Making customer feedback actionable - how can AI help?

In many companies, customer experience is measured, but the results are not actionable. The most widely used customer experience metric NPS (Net Promoter Score) actually gives all the necessary ingredients for the actionability – when customers are asked not only the score but also why they gave that score, they provide very useful information about the drivers of satisfaction or dissatisfaction. The problem, therefore, is not the lack of data, but the fact that the data is not analyzed and utilized as it should. 

It is very typical that a customer experience metric is followed on leadership team level only as a number. As long as the metric goes up, everybody is happy. But as soon as it starts to decline, there is panic in the air. No-one seems to know what should really be done to improve the metric.

There are a lot of anecdotes floating around, strong opinions and some blaming as well: marketing is confident that the product is to blame while the product team sees marketing as the culprit. Task forces are being initiated to investigate the reasons behind negative development. Market research is being commissioned and market research agencies start doing both quantitative and qualitative studies to get to the bottom of the issue.  

All of this takes both time and money. Luckily, with the help of modern technologies, that should no longer be needed. And artificial intelligence has a role to play in this. 

Relevant fields of AI.JPG

The most important AI technologies relevant for analyzing customer feedback fall in the area of natural language processing (NLP) and machine learning. Both groups of technologies can be utilized to make analytics more actionable.

When it comes to text feedback, they are beneficial even on the basic level of descriptive and diagnostic analytics, but the benefit becomes more obvious when we aim at predictive analytics. Machine learning is can be utilized to make predictions based on the historical data as long as there is a large amount of it. It can help us to understand how likely is a person with certain characteristic to churn, what could they buy next or what would be the ways to improve their loyalty. 

Some routine decisions can be automated as well: the machine can respond to the most typical complaints and suggest solutions to problems that many people have. However, human judgment is and will be needed in strategic decisions and in seeing the results of analytics in the full context. 

Analytics steps.JPG

Today, the most optimal analytics solution is typically a combination of machine and human intelligence. While the machine learning brings in speed, cost-effectiveness and ability to process massive text volumes in a split second, the humans are needed to interpret and make judgments. 

Heart.JPG

Lumoa’s analytics solution is built on top of this philosophy. We combine machine learning and other AI technologies with human-built rules that are specified for each industry and sometimes also tweaked to match the business of our biggest customers. This way the results are always relevant for the businesses we work with, but the modern technologies help us in increasing the efficiency and reducing cost level. 

We recently talked about this topic in a breakfast seminar hosted by Sininen Meteoriitti. If you are interested in reading more about the topic, you can download our slides here: