How natural language processing can help marketers hear the voice of the customer

NLP voice of the customer
0 Shares 0 Flares ×

The Six Sigma methodology of improving processes includes, as a first step, capturing the voice of the customer (VOC) and using it as a guideline to provide the best experience and quality. The VOC is essential since it includes both stated and implied requirements. It varies continuously and requires frequent updating. With the rise of the Internet, social media and online discussion forums, companies found that the clients were stating their opinions, reviews, and claims openly, online. Capturing the VOC was no longer a problem of convincing people to talk about the product. It became a matter of collating all the opinions from different channels and analyzing them in a unitary setting.

Customer Feedback: Importance and Methodology

Having quality customer feedback is similar to reading the minds of clients and foreseeing the future. Companies can create better strategies, manage stocks, control costs and hedge risks. This, of course, translates into increased revenue and better management of the cash flow.

However, the real problem is what method is best to collect data from customers. These use different channels, ranging from reviews on online shops to social media or even discussion forums. Some are easier to retrieve, due to a semi-structured organization, using hashtags, like Instagram or Twitter. All the others, which are unstructured need to be further analyzed just to get the underlying message.

Challenges of Analyzing Unstructured Data

While humans have no problem understanding communication in various forms, ranging from slang to academic, machines have a much harder time making sense of sentences which are not in a standard format. It requires formatting, parsing, labeling words identifying negative forms and much more.

Until recently, this analysis was performed by marketing assistants, via interviews, focus groups or looking through some reviews. They used to hand-code responses according to the degree of satisfaction and sometimes even used numbers for some statistical computations.

Right now, considering that Yelp users alone add about 30 million reviews per year, it would be impossible to continue performing analysis by hand. It would take years to go through these, and the salary costs for the operating personnel would most likely be larger than the benefits.

Not only it would cost a fortune to use people to gather and analyze customer feedback, but, apart from the inherent mistakes, by the time the results were ready, they would already be outdated.

Also, this approach would require defining some fixed categories which do not necessarily reflect reality or remain constant.

So how can a company address the volume, velocity, and variety of information included in customer feedback? These are precisely the characteristics of Big Data, and the solution comes from a tool designed specifically for this purpose: Natural Language Processing (NLP).

How Can NLP Make a Difference?

NLP can help companies make sense of the thousands of feedback they receive directly or indirectly from customers.

It does this first by filtering information, removing duplicates and cleaning unnecessary punctuation. The next step is splitting the message into individual words. Not all words are essential for communication, while others are extremely relevant, especially those containing names, actions, describing sentiments or acting as negations. Using an embedded dictionary, the algorithm classifies each word as a part of speech. Next, it looks at verbal tenses, meanings and tries to include the relationship between words in this step.

Once such data is in place, it goes deeper into analyzing the underlying sentiment by scoring each emotion. Emotions are grouped into clusters, and the same word can change the scores in multiple clusters.

Advantages

Relying on an automated tool means that the company is no longer tied to its available workforce. The sky becomes the limit for the number of channels or comments which can be analyzed. This opens up the possibility of sourcing feedback from numerous channels at the same time and putting everything into the same database for a global understanding of customer perception.

Having such a vast assortment of customer opinions drastically reduces the potential bias which is typical for focus groups and interviews. Also, if specific problems emerge, it means those are real issues and should be addressed as soon as possible to completely satisfy the customer.

Last but not least, this method can be used to get almost instantaneous feedback on things such as the impact of a commercial or a particular campaign. It can be used to test marketing tools on a smaller number of target clients before going full-scale. This saves time, money and sometimes even the reputation of a company.

NLP In Action

A good illustration of the utility of the NLP method is a case described by InData Labs for a gaming company. The feedback was collected from YouTube reviews and forums regarding the release of a new game. This helped analyze the sentiment of the players, identify bugs and flows which were not working smoothly and launch the updated product in time, while contributing to significant cost and time reductions compared to traditional customer feedback collection methods.

Conclusion

To remain competitive in analyzing customer feedback created at a rate of thousands per second automation is no longer an option, but a necessity. Capturing the VOC is no longer about interviews and surveys, but about web scraping tools to gather opinions from forums and social media combined with NLP to analyze them. Sentiment analysis brings a new dimension to understanding what clients want and need from companies.

The savings regarding time and costs are so significant that the initial deployment expenses are covered after just a few weeks. The only downside is that currently there are no user-friendly tools to give companies the possibility of performing this analysis independently.

0 Shares Twitter 0 Facebook 0 Google+ 0 LinkedIn 0 Email -- 0 Flares ×
The following two tabs change content below.
Sophia Brooke

Sophia Brooke

Sophia Brooke has over 3 years experience as a technology journalist; writing on subjects such as science & innovation, big data, the Internet of Things (IoT), robotics & AI for a range of publications.