Tableau Cloud Help

Explore groups of metrics with enhanced Q&A (Discover)

When enhanced Q&A (Discover) is turned on, it's surfaced on the Tableau Pulse home page. For metrics that are sorted by data source or definition, the Discover with AI button appears. When you ask a question with this button, Tableau Pulse looks across all of the metrics in the group to find insights. For metrics sorted by data source, definition, or time range, there are suggested questions for you to explore. These questions look across a sub-set of metrics in the group.

Enhanced Q&A highlights shared contributors, detects when trends move together or in opposite directions, flags entities that are outliers across multiple metrics, and identifies metrics that are jointly above or below expected ranges.

Important: Enhanced Q&A (Discover) is a Tableau+ feature that must be turned on in Tableau Cloud site settings. It requires a connection to a Salesforce org that has Einstein generative AI set up for the selected Salesforce org. For more information, see Turn on Enhanced Q&A (Discover) in Tableau(Link opens in a new window).

Ask questions with enhanced Q&A (Discover)

To start using enhanced Q&A (Discover), ask a question about the metrics that you follow. For examples of the types of questions that you can ask, see Example questions for enhanced Q&A (Discover).

1. On the Tableau Pulse home page, under a group of metrics, select Discover with AI. The Discover with AI button appears only if you have your metrics sorted by data source or metric definition. If you aren’t sure what to ask, you can select one of the suggested questions.

   

2. In the text box that appears, enter a question about your data. You can ask questions in natural language to explore different regions, segments, or time periods. You can also ask time-based scoping questions such as "What happened last month compared to this month?" and questions that filter metrics based on dimension values such as "Show sales for California" or "Compare APAC vs. EMEA".

3. On the Discover page, an insight brief answers the question. Click a link at the end of an insight to see its source. Click Metric Sources to see metrics considered as part of the analysis when the response is generated. Occasionally, a metric source might not be clickable if Tableau Pulse couldn't generate a metric with those specific filters, but the insights that use that metric source are still rooted in your data.

   
   

4. Select other questions to explore more insights, or enter a question in your own words in the text box located near the bottom of the Discover page.

Example questions for enhanced Q&A (Discover)

Use these questions as inspiration to guide your conversation with enhanced Q&A.

Insight Type	Example Question
Period Over Period Change	What happened to admission rates in the last 30 days?
Correlated Metrics	How are revenue and deliveries relating to each other in the last 90 days?
Record-level Outliers	Which transactions are outliers?
Current Trend	What's the trend for call time this week?
Unexpected Values	Is CSAT lower than expected?
Goal and Threshold Breakdown	Are refunds in the southwest below the threshold?
Pace to Goal	How are we trending towards our donation target for this month?
Top Drivers	Which product lines are driving revenue growth?
Top Detractors	Which product lines detracted the most from revenue?
Concentrated Contribution Alert	Are there any risky monopoly scenarios?
Top Contributors	Which account executives contributed the most to sales this quarter?
Bottom Contributors	Which menu items contributed the least to profit this year?
Cross-Metric Analysis	What's up with churn rate and new applications?

How to set up enhanced Q&A (Discover)

Enhanced Q&A (Discover) requires a Salesforce org to be connected to Tableau Cloud. The connection settings are listed at the beginning of the AI in Tableau section of General settings in Tableau Cloud. Also, Einstein generative AI in Salesforce must be set up for the selected Salesforce org.

To make Discover available in Tableau Cloud, your Tableau Cloud admin must make configuration changes in both Salesforce and your Tableau Cloud site. The complete instructions are described in the help topic Configure Tableau Cloud sites for AI(Link opens in a new window).

The trial of enhanced Q&A (Discover) doesn't require a dedicated Salesforce org. For information about how to start the trial, see Start your trial of enhanced Q&A (Discover).

Ideal (and less ideal) data scenarios for Discover in Tableau Pulse

To get the best experience when using Discover in Tableau Pulse, use data that is clean, structured, and validated. Connecting to data that is consistently formatted and accurate produces better metrics and more reliable, actionable insights.

Data with the following characteristics provides the best experience with Discover:

• Time-series data: Data that evolves over time (for example, sales, traffic, financial metrics).

• Aggregated metrics: Key business numbers that can be summarized as metrics (for example, total sales, average satisfaction, loan approval counts).

• Clear dimensions: Data sets with well-defined categories such as region, product, or customer segment.

Data scenarios that provide a less ideal experience include:

• Highly granular data: Data at the highest level of detail with very fine granularity such as minute-by-minute logs.

• Single, point-in-time data: Static data sets such as one-time survey results.

