Analyzing forum content: a new customer group and product development
Product development and customer acquisition
Industry and type of organization
A company developing its own product and looking for new audiences for it. The knowledge of what users really want was scattered across forums and topic groups, where people describe their problems in their own words.
Business problem
Decisions about what to add to the product and whom to take it to were made on the basis of single signals: a louder customer, an internal hunch, whatever the competition happened to be doing. Meanwhile users on forums describe directly where it pinches and what they would be willing to pay for.
This material, however, is unstructured. It is free text, in varying styles, with no categories at all. Reading and sorting such a mass of posts by hand is unrealistic, so the knowledge lay untouched.
Approach and scope of work
We gathered the content users write themselves on topic forums and processed it with natural language processing (NLP) methods. Instead of reading post by post, we first turned the statements into vector representations (embeddings) that capture their meaning, and then grouped them by clustering: the algorithm arranges similar statements into clusters on its own, so recurring themes and needs emerge from the raw mass of posts.
Each cluster is one group of expectations: a concrete problem, the way people talk about it, and the language they use. On this map you see two things at once: audiences the product had not yet spoken to, and features missing from the current product.
We processed the text data in a secured environment.
Business effect
tens of thousands
statements analyzed
content from forums and topic groups
two directions
a new customer group and new features
entering a new segment and extending the current product
The company received user expectations described in their own words and grouped into clear themes. Some clusters pointed to a segment the product had not previously reached, which opens up a new customer group. Some showed concrete gaps in the current product, that is a ready list of features to add.
Extending the product with features for which there is real demand raises its value for the customer and has a positive effect on margin, because the company builds what people are actually looking for.
Scale of the implementation
The analysis covered tens of thousands of statements from forums and topic groups related to the product, so the scale was large enough for the patterns to be reliable. The more statements in the source, the more reliable the clusters, because a single voice is an anecdote, while a recurring pattern is a signal.
If your users describe their needs online and you make product decisions by gut feel, describe your case to us. We will check what can be read out of what people already write about you.
Related articles in the cm-opti knowledge base
Service area: Data analysis and NLP in product development.