---
schema_version: "1"
id: fa123e4a1b68507c454dd3d71a2057a241a6421287859b0b1571e7567dbb3686
title: Engagement Score Thresholds
collection: admin
version: current
locale: en-us
portal_url: https://docs.salesforce.com/admin/current/en-us/mktg/264-0-0/mc_anb_engagement_scores_thresholds.md
canonical_url: https://help.salesforce.com/s/articleView?id=mktg.mc_anb_engagement_scores_thresholds.htm&release=264.0.0&type=5
source_site: help.salesforce.com
content_type: documentation
content_hash: ebb29264c0e6756869b115254fe03fa45ee8f3ad21931d0d9189a747b819ce23
description: Each key performance indicator (KPI) in Einstein Engagement Scoring has an Audience Health indicator. This indicator shows whether the average grade for that score across all your subscribers is Excellent, Good, Fair, or Poor.
products:
  - Marketing
  - Marketing Cloud Intelligence
release: "264"
---
# Engagement Score Thresholds

Each key performance indicator (KPI) in Einstein Engagement Scoring has an Audience Health indicator. This indicator shows whether the average grade for that score across all your subscribers is Excellent, Good, Fair, or Poor.

The grades are equivalent to the likelihoods that are assigned to each subscriber:

- Excellent: Most Likely
- Good: More Likely
- Fair: Less Likely
- Poor: Least Likely

To make that determination, the Audience Health indicator uses score thresholds based on your threshold settings for Einstein Engagement Scoring.

Email

- Local Averages (default)
- Global Averages (alternative default)
- Custom Thresholds

MobilePush

- App’s Averages (default)
- Global Averages (alternative default)
- Custom Thresholds

Default averages use your business unit’s email or, for MobilePush, the app’s own historical performance. It subjectively compares your customers’ likely engagement.

For the alternative defaults, Salesforce trains each customer’s model using that customer’s behavioral data. If a customer opts in to the Global Models feature, they gain a wider pool of probabilistic contact matching using a greater range of training data. [Learn more](https://help.salesforce.com/s/articleView?id=mktg.mc_anb_einstein_sto_global_data_model.htm&release=264.0.0&type=5 "Einstein intelligence is built on data. The more data Einstein has, the more powerful the resulting predictions. Most organizations don’t have enough data to build their own models. To give you more meaningful predictive AI, Salesforce builds global predictive models that scan for aggregate, anonymous trends across many Salesforce customers. To protect your customer data, your account is opted out of the global data model by default.") about Einstein’s Global Models feature in Marketing Cloud Engagement.

Local and global averages replace the industry averages for engagement scoring for email thresholds. The local and global average default settings are optimized for the blended scores that Einstein uses.

#### See Also

- [Activate Einstein Engagement Scoring](https://help.salesforce.com/s/articleView?id=mktg.mc_anb_activate_einstein_engagement_scoring.htm&release=264.0.0&type=5 "Turn on Einstein Engagement Scoring for Marketing Cloud Engagement in Setup. Einstein requires historical contact or subscriber engagement data, including at least 1,000 events in the 90 days before activation. It can take up to 90 days to generate initial scores, but engagement scores are typically available 24–48 hours after implementation.")
- [Einstein Engagement Scoring Dashboard](https://help.salesforce.com/s/articleView?id=mktg.mc_anb_einstein_engagement_scores_dashboard.htm&release=264.0.0&type=5 "Use the dashboard to view engagement predictions for email and MobilePush channels, including the likelihood to click an email or spend time in a mobile app, and other key performance indicators (KPI). To enable Web Conversion scoring, implement Collect Tracking Code. You can analyze subscriber Persona changes over time.")
- [Einstein Engagement Scoring Model Confidence](https://help.salesforce.com/s/articleView?id=mktg.mc_anb_model_confidence.htm&release=264.0.0&type=5 "Model confidence is a way to express the relative strength of a prediction generated by a model. Understanding the concept can help you interpret Einstein Engagement Scoring for email model confidence scores.")