Reviews help organizations improve their products and services, nothing new. But what are the exact features your customers want? An essential part of the strategy of this client was to listen to their customers. They asked us how to tackle the classification of their customer reviews.
One of our clients faced a problem where they found it difficult to classify open customer reviews about their brand. We helped them by building a text classifier that could quickly classify these reviews. With these insights, our client was now able to better steer the organization and prioritize business process improvements.
Our open text unsupervised clustering model that can categorize a wide range of customer reviews. Without human labelling efforts (pre-annotation) our model was able to predict the correct journey and topic in 80% of the cases. We developed a reviews model that was able to give a detailed overview on what topics your customers are complaining about. So, you can address these issues on a focussed way.
- When complaints are resolved properly in 60-70% of the cases detractors became passive and passive became promoters
- Detractors that become passive have 30% less change to churn
- Passive that become promoters have 40% less change to churn
Better insights: Categorise customer reviews and actively steer the organisation based on these subjects
Better decisions: Organisational plan on which improvements tackle first.
100.000 customer reviews
Unsupervised NLP model (Autoencoder & Transformers)
Modelling, Testing and Integrating
Realized in 24 weeks