How we surface customer complaints, wants and needs through text analysis?


The Problem

Customer feedback drips in from various sources including customer surveys, customer reviews, Twitter and customer support conversations. All this data put together is actually a gold mine for understanding what customers REALLY want. It can be a powerful tool for discovering new opportunities for innovation and can help pin point nagging problems in your products or services.

Let’s take a healthcare example. If a hundred patients at a hospital complained about parking being limited at the hospital facility, then a corrective action by the administration could be to expand the parking structure to address this pain point. Similarly, if several hundred patients complained about hospital staff being rude, then the hospital could decide to send their staff for training on how to better accommodate patients and their needs and dig further into the reasons for the rude behaviors. Are they being overworked? Or do they need to change their hiring approach?

Unfortunately, manually discovering such patterns, is not easy. You will have to read feedback from hundreds of users and this can become an error prone process and can quickly become exhausting. Luckily, with the help of Natural Language Processing, Text Mining and Machine Learning techniques, we can automatically analyze all this text data surface repeating patterns.


From Messy Data to Deep Insights

On an ongoing basis, we help organizations gather insights about customers from volumes of messy, unstructured feedback data consisting of user reviews, survey responses, support conversations and others. Our insights have typically included:

  • Customer Wish List – what are people looking for?
  • Sentiment Trends – customer perception over a period of time
  • Common Complaints – what are customers commonly complaining about?
  • Areas of Excellence – what is your brand or organization doing well in?
  • Common themes – what are the common discussion topics ?


Our Approach

Our approach to customer feedback insights is highly customized and thus accurate and actionable. We build new models for specialized insights and we re-train most of our models for each of our customers. A one-size-fits-all approach generally does not work as the data can greatly vary between industries. Here is a summary of our approach:

  1. We first merge, clean and standardize data from various feedback sources.
  2. We then re-train our in-house classifiers to detect wish lists, sentiment polarity, complaints, areas of excellence and others.
  3. We build additional models as needed
  4. We then compile a highly actionable summary so that our customers can immediately make data-driven decisions.
  5. We repeat this process every month, quarter or year.



From this ongoing project, our clients have been able to:

  • Detect organizational and product problems and come up with a plan of action before problems get bigger.
  • Find areas where they could introduce new features and services based on what customers were complaining about and wishing for.
  • Understand areas they have been excelling in and incorporate more of that in other parts of their business.



Here are some examples of insights from our work in a healthcare domain.


If you are interested in gaining meaningful insights from your unstructured feedback data, please get in touch with us for a free consultation.