Skip to main content

Understand customer experience at scale with the CX Score

Get full visibility into customer experiences with CX Scores—across both human and Fin interactions.

CX Score measures how customers feel about their support experience - across both Fin and human interactions. It gives support leaders and teams full visibility into customer sentiment, without relying on surveys.

Use this article to understand how the CX Score works, view your scores in the Performance report and Topics Explorer, compare your performance against industry peers using the Benchmark panel, and apply best practices to improve your score over time.

Note: The CX Score is currently available to all Fin customers who meet the following criteria:

  • Have imported conversations from your source platform (e.g. Zendesk or Salesforce) which you can toggle on and off.

  • Have deployed Fin.

  • Have the Pro add-on included in your subscription.


What challenges does the CX Score help solve?

The CX Score is helping support leaders and teams address the following issues:

  • Surveyed CSAT is incomplete and biased - It captures feedback from a small, self-selecting group, often over representing extreme opinions.

  • Customers score AI and human agents differently - Customers tend to score AI support more harshly, making it hard to assess AI fairly.

  • Disengaged customers go unmeasured - Customers who don’t respond to surveys - including those who had a neutral or frustrating experience - are often left out of satisfaction metrics.

  • There’s no scalable way to measure satisfaction across all AI conversations - Without full coverage, teams struggle to track trends, identify friction points, and improve overall support quality.

  • Survey design can skew results - Small changes, like survey timing or phrasing, impact scores without support quality having changed.

The CX Score addresses these gaps by scoring every meaningful interaction—without relying on surveys—so support teams can get a fuller, more consistent view of customer sentiment across both Fin and teammate conversations.

Diagram showing the CX Score model — a 1 to 5 rating scale with positive and negative CX Score reason categories mapped across the range.

Note: Fin needs to be set live to see your CX Score across all conversations.


How CX Score works

The CX Score uses machine learning to score meaningful conversations based on specific factors called CX Score Reasons — the named signal categories the model evaluates in each conversation.

The model analyzes the conversation to identify clear signals—positive or negative—across the following categories:

  • Answer quality (Fin AI Agent): Measures how clear and accurate the Fin AI Agent’s response was, and if it resolved the issue without contradiction.

  • Answer quality (Teammate): Measures how clear and accurate a human teammate's response was, and if it resolved the issue without contradiction.

  • Customer effort: Measures how much work the customer had to do to get help (e.g., repeating themselves or multiple handovers).

  • Strong emotion: Detects if the customer expressed strong positive or negative feelings, such as joy, gratitude, frustration, or anger.

  • Product/Service feedback: Captures praise or criticism of the product (features, bugs) or the service provided (delivery, reliability).

  • Policy feedback: Highlights positive or negative responses to company policies, such as refunds or account rules.

Each reason is evaluated independently. If a clear signal is found, the model assigns a value (positive, negative, high, or low). If there is no strong indication, the attribute is considered neutral.

There is no hidden formula or weighting where one reason influences the score more than another. Instead, the model looks at the combination of non-neutral reasons to derive a holistic CX Score rating from 1 to 5.

Specific criteria for CX Score

A conversation will receive a CX rating once the following conditions are met:

  • The conversation has been closed.

  • There is enough conclusive information in the conversation to determine a rating.

  • The conversation is over messenger or email (not phone). Phone conversations are not scored and won't appear in your CX Score data.

Note: We recently expanded our criteria to include broader coverage. Short or low-context conversations that previously went unscored may now receive a rating, provided there is a clear signal to evaluate.

To keep scores accurate, the model automatically filters out spam.


How to access your CX Score

How to view your CX Score in the Performance report

To see your CX Score, go to Analyze > Performance and you will see the CX Score chart in the first section of the report.

You'll see the percentage of positive customer ratings out of all ratings received for conversations handled exclusively by Fin, without any human agent involvement.

If the CX Score chart isn't visible, check that you meet all three eligibility criteria in the note at the top of this article.

Screenshot of the Fin Performance report showing the CX Score chart in the first section, with a percentage of positive ratings and a breakdown by CX Score reason category.

This score reflects how satisfied customers are with Fin's service. In the CX Score chart, click Drill-in to view the conversations behind Fin’s CX Score, along with a summary that highlights key moments and explains why each score was given.

Screenshot of the Drill-in view showing a list of rated Fin conversations. Each row includes the CX Score rating and an AI-generated explanation of the key factors that influenced it.

How to track CX Score by topic

The Topics Explorer (Analyze > Topics Explorer) lets you track your CX Score performance by topic. Each topic shows a CX Score so you can see which issue areas are handled well and which need attention.

Focus where it matters most by identifying high-volume, poor CX Score topics and click on them to see the tree map and line charts broken down by subtopics. This enables you to make targeted improvements to the most impactful subtopics by addressing the root cause of the volume and negative CX.

