Analytics

Overview

Once you've published a version of your Actions project and users have invoked your Action, you'll start seeing analytics data in the Actions console, usually within 24 hours. These analytics provide information about your Action's usage, health, discovery, and directory listing.

Usage

The Usage page displays three graphs, which pertains to the usage data of your Action. Data in the following graphs can be exported in CSV format.

This graph shows usage over time.

Customize the data displayed in the following ways:

  • You can see daily, weekly, and monthly usage by clicking the corresponding button at the top of the graph.

  • Clicking on the date range in the bottom left corner of the graph, allows you to set specific time frames to view.

  • An insight for the total numbers of unique users for the selected date range is shown at the bottom of the chart.

  • Clicking Filter alters the graph using the following data:

    • Surface: Filter usage by surface used (for example: mobile device, smart speaker, smart display, TV, wearable, auto, feature phone, etc).
    • Action: Filter usage by a specific intent.
    • Language: Filter usage by specific languages your Action is available for.
    • Country / Region: Filter usage by specific countries and regions your Action is available for.

Data can be exported from the charts, in CSV format.

User retention

This graph shows user retention for the last five weeks. This includes all users (including new users) that have used your Action.

Customize the data displayed in the following ways:

  • Clicking Filter alters the graph using the following data:

    • Surface: Filter usage by surface used (for example: mobile device, smart speaker, smart display, TV, wearable, auto, feature phone, etc).
    • Action: Filter usage by a specific intent.
    • Language: Filter usage by specific languages your Action is available for.
    • Country / Region: Filter usage by specific countries and regions your Action is available for.

Data can be exported from the charts, in CSV format.

Conversation data

This graph shows conversational data related to your Action and includes the following tabs:

  • Conversations: The number of conversations users had with your Action. A conversation is a single, complete interaction with your Action, starting from invocation and ending with the user exiting the context of your Action.
  • Messages: The number of conversational turns across all conversations.
  • Avg length: The average length, in seconds, that isers engaged with your Action before exiting the conversation.
  • Abort rate: The percentage of time users exited from or aborted from a conversation with your Action.

Customize the data displayed in the following ways:

  • Clicking on the date range in the bottom left corner of the graph, allows you to set specific time frames to view.
  • Clicking Filter alters the graph using the following data:

    • Surface: Filter usage by surface used (for example: mobile device, smart speaker, smart display, TV, wearable, auto, feature phone, etc).
    • Action: Filter usage by a specific intent.
    • Language: Filter usage by specific languages your Action is available for.
    • Country / Region: Filter usage by specific countries and regions your Action is available for.

Data can be exported from the charts, in CSV format.

Health

The health section helps you track the following metrics related to the quality and health of your Action:

  • Errors: The number of errors your Action's cloud endpoint returned on a given day. If you have a large number of errors, you may want to look at your logs to identify what is causing your endpoint to crash or behave unexpectedly.

    • Select the View logs link to view your project's logs in the Google Cloud console. You can learn more about reading these logs on the Troubleshooting page.
  • Action latency: The latency of your Action's endpoint. If latency is very high or regularly spikes, your users may be experiencing delays while interacting with your Action.

  • Experienced latency by user: The latency felt by a user on each request to your Action. This metric illustrates what users experience when interacting with your Action.

Data can be exported from the charts, in CSV format.

Clicking on Filter gives you the following additional viewing options:

  • Surface: Filter by mobile, smart speaker, and smart display devices.
  • Action: Filter by specific Actions.
  • Language: Filter by the languages your Action supports.
  • Country/Region: Filter by country/region where your Action is available.

Discovery

The discovery section shows information about how users are discovering your Action.

When users say a phrase related to your Action, Google will sometimes recommend that the user try your Action, even if the user did not explicitly specify your Action's invocation name. The discovery page displays a table of which phrases led to Google recommending your Action. The columns in this table are as follows:

  • Invocation: The user query that led to Google recommending your Action. Aside from the actual user queries, this list includes the following values:

    • BUILT_IN_INTENT - This listing indicates your Action was invoked through a built-in intent.
    • AUTO_MATCHED_BY_GOOGLE - This listing indicates when implicit invocation was used.
    • ACTION_LINK - This listing indicates when your Action was invoked through an Action link.
  • Intent: The intent the user’s query matched.

  • Impression: The number of times this phrase led to Google recommending your Action.

  • Selection: The number of times a user invoked your Action after Google recommended it. This number cannot exceed the number of impressions this phrase has.

  • Selection rate: The percentage of impressions that led to a selection. A low percent rate indicates that many users are opting to use other Actions for this particular query, whereas a high rate indicates that your Action is popular for this query.

Data can be exported from the charts, in CSV format.

Directory

The directory section shows traffic and ratings for your Action in the Assistant directory.

  • Ratings summary: Your Action's average rating, total number of ratings, and number of users who rated your Action at each score.
  • Average rating: Your Action's cumulative average rating over time, or the average of all reviews on a given day.
  • Ratings over time: The distribution of your Action's cumulative rating scores over time or on a given day.
  • Assistant directory page views: The number of times your Action's directory page has been visited per day. Views can be filtered by country/region, language, and other useful categorization options.
  • Invocations initiated from Assistant directory: The number of times your Action has been invoked by the user pressing the "Try it" button in the directory. Invocations can be filtered by Action, country/region, and other useful categorization options.

Data can be exported from the charts, in CSV format.

Additional analytics tools

Once you've deployed your Action, you may want to track things like adoption, usage, and quality to help you analyze the current state of your Actions and answer questions like:

  • How many users do I have? Are my users increasing?
  • Are users confused about what to do or do they say things that my Action doesn't understand?
  • Do users spend a long time with my Action, or do they regularly cancel the conversation?
  • Which of my Actions are most popular, and which have low engagement?

This data helps you identify opportunities to improve the long term success of your actions. To let you obtain this data, we've let analytics tools providers integrate with Actions on Google, and you can set up these tools in as little as five minutes:

Additional monitoring tools

As you develop and deploy your Actions, you might benefit from adding logging, monitoring, and alerting. Monitoring and alerting tools track uptime and notify you of errors and exceptions. With these tools, you can also debug and test Actions with real requests from the Assistant. The following monitoring tools have integrated with Actions on Google: