Query your data in natural language with Gemini assistance

Preview

This product or feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA products and features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

Conversational Analytics is an AI-powered data querying tool that helps you write questions in natural language, empowering those with no expertise in business intelligence to gain value from your data. To produce the most reliable answers possible, Conversational Analytics uses your LookML models to understand how it should query your data.

Learn how and when Gemini for Google Cloud uses your data. As an early-stage technology, Gemini for Google Cloud products can generate output that seems plausible but is factually incorrect. We recommend that you validate all output from Gemini for Google Cloud products before you use it. For more information, see Gemini for Google Cloud and responsible AI.

Before you begin

To use Conversational Analytics, you must meet the following requirements.

1. You must be a user under a Looker Studio Pro subscription. Looker Studio Pro licenses are available at no cost to Looker users.
2. An administrator must have enabled Gemini in Looker for Looker Studio.
3. The dataset that you want to analyze must be in Google BigQuery, Google Sheets, a CSV file, a data extract, or Looker (which is a different product than Looker Studio).

Supported data sources

Before you can ask questions of your data, you must have data sources connected to Looker Studio. Conversational Analytics works with the Google BigQuery, Google Sheets, CSV, data extracts, and Looker data connectors.

Set up connections to these data sources by following the instructions on the following pages:

• Connect to BigQuery
• Connect to Google Sheets
• Upload CSV files to Looker Studio
• Extract data for faster performance
• Connect to Looker

You can view the data sources that have already been added to your Looker Studio by navigating to the Data sources page in Looker Studio.

Limits of data sources

Be aware of these data source limitations:

• Conversational Analytics doesn't support BigQuery's Flexible Column Names feature.
• Conversational Analytics doesn't work well with data sources that have field editing in reports disabled because this setting prevents Conversational Analytics from creating calculated fields.
• When the data source is Looker, Conversational Analytics cannot override the default value of an always_filter parameter or a conditionally_filter parameter.
• When the data source is Looker, Conversational Analytics cannot set the value of a filter-only that is defined using the LookML parameter parameter.

Navigate to Conversational Analytics

You can access Conversational Analytics in the following ways:

• Navigate directly to Conversational Analytics.
• Choose Conversational Analytics from the navigation panel of Looker Studio.
• Choose Conversation from the Create menu of Looker Studio if you are in your Sandbox workspace.

Note: If Data agents have been created for your instance, navigate to the Data sources tab on the Chat with your data page to use Conversational Analytics.

Manage conversations

Sets of questions that you ask about a dataset are organized by conversation. Splitting work into multiple conversations can be useful for organizing lines of inquiry. Previous conversations are listed in the schedule Recent panel of Conversational Analytics. Clicking on any existing conversation lets you return to the conversation and ask additional questions.

Create a new Conversation

To create a new conversation, follow these steps:

1. Navigate to Conversational Analytics.
2. Click + Create conversation within Conversational Analytics. Questions and results are saved for future viewing in the "Sandbox" location of the project that you choose.

3. Select the data source that you want to investigate or the data agent that you want to use for your conversation:

   • Data source: To start a conversation based on an existing data source, select the Data source panel, and then select a data source. To create a new data source, select Connect to data.

     Tip: If you're having trouble making a data source appear, validate that it's one of the supported data connector types.

   • Data agent: To start a conversation with an existing data agent, select Agents, and then select a data agent. To create a new data agent, select + Create agent.

4. To start the conversation, enter your question and press return (Mac) or Enter (PC).

You can return to the conversation from the schedule Recent section.

Name a conversation

Conversational Analytics automatically generates a conversation's title that is based on your first question and response. To change the generated name, follow these steps:

1. From the Recent panel, open the conversation.
2. Click the title at the top of the page.
3. Enter a new conversation name.
4. To save your changes, click elsewhere on the page, or press return (Mac) or Enter (PC).

Delete a conversation

You can move a conversation to the trash, restore a conversation from the trash, or permanently delete a conversation.

Note: Conversations that remain in the trash for more than 30 days are automatically deleted.

Move a conversation to the trash

To move a conversation to the trash, follow these steps:

1. Navigate to Conversational Analytics.
2. From the Recent panel, open the conversation that you want to move to the trash.
3. Within the selected conversation, click Move to trash.

Restore or permanently delete a conversation

To restore or permanently delete a conversation from the trash, follow these steps:

1. Within Conversational Analytics, select Trash in the left navigation panel to view the list of conversations that have been moved to the trash.
2. In the Trash section, click the name of the conversation that you want to restore or permanently delete.
3. In the Are you sure? dialog, select one of the following options:
   • Cancel: Cancels the action.
   • Restore: Restores the conversation. The conversation can be accessed from the Recent section of the left navigation menu within Conversational Analytics.
   • Delete forever: Permanently deletes the conversation.

Search conversations

To search for a specific conversation by title, follow these steps:

1. Navigate to Conversational Analytics.
2. In the Search Conversational Analytics search bar, enter your search query. As you type, a list of conversations with titles that match your search query will appear.
3. Select a conversation from the search results to open that conversation.

Ask questions

You can ask questions to get insights from your data. You can use suggested questions as a starting point for exploring data and building familiarity with Conversational Analytics.

Ask questions about a data source

Once you have created a conversation, you can ask questions about the data in the send_spark Ask a question field at the bottom of the screen.

