Overview

Analytics displays conversation outcomes, Evaluation scores, resource usage and spend for your project. Choose the dashboard cards to display, filter data by Environment and date, and open Evaluations or Transcripts to inspect individual conversations. Open Analytics in the project sidebar.

Read outcomes alongside conversation volume and cost. Photo:

Examples of when to use Analytics

Use Analytics to review agent performance, investigate conversation results and track operating cost. For example:

Analytics, Evaluations and Transcripts

How to use Analytics

Use the filters to select activity, configure the cards, then open individual results.

1

Choose a review scope

Choose the Environment and date range at the top right of Analytics. To review one release, select its Environment and version.

Choose the Environment and review period at the top right. Photo:
2

Build a useful view

Open Edit view, select the dashboard cards, arrange them within their groups and click Save.

Choose visible cards within Quality, Usage and Spend, then save the view. Photo:
3

Compare the movement

Select a headline metric to change the chart. Hover a point to read one time bucket; move away to return to the range summary.

Select a summary metric to inspect its trend and comparison period. Photo:
4

Check the underlying evidence

Open a custom Evaluation from Quality, then open a scored conversation to read its reasoning and Transcript. The Results view remembers its own filters; set its dates and Environment for this review.

Check the filters and scored population before interpreting an Evaluation. Photo:

How Analytics measurements relate

Understand why totals, judged outcomes and customer responses can differ.

These measures use different populations:

Range rates use the whole range’s counts, rather than a simple average of daily percentages. Sessions without an Evaluation verdict are excluded from the rate.

Why are Analytics results missing or different?

Start with the symptom, then follow the matching scope, scoring or loading check.

Building your dashboard

Choose and arrange the cards that support your review.

An Analytics view is the set of cards you choose to display, arranged within Quality, Usage and Spend. Open Edit view at the top right to configure it.

Saved layout and review scope

  • Saved layout: card selection and order are saved in the active version’s Analytics configuration and shared by people reviewing that version.
  • Temporary choices: the selected headline metric, Tools tab and Spend tab are remembered separately in the browser session.
  • Scoring: hiding an Evaluation card changes the dashboard; it does not disable the Evaluation’s automatic runs.
  • Shared access: recipients of a review link still need access to the project. Copying a link does not create a public dashboard.

Choose and arrange dashboard cards

Select cards, arrange them within their groups and save the view.
1

Open Edit view

Click Edit view at the top right of Analytics.

Open Edit view from the toolbar at the top right of Analytics. Photo:
2

Choose the cards and their order

Turn a card on or off. Use its drag handle to change its order within that group.

This example keeps Return policy clarity, Conversation efficiency and Customer effort visible alongside Spend for a custom-quality review, and turns the other dashboard cards off.

Choose the cards for this review; use a row's drag handle to change its order. Photo:
3

Save or discard the draft

Click Save to apply the configuration. Save and Reset are disabled while loading or saving.

To restore the standard selection and order, click Reset to default. This changes the draft in the open dialog; it still needs to be saved.

Save commits the current draft; Reset to default changes the draft before you save. Photo:

Closing the dialog without saving discards its draft changes.

4

Check the saved layout

Confirm the chosen cards appear. Reopen Edit view to check the selected cards and their order.

Reopen Edit view to verify the saved card selection and order. Photo:

Share a review

Copy a scoped link and verify the review it opens.
1

Copy the review link

Set the Environment, version and dates, then click Copy link to analytics in the toolbar.

2

Check the destination

The URL preserves the selected scope and dates. Recipients need access to the project; the selected headline metric and Tools/Spend tabs remain browser-session choices.

Adding custom metrics

Bring your own Evaluation criteria into the dashboard.

Custom metrics bring your Evaluation criteria into Analytics. Define them in Evaluations, then choose which cards appear under Quality.

Supported metrics

Analytics supports these Evaluation metric types:

  • Binary: measures a yes/no outcome.
  • Rating: measures a numeric scale.
  • Options: separates defined categories.

Text Evaluations do not become dashboard cards. The primary Resolution or Conversion measure has its own outcome card.

A custom Rating shows its aggregate score and the conversations judged. Photo:

Scoring and card visibility

  • Historical coverage: enabling an Evaluation does not backfill existing conversations; those need a batch Evaluation.
  • Empty cards: no results are shown until the Evaluation has scored conversations within the selected period.

Read an Options distribution

The distribution uses the Evaluation’s included options:

The Result filter in the destination identifies the selected category.

