How we used Text Analytics to prioritise Decision Making

Text Analytics, in the context of Market Research means analysing continuous stream of free, ‘unstructured’ text (feedback and comments) written by the customers.

Asking the customers for direct feedback is one of the best ways for companies to get actionable information. This data can be leveraged easily using Machine Learning and Text Analytics. Therefore, companies must focus on open-ended comments rather than just closed rating questions, to better understand customer issues and challenges.

WHY CLOSED SURVEYS/ RATINGS ARE INSUFFICIENT

1. LIMITED PARAMETERS– They limit the parameters (such as politeness, informative) on which the customers can assess a product or service. These surveys pigeonhole the customers by providing them with limited options. Furthermore, these options/parameters are based solely on the company’s understanding of possible issues/ problems that could arise.
   
   
2. CAN NOT CATCH EMERGING ISSUES– These surveys cannot catch new, emerging issues.
   
   
3. NO DIFFERENTIATION– Customer ratings on a 5-point scale do not offer any differentiation.
   
   
4. TOO LONG- Because, companies try to be thorough and cover all possible parameters, these surveys often end up being lengthy and tiring. Consequently, they lead to survey fatigue and prevent companies from deriving actionable data.
   
   

Numr recommends that you use a 2-Question survey for your NPS® program. Therefore, the first page of any survey should have a rating (NPS®) question as well as a mandatory open-ended text box for comments and suggestions.

HOW TO DO TEXT ANALYTICS

Once, a continuous flow of comments has been established, the process is started.

Primarily, it comprises of two steps-

1. SENTIMENT ANALYSIS- The first step is to categorise comments as positive or negative using NLP. During this process each comment is assigned a score between 0 and 100, 0 being extremely negative and 100 being exceedingly positive.

Therefore, we can raise trigger/alarms based merely on the tonality of the comments.

2. CATERGORISATION- After the sentiment has been established, the machine learning system can begin categorising the comments.

GIVEN BELOW IS A CHART OF TEXT ANALYTICS DONE FOR A TAXI AGGREGATOR.

‍

By leveraging Text Analytics, companies can analyse comments and as a first step pick out recurring topics (issues) that most customers are talking about (This is represented by the size of the bubble). Now, what Numr Research next is we utilise Regression Analysis to help companies prioritize the issues that have the biggest impact.

Notably, this Regression Analysis model isolates how much impact a topic will have on the overall Net Promoter Score®. For instance, it predicts whether a customer talking about trip cancellations is more likely to give a low NPS, as opposed to the one talking about fares and vice versa. As emphasised in the chart, the customer having issues with trip cancellations is a lot more likely to be negative in their outlook as compared to say car quality. And, this negativity is clearly reflected in the NPS.

Most importantly, this tells companies the topic that they should prioritise and handle first.

TEXT ANALYTICS OR NPS®?

In a word, both.

NPS is a clean, manageable metric that an entire organisation (from the mangers to the frontline) can utilise.

However, relying only on open-ended comments is extremely effective in understanding the issues that need fixing. In conclusion, Text Analytics enables companies to get a firmer handle on everything that is going on with the customers.

*NPS® is the registered trademark of Bain and Company, Satmetrics and Fred Reichheld