric or rating score available. Let's say you measure NPS, for example. You ask customers how likely they are to recommend you, and then you have an open-ended question that asks why the customer provided that rating. Using text analytics software, we can identify what the common nouns and noun phrases are in the data, essentially identifying the main themes being discussed. We can then filter those nouns by the ratings that were provided; for example to look at what Detractors versus Promoters are saying, and if the themes vary.
No manual coding. No taxonomy or rule-set creation. No extensive processing times. Minimal confusion.
Simple? Indeed. Life-changing? Probably not. However, not every single text analytics venture has to change the direction of your company – a pitfall I often see customer experience professionals encountering. Use some of these small insights to inspire an overall understanding of where your customers are coming from. Then, maybe follow up with certain groups individually to get more "layers" to the story. Start small and build from there.