Text Analytics 101
The explosion of customer feedback from social media, text boxes in surveys and the data from call centre voice recognition software have driven innovation in the field of text analytics.
Text analytics is essentially a way to automate the process of reading the comments and coding them with the underlying message – both the subject and the sentiment. It uses language algorithms that are surprisingly accurate to understand what a customer is talking about and tag their comment with its meaning. This turns open text into categorical data that analysts need to perform their magic for example, it can be fed into data cubes, visualisation engines and predictive models.
Before text analytics was available we had to manually code all the responses and create a man-made database with the meaning and sentiment. This would often be limited to a sample of the data because the process is very slow. There is also the risk of human error as the task is delegated to a team of people who have different ideas about what the customer is saying. It could take us days to find out that, for example, 25% of Detractors complain that the checkout staff were rude. In situations where a customer is about to churn, this time delay can make all the difference. Customers expect the business to act on their feedback very quickly.
Now that accurate text analytics is a reality, we can very quickly code all the open text and trigger message to call centres for any customers that are about to churn. The analysts can spend their time running more sophisticated analysis. For example, we can slice and dice the data to see how this 25% varies by store/region/day of week/time of day/customer value segment/channel. We can also see how all the different customers complaints are correlated with each other and create a Decision Tree to show how the different combinations of complaints affect Net Promoter Scores. This multivariate analysis is very important for the identification of the critical factors (pain points) for the most valuable customers.
By holding the comments as categorical data, you can also monetise them within your data/CRM – see the white paper for more details.
Net Promoter, NPS, and the NPS-related emoticons are registered service marks, and Net Promoter Score and Net Promoter System are service marks, of Bain & Company, Inc., Satmetrix Systems, Inc. and Fred Reichheld