The Experience Score is a 0–10 metric that estimates customer satisfaction for each interaction.
Unlike survey-based NPS, it’s available for every conversation ingested into EdgeTier, because it’s predicted from signals inside the interaction itself. It combines multiple indicators of experience (for example: resolution, frustration/effort, how the conversation ends, when gratitude appears, and channel context) using a machine-learning model trained on labelled customer outcomes from past interactions.
The EdgeTier experience score model produces a single score per interaction that you can track over time and use to spot changes in customer experience.
In plain terms: it’s an “estimated satisfaction score” for every interaction, even when a customer doesn’t fill out a survey.
Experience Score is based on qualitative and quantitative signals, so it isn’t calculated using a simple weighted equation (e.g., “30% sentiment + 20% resolution”). Instead, the final score comes from a machine-learning model. Unlike a formula with fixed weights, the model automatically learns the relative importance of different signals within interactions. For example, frustration often has a higher influence on the score than other factors.
No personal data or text from transcripts will ever be passed to the Experience Score model. The model is not a generative AI model, it's a traditional machine learning model that our Data and AI team train from scratch. It sits within our AWS infrastructure.
The model is trained on past interactions where a “true” outcome is known, typically NPS, or a labelled tag such as good / neutral / bad. From these examples, the model learns which interaction signals are most predictive of satisfaction for your customers.
The model uses all signals together to calculate the final experience score on a scale of 1-10.
For example, these factors typically increase the score:
While these factors typically decrease the score: