> For the complete documentation index, see [llms.txt](https://docs.churned.io/churned-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.churned.io/churned-docs/models/sentiment-model.md).

# Sentiment model

NLP that scores the sentiment of contact moments and feeds it into health and impact

## What it predicts

The sentiment of a piece of customer text, classified into five classes: Very dissatisfied, Dissatisfied, Neutral, Satisfied, Very satisfied. Per customer the recent sentiments are aggregated into a sentiment value and, where surveys exist, into NPS detractor, passive and promoter counts.

## Data it needs

Text from contact moments: support tickets and their messages, emails and chat conversations, call notes, survey comments. Delivered through the support and CRM integrations. The model runs per language, so mixed-language bases are supported.

## How it works

A natural-language-processing model reads each contact moment and assigns a sentiment class with a confidence. Aggregation weights recent contacts more heavily. Sentiment is an input feature of the [health scoring model](/churned-docs/models/health-scoring-model.md) (typically in the Support or CSM engagement breakdown group) and a metric in its own right.

## How often it runs

New contact moments are scored in the daily run.

## How to read the output

On the customer page, sentiment explains a health drop that payments and usage do not. On the [Impact Dashboard](/churned-docs/measure-impact/impact-dashboard.md) the *Sentiments given* chart shows the distribution over the selected period, which is the fastest way to see whether a service change landed well.

## Where it surfaces

Customer page, Impact Dashboard, health breakdown, audience conditions. The sentiment model is part of the NLP on text [add-on](/churned-docs/getting-started-with-churned/add-ons.md).