Today, there is so much information available and in business, around 80% of information comes in the form of unstructured data. Unstructured data is qualitative, textual content and can be hard to make sense of. The content can often come through social media, email, images, surveys, task management, chat, etc. With the average tech stack at around 20 tools per user, it’s not hard to imagine why the 80% is expected to grow to 93% by 2020. Customer feedback, for example, is very important but hard to keep track of. Analyzing customer reviews is difficult because the feedback keeps growing and there is usually too much information to process manually. Problems can also arise with many reviews, as they can lack consistency and depth.
In a recent joint webinar, Diego Ventura from MonkeyLearn demonstrates how you can structure and make use of this type of data through machine learning and natural language processing.
Using MonkeyLearn to Analyze Feedback at Scale
With MonkeyLearn, you can sort through all of the reviews that come in and can inform your product decisions by analyzing the aspects, sentiment, emotion, opinion units, etc.
MonkeyLearn also allows you to centralize criteria about how to tag and structure certain data and feedback. You can apply the same lens and criteria to all of your feedback, reduce errors in classification and processing with centralized text analysis modules, and allow your team to control further training to help improve model accuracy.
In this demonstration, Diego leverages Slack feedback left in Capterra. Some reviews include:
- “It’s really easy to integrate with”
- “Fantastic collaboration tool”
- “Terrible notification sounds, you will hear it even in a noisy room”
The idea is that you can grab these reviews and analyze them at scale, which can help inform product decisions. For example, if these three reviews are indicative of a larger trend, and you are a competitor of Slack, you might want to take a look at your integration capabilities. Or if you are Slack, you might want to address notification sounds.
Summary of the Process of Analyzing Customer Reviews
The process involves classifying reviews using models that have been trained with the Slack data taken from Capterra. For example, a review saying “Slack helps improve coordination among team members” could be classified, using sentiment analysis, as either negative, neutral, or positive. In this case, it is positive.
You can also use the models to classify the reviews into topics such as ease of use, integrations, performance quality, and UI-UX. For example, a customer saying, “I work across several workspaces and switching is super easy,” could then be classified as ease of use. Using RapidMiner, you can then connect all these individual results. Check out the webinar to learn how to use pre-trained models in MonkeyLearn, how to train custom models with your own data, and how to set up this process in RapidMiner.