Understanding the perceptions carried by your Customers and Prospects is critical in controlling customer churn. Our Perception Analysis Engine computes the perception from everyday conversations from the comments on your Facebook page, chat transcripts, customer support interactions and any other textual data that your organization may have.

These sentiments expressed by your Customers and Prospects carry a strong signal towards predicting the likelihood of Customer acquisition and retention.

There are four aspects of interactions that are relevant to determine the probability of a Prospect turning into a Customer or your existing Customers renewing.

  • Volume
  • Frequency
  • Sentiment
  • Recency

Volume and Frequency of interactions are captured using the data extraction processes of the Metadata driven ELT Engine. The Perception Engine identifies the sentiment of the interactions. All the sentiments are finally aggregated in the configured set of time buckets to capture the recency of interaction by applying our weight process.

How it Works?

Perception Analysis involves determining the evaluative nature of a piece of text. For example, a product review can express a positive, negative, or neutral sentiment (or polarity). Mining Opinions and Sentiments from Natural language is challenging because it requires a deep understanding of the explicit and implicit, regular and irregular, and syntactical and semantic language rules.

  • Our Perception analysis algorithms use a combination of English dictionary for single words, a repository for phrases, most commonly used idioms and phrases and NLTK’s Corpus.
  • We separate all the Stop words, special characters from a sentence. We then apply Lemmatization to derive a contextual meaning of a sentence/paragraph.
  • We analyze the lemma scores using at least three different varieties of algorithms and compare the scores and apply weight to each of those. We derive the sentiment precisely in 5 categories of emotions but group them into two final buckets of Positive or Negative binary signal that matters to the predictions.