KNIME logo
Contact usDownload
Read time: 6 min

What Marketers Need to Analyze Customer Sentiment

May 12, 2022
ML 201 & AI
sentiment analysis_0.jpg
Stacked TrianglesPanel BG

B2B buyer behavior is increasingly shifting to digital channels. To adapt and provide these customers with rich – more Amazon-like – buying experiences that customers take for granted, businesses need to find out what content they find valuable, as Gartner points out in their Future of Sales 2025 report.

But it isn’t easy. Stuck in spreadsheets, going through customer feedback line by line – marketers know how time-consuming it is.

To see how marketers can successfully overcome real-life challenges of customer data analysis, we look over the shoulders of two members of a fictional marketing team at a B2B company: Aline, a data scientist, and João, a marketing analyst.

Aline, just hired on the team as a Python programmer to improve customer feedback analysis with data mining techniques, digs into social media sites where the target audience hangs out in order to discover new information that the rest of the marketing team can turn into actionable insight. The challenge is that this team has no common background. So while Aline, a data scientist, spends a lot of time interpreting what marketers want, turning it into an analytic model, and translating that data into insight for the team, all this “talking” takes up a lot of time. Also, Aline’s work is written in code.

João, the marketing analyst, wants quick insights. And while he can understand the results of Aline’s model, if he wants to make any change or see a different view, it’s hard for him to tweak the code without her help. And when she’s out, there’s no one else who can jump in. It would be simpler and save time if the marketers could work together and even build applications themselves, injecting their valuable industry knowledge into the process. It’s clear that Aline and João’s team needs to change the way they’re collaborating.

Accessibility to advanced data science across the team

For João and Aline’s team to be more successful, they need a tool that everyone can use – one easy enough for marketers to start doing simpler analytics themselves, but advanced enough to continue using the AI/ML-based analysis Aline has introduced. It should be intuitive enough for João to extend his skills and learn to apply machine learning himself, and open enough for Aline to continue using Python if she wants to. Advancements in open source analytics tools led to sophisticated techniques like sentiment analysis becoming potentially available to anyone. But people mostly had to learn high-level programming skills to use them. Now no-code/low-code tools also support business users as well, giving them self-sufficiency to build advanced applications, even with AI/ML techniques, in a drag-and-drop manner. The team starts to use KNIME to gain access to advanced data science.

The marketing team learns to use sentiment analysis with KNIME

To provide the meaningful content people are looking for, the content writers on the team need to immerse themselves in the customers’ world, listening to and observing conversations around their product. That means not just analyzing internal customer data, but also going to where the customers hang out – Twitter, Instagram, LinkedIn, Reddit and Quora conversations – and analyzing the posts and dialog that take place there.

The team wants to start with sentiment analysis to understand the customers better, as it can be used in multiple use cases that all contribute to improving the customer experience. For example, they can gauge sentiment around a new product launch, evaluate brand reputation, compare themselves to competitors, provide proofs of concept, measure marketing/PR efforts, detect or predict potential PR crises, or identify influencers.

Content writers build automatic tweet sentiment predictor by drag-and-drop

The amount of data produced by tweets alone is huge. Combing through 14,000 tweets manually in spreadsheets is not effective. The team can save a lot of time with a solution that can process large amounts of data, then measure and predict sentiment.

Using a codeless application for simple sentiment analysis, the content writers can learn to run and adjust the analysis easily themselves, depending on requirements. They don’t need Aline or even João to tweak the application for them. Instead they are self-sufficient. They can develop a dashboard, for example, showing a word cloud of the predicted sentiment. Color coding enables them to easily see when complaints start to exceed positive sentiment. The content team can dive deeper into the visualized data in the dashboard to see exactly what customer opinion is. They use this insight to improve web content that addresses the customer feedback – for example, by improving FAQs or developing landing pages on the relevant topic.

Tip: Download a blueprint to build a sentiment predictor for tweets, described in this tutorial.

Marketing analysts extend skills to ML for fine-grain customer review analysis

The longer and more complex a text is, the greater the need for more advanced sentiment analysis. Before, the marketing team had to hand over tasks needing more refined techniques to Aline, which created a bottleneck. The low-code tool gives João access to these advanced techniques and the data analytics bottleneck is eliminated.

João learns to build a machine learning model that can analyze the sentiment of longer texts, like product reviews left online by customers. In the future, this application will even be able to analyze the sentiment of reviews on its own, with no input from João required. With the insight from the reviews, the content writers can craft more targeted customer web content. With more time on his hands, João can learn to apply even more advanced and accurate sentiment analysis techniques to get even more precise results for his team.

Tip: Read more about the machine learning approach to sentiment analysis in this tutorial.

Data scientist and marketing analyst connect to develop AI applications

Our marketing analyst, João was soon able to extend his analytics skills to learn how to apply advanced algorithms in the visual programming environment of KNIME, reducing the complexity of these techniques. He can now collaborate easily with Python programmer, Aline, sharing functionality within the same tool. KNIME’s Python Integration allows Aline to continue programming in Python if she wants to.

→ Advanced deep learning refines customer complaints analysis

In their first collaboration using KNIME, they apply deep learning to emails and detect data quality issues with spam. It was important to the team that they be able to analyze sensitive data like customer complaints as accurately as possible. They also need a solution that would ensure the integrity of the data and accurately detect data quality issues arising from spam. This type of resource has the potential to save marketing analysts a lot of time.

The team can now go on to use this deep learning technique for other types of text analysis, like in foreign language translation or speech recognition.

Tip: Explore a tutorial on how to set up a deep learning approach to sentiment analysis.

→ State-of-the-art customer analysis with BERT

In addition to finding out the sentiment of customer tweets and comments in reviews, the team can go up a level and use a technique called Bidirectional Encoder Representations from Transformers (BERT) to improve the accuracy of the tweet sentiment predictor. The BERT nodes in KNIME remove the technical complexity of implementing this advanced technique.

They decided to test the technique on the same customer tweets, and found that BERT was much better at accurately predicting the sentiment. Their success motivated them to consider how they could go on to use this technique for content moderation, answering questions with bots, or identifying key topics customers talk about.

Tip: Read more about how to implement BERT technique in KNIME in this tutorial.

Cross-skills collaboration for sophisticated sentiment analysis

The barrier to adopting sophisticated approaches like sentiment analysis is often a lack of knowledge about the opportunities such approaches offer, and a dearth of the skills needed to apply them. KNIME software allows teams to combine domain expertise with analytic skills without any worry. Team members with different analytic skill levels can collaborate to build workflows entirely using the drag-and-drop interface, and more technical users can even code in their favorite language with one of the many scripting integrations.