The questions don't need to be in a specific format or use a specific syntax. However, they do need to relate to the data source that you've selected. Conversational Analytics may rephrase your question after you've written a query, and the rephrased question will be displayed in the conversation window following your original question. For example, Conversational Analytics might rephrase the question "What is the mean of user ages?" to "What is the average user age?"

Conversational Analytics will take previous questions and answers into account as you continue the conversation. You can take previous answers and build on them by further refining results or changing the visualization type.

Note: You cannot reference previous questions and answers from other conversations, only the one that you are actively in.

Supported questions

Conversational Analytics supports questions that can be answered by a single visualization, for example:

• Metric trends over time
• Breakdown or distribution of a metric by dimension
• Unique values for one or more dimensions
• Single metric values
• The top dimension values by metric

Conversational Analytics doesn't yet support questions such as the following:

• Percent change of a metric over time, including period-over-period analysis
• Prediction and forecasting
• Advanced statistical analysis, including correlation and anomaly detection

For more guidance on creating questions, see Best practices.

Suggested questions

Starting with a suggested question can help you get a conversation started if you're unfamiliar with the data or unsure how to start. When you're beginning a new conversation, Conversational Analytics suggests starting questions under the heading What questions can I ask?. Click a suggested question to generate an answer.

You can also find suggested questions once a conversation has started in the collapsable storage Data panel under the heading Try asking:. Click a suggested question to generate an answer.

Get additional insights

When Conversational Analytics is able to provide additional data insights about a response, an Insights keyboard_arrow_down button will appear below the response. Click Insights keyboard_arrow_down to see additional information about your query. Insights only analyzes the data that was returned by your prompt and won't run additional queries to fetch additional data. Insights can be a useful source for ideas for follow-up questions to continue the conversation.

The following is an example of some insights that might be returned by the prompt "How many users are in each state?":

• A general summary of high and low data volume areas. For example:
  • "California, Texas, and Ohio are key states for business operations based on the data provided."
  • "England and specific regions in China, namely Anhui and Guangdong, show significant business activity."
  • "Some states, including Mie, Akita, and Iwate, have minimal presence based on the data."
• An assessment of the variability of the dataset. For example, "The data indicates varying operational scales across different locations."

Determine how an answer was calculated

To see how Conversational Analytics arrived at an answer or created a visualization, select How was this calculated? below the response. The How was this created? section includes the following tabs:

• Text: Provides a plain text explanation of steps taken by Conversational Analytics to arrive at the given answer. This explanation includes the fields that were used, the calculations that were done, the filters that were applied, and other details.
• Code: Provides the exact query that was sent to the database.

Best practices

Review the following best practices to help Conversational Analytics provide the most helpful answers.

Set up a data source

Setting up a data source in the ideal manner can help Conversational Analytics provide the most helpful answers. Consider following these best practices when creating a data source:

• Only include fields in the data source that should be used for analysis by end users.
• Give each field a clear and concise name. For Looker data sources, the field labels that are defined in Looker are automatically used by Looker Studio.
• Give each field a clear description, including sample values where relevant. These field descriptions are included in the prompt that is sent to Conversational Analytics, and they can be helpful for providing context. Sample values are especially helpful for string fields.

If you are using a Looker Studio data source, consider these additional best practices:

• You can exclude fields that shouldn't be used for analysis entirely, or hide them in the data source.
• Field descriptions can be added or edited to give context to Conversational Analytics.
• If you are seeing unexpected results, check your data source and confirm that the field types and default aggregation settings are correct.

Prompting

When writing questions for Conversational Analytics, consider following these best practices:

• Use the exact field names that are included in the data source when possible. This will help Conversational Analytics disambiguate similarly named columns.

  Tip: You can view these field names by clicking open_in_new View fields in the collapsable storage Data panel.

• To narrow down results by including or excluding specific data, state the field and the filter value directly when possible. For example, instead of asking for "German sales", ask for "sales where the country is Germany" or "sales where the region is DE".

To account for things like similarly named columns, or to let users use the terms "German," "Germany," and "DE" interchangeably, consider creating a custom Data agent that can handle these variations.

Data agents

Data agents build on the power of Conversational Analytics to further refine the experience of users with no expertise in business intelligence to gain value from your data. With Data agents, you are able to customize the AI-powered data querying agent with context and instructions specific to your data.

For example, maybe you define a "loyal" customer as one who has made more than five purchases within a certain timeframe. Or, you want to save your users time, so all responses from your data agent should be summarized in 20 words or fewer. Also, you want numbers to be formatted to match company standards. These types of instructions and more can be used to build a data agent that knows how your users want to interact with your data. Visit the Data agents documentation page to learn more.

Provide feedback

You can provide feedback to Google about Conversational Analytics in either of the following ways:

• Rate individual responses by selecting one of the following options:
  • thumb_up Good response: Indicate that the response was helpful.
  • thumb_down Bad response: Indicate that the response was not helpful.
• To send detailed feedback, click Send Feedback at the bottom of the left navigation panel in Conversational Analytics.

If you're sharing negative feedback, you'll have the option to include additional details, including a copy of the conversation that you were having.

Important: When sending feedback, don't share sensitive information with Google such as passwords, payment information, or personal customer details.