• Unstructured data: Free-form text or multimedia that lacks structure.

Time-based filtering

Discover can handle time-based scope adjustments. For example, you can refine your question to compare data from different time periods, such as asking, "What happened last month compared to this month?" Use questions to adjust the granularity or breadth of the insights you're exploring.

Dimension-based filtering

Discover can handle questions that filter metrics based on dimension values (region, product, segment) or compare dimension members (West versus East) to uncover meaningful patterns.

Discover supports filtering and compare metrics by dimension members. For example, you could ask "Show sales for California" or "Compare APAC vs. EMEA" to automatically expand your exploration to the full set of compatible metrics.

Note: Questions you ask in the Discover page can only filter or compare metrics that share key traits, like the same time granularity or the same set of dimensions. If a metric doesn’t match, it won’t be included in the answer.

Cross-check insight sources for transparency

When using Discover, you can cross-check the insights it generates by clicking the link at the end of each insight. These links cite the metric source and display the insight chart when available. You can use this information to verify that the insights Discover provides are grounded using the correct data.

Multilingual support for questions and responses in Discover

Enhanced Q&A (Discover) supports multilingual questions and responses, aligned with Tableau Cloud supported languages. Suggested questions, follow-up questions, and insight briefs adapt to the site’s language setting. Metric names and filters always remain in their original, authored language for clarity and consistency.

Discover can fluently adapt to the language that you enter as a query. It detects the input language as you ask questions and returns LLM responses in that language—even if it differs from the site or data language.

Note: Einstein Trust Layer masking supports 6 core languages; others may reduce response quality. Metric names stay in original language. Input language detection may vary in unsupported locales.

When the LLM attempts to negotiate differences between language input, it can get confused with certain languages and sometimes blend those languages in its response. With current LLM capabilities, for certain languages such as Chinese, the results may be inconsistent. Specifically, entering questions in traditional Chinese can result in responses that contain simplified Chinese, which can make the response difficult to read.

Limitations of Discover in Tableau Pulse

Enhanced Q&A (Discover) is limited to working with insights from Pulse metrics, so it can't generate insights from data outside the Pulse framework. It can't answer questions about columns that aren't used in the metrics or metadata about the data source such as "Is this data certified?"

It can filter or compare metrics that share key traits, like the same time granularity or the same set of dimensions. If a metric doesn’t match, it won’t be included in the answer.

While Discover strives to provide accurate and actionable insights, occasional hallucinations (inaccurate or off-topic answers) may occur, especially when you ask questions that create more complex queries. LLMs can hallucinate because they generate responses based on patterns in data and probabilities, not direct access to real-world facts. As a result, generative AI can be prone to creating output that sounds plausible but contains incorrect or irrelevant information.

For more information, see AI in Tableau and Trust(Link opens in a new window) and Reviewing generative AI outputs(Link opens in a new window).

Einstein Trust Layer masking for Discover

The Einstein Trust Layer uses advanced pattern matching and machine learning techniques to detect sensitive data in prompts. Personally Identifiable Information (PII) such as names and email addresses are automatically masked using machine learning. For best results, consider structuring your data with clear entity identifiers and consistent field labels to minimize ambiguity and ensure accurate results when masking is applied.

Ask questions about a single metric with Ask Q&A

Ask Q&A is available for single metric exploration.

If you don't see an insight you are looking for, click the Ask button to the left of the available questions, and then enter a new question about the metric. Based on your question, Tableau Pulse will look for and rank semantic matches for insights that it detects for the same metric you're investigating. It then lists a question for each detected insight. The filter settings and data available for the metric affect the insights that Tableau Pulse can detect. For more information, see How Tableau Approaches NLP with Q&A to Empower Everyone(Link opens in a new window).

Ask Q&A can be enhanced by turning on the AI in Tableau setting for better semantic matches in Tableau Cloud site settings. For more information, see Turn on better semantic matches for Ask Q&A in Tableau(Link opens in a new window).

Explore a single metric with Ask

1. In the Tableau Pulse home page, click a metric of interest.

2. In the Insight Exploration page, under Discover Top Insights (following the chart visualization), click a suggested question.

3. To ask your own question, click the Ask button.
   

4. Enter your question about the metric.

   Try asking questions about trends, changes, outliers, or categories that may be contributing to the metric. You can also ask questions that adjust the time granularity and apply dimensional filters for the current metric view. For example, you could enter "Sales in California last week," "Zoom into technology," or "What about yesterday?"

5. Based on semantic matching and ranking, Pulse lists questions for each detected insight. Click a question to see the insight.

   If Pulse can't find a match, it might suggest other questions of interest.