Screenshot of the Topics Explorer showing a list of conversation topics ranked by volume, with CX Score overlaid for each topic.

To go deeper, filter the Topics Explorer by a specific CX Score reason to pinpoint which topics are driving scores for that reason. For example, filter by "CX Score reasons is Negative product feedback" to see which topics generate the most negative CX Scores for that reason specifically, then drill into those topics to understand the root cause and take targeted action.

Note: You must have enough eligible conversations to start seeing topics with CX Scores. Learn more about AI-generated topics and subtopics.

CX benchmarking

The Benchmark panel in the Fin Performance report lets you compare your CX Score against anonymised peer groups, segmented by industry. It's available to Pro plan customers only.

How to compare your CX Score to industry benchmarks

Screenshot of the Benchmark side drawer in the Fin Performance report, showing the industry selector dropdown, score distribution bars comparing your workspace to peers, a density curve with a 'you are here' marker, and 'What's standing out' cards.

How to open the benchmark panel

The panel includes:

If you don't see the Benchmark button, confirm your workspace includes the Pro add-on.

  • Industry selector — a dropdown at the top of the panel, defaulting to "All Industries". Over 15 industries are available with search and filter. Industries with fewer than 30 qualifying workspaces don't appear as options to protect customer privacy.

  • Score distribution — two horizontal stacked bars comparing your workspace's CX rating breakdown (ratings 1–5) against your peer group's distribution.

  • Density curve — a smooth chart showing how peers are distributed across the score range, with a "you are here" marker indicating where your workspace sits.

  • "What's standing out" cards — highlights CX reason rates where your workspace differs meaningfully from peers. Hover over a card to see your workspace rate vs. the peer median.

Limitations

  • Benchmark limitations and eligibility

    The following limitations and eligibility requirements apply to the CX Score benchmark panel:

  • Minimum conversation threshold — workspaces with fewer than 75 Fin-involved conversations are excluded from benchmark data.

Note: Your data is always anonymised. No individual workspace's score is visible to others — only aggregated, anonymised distributions are shared across peer groups.


Getting the most from your CX Score

To maximize the value of your CX Score, we recommend the following:

Action

What to check

Track trends and patterns

  • Are scores improving or declining after changes in support content, workflows, or product updates?

  • Do certain categories (e.g., AI Agent vs. Teammate) perform better over time?

Investigate low-scoring conversations

  • Are common themes emerging in low-rated interactions?

  • Are customers dissatisfied with specific responses or resolution times?

Explore AI-generated explanations

  • What specific CX Score reasons (e.g., Answer Quality, Strong Emotion) influenced the rating?

  • Are there common friction points affecting satisfaction?


FAQs

How is the CX Score different from CSAT?

Traditional CSAT surveys offer a narrow and often biased view of customer satisfaction. Response rates are low, feedback tends to reflect only extreme experiences, and AI support is often rated more harshly than human-led conversations.

The CX Score addresses these gaps by scoring every meaningful interaction—without relying on surveys—so support teams can get a fuller, more consistent view of customer sentiment across both Fin and teammate conversations.

I don’t agree with some of the CX Scores. What can I do?

Just like with CSAT, if you believe a CX Score isn’t accurate or shouldn’t be counted, you can tag the conversation (e.g., with “exclude_from_cx”). Then, use that tag to filter out those conversations in your reports.

Why can’t I filter by CX rating on some charts in the Custom Report?

"CX Score rating" and "CX Score rating explanation" attributes are only available on metrics that are derived from Conversations dataset. We have also added a new attribute called "CX Score Reasons" which allows you to filter or segment charts to see which reasons are influencing CX Scores.

Why don't I see my industry in the dropdown?

Industries only appear when there are at least 30 qualifying workspaces in that group — a threshold that exists to protect customer privacy. If your industry isn't listed, select All Industries to compare against the full benchmark dataset.

Why isn't my workspace included in the benchmark?

Your workspace needs at least 75 Fin-involved conversations to qualify for inclusion in benchmark data. If you're below this threshold, the benchmark panel won't reflect your workspace in the peer comparisons.

Is my data shared with other customers?

No. Only anonymised, aggregated distributions are used in benchmarks. No individual workspace's score is ever visible to others.

Why does my benchmark data feel outdated?

Benchmarks currently use a fixed 4-month data snapshot (December 2025–March 2026) rather than live data. This means peer group distributions won't reflect very recent changes in the industry.

Can I benchmark human agent conversations?

No. The benchmark panel only includes Fin conversations. Human agent, workflow, and other agent type conversations are excluded from the benchmark data.


💡Tip

Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts


Did this answer your question?