Sentiment

What it is for

Sentiment automatically classifies every inbound message (WhatsApp, email) as positive, neutral or negative, with an intensity signal. The results feed the health index (the "relationship" dimension) and raise alerts when a customer shows sustained deterioration.

It is not a detector of momentary anger. The useful signal is the trend over time.

When to use it

• Background monitoring: you do not have to do anything, sentiment runs on its own.
• Reviewing alerts: when LealUp flags a "sentiment drop" on a customer.
• Churn post-mortem: checking whether there were early signals that got missed.

How to reach it

• On the customer record: Health tab → Relationship dimension → weekly sentiment chart.
• In the alert list: Command Center → Alerts block → "sentiment drop" filter.
• Per contact: contact record → Recent messages, each message shows its sentiment icon.

Example

Valentina (director) opens the Command Center on Monday at 8:30. There is a new alert: "Industrias Bravo (champion Luis Pérez): sentiment down 55% over the last 2 weeks". She opens the customer and sees Luis replied to 3 emails in a frustrated tone about an export bug. She assigns a task to Joaquín, the CSM, to call today and offer a fix.

What it does not do

• It does not replace reading the messages. The classification is an aggregate signal, not a verdict on each sentence.
• It does not infer churn intent. The health model and retention playbooks are for that.
• It does not classify outbound messages (yours to customers), only inbound ones.

Known limitations

• Supported languages: Spanish (LATAM and Europe), English, Portuguese. Other languages default to neutral.
• Very short messages ("ok") may come out neutral even when the context implies otherwise.
• Classification runs in batches of roughly 5 minutes. Very recent messages may not have a sentiment yet.

---

Last updated: 2026-04-21.