Add a custom metric

Define an Evaluation, collect results and show its card in Analytics.
1

Define and test the Evaluation

Open Evaluations and define an Evaluation with a name, supported Metric and Criteria. Use its test controls to score a Transcript.

2

Choose the conversations to score

Enable automatic runs for new conversations, or run a batch Evaluation on existing Transcripts.

3

Show the card in Analytics

Return to Analytics. If the card is missing, open Edit view, show the Evaluation under Quality and click Save.

4

Check the results

Open the card, then select a conversation result to read its score and reasoning.

Measure return-policy clarity

Follow one Rating from its criteria to a useful Analytics review.

The fictional retail project uses Return policy clarity, a Rating from 1 to 5. It asks whether the agent explained the return window and the next available action. It does not judge whether the customer qualifies for a refund.

Evaluation criteria
markdown
Score how clearly the agent explained the return window
and the next available action.
Use only evidence from the conversation.

Minimum (1): No usable explanation of the policy or next step.
Maximum (5): A clear return window and an appropriate next action.

Test these criteria on a Transcript that explains the return window but omits the next action. Its reasoning should identify the missing next action. Then show the Rating in Analytics to review its average and individual results. These criteria are illustrative.

Tracking outcomes

Measure resolution or conversion against the agent’s selected objective.

The project objective determines whether Analytics displays Resolution or Conversion as its primary outcome.

  • Resolution: tracks whether conversations reach a resolved outcome.
  • Conversion: tracks the outcome defined for the agent, such as completing a qualified request.
The selected objective determines the outcome shown in the headline. Photo:

Choose the project objective

When the KPI card asks you to choose an outcome, select Resolution or Conversion. This configures the project’s objective and its corresponding system Evaluation. To change an existing objective, review the agent’s Objective in project settings.

Resolution scoring

  • Scoring off: if the Resolution card offers Turn on, click it to enable automatic scoring for new conversations. Showing the card in Edit view alone does not turn scoring on.
  • Historical scoring: use batch Evaluations to score existing Transcripts with new criteria.

Define a conversion

Set the success criteria used to score conversions.
1

Open the definition

Select the Conversion headline metric and use Define to describe the outcome.

Define the successful outcome before interpreting the Conversion rate. Photo:
2

Save or cancel the draft

Click Save to update the Conversion criteria, or Cancel to discard the draft.

3

Check the saved definition

Reopen Define to check the saved criteria. Use Evaluations to test them against a Transcript.

Filtering Analytics

Focus the dashboard on an Environment, version and time period.

Filters control which activity is summarized in the dashboard’s totals, trends and cards. Open Analytics and use the Environment and date controls at the top right.

Choose the Environment and review period at the top right. Photo:

Interpreting filtered results

  • Live activity: conversation and call totals include ended live sessions. Internal Run tests are excluded.
  • Dates and timezone: sessions are assigned to their start date in the viewer’s timezone. The range includes the full start and end dates. A session that started before the range is excluded, even if it ended within it.
  • Updated results: reload Analytics after a live session ends or an Evaluation completes to request fresh results.
  • Detail-view filters: Evaluation results and Transcripts retain their own filters independently of Analytics.

Choose an Environment and version

Review all activity, one Environment or a specific version.
1

Open the Environment filter

Click All environments, or the currently selected Environment, at the top right of Analytics.

2

Choose the activity to include

  • Across the project: select All environments.
  • One Environment: select an Environment and leave All versions selected.
  • One version: select an Environment, then choose the specific version to review.
Select an Environment, optionally choose a version, then click Apply. Photo:
3

Apply and check the selection

Click Apply to update the dashboard. Confirm the Environment and version shown in the filter’s label before reading the results.

For projects with the fixed Development, Staging and Production Environments, choose the Environment directly from the menu. This menu applies the Environment directly, without a version picker.

Choose a review period

Use a preset or choose start and end dates in the calendar.
1

Open the date filter

Click the displayed date range beside the Environment filter.

2

Choose the dates

  • Preset: select Last 7 days or another preset. The dates apply immediately and the picker closes.
  • Custom range: select the start and end dates in the calendar. Use its month controls to reach dates outside the visible month. The range updates as you select dates; there is no separate Apply step.
Choose a preset or select the start and end dates in the calendar. Photo:
3

Check the review period

Confirm the start and end dates displayed in the date control before comparing the dashboard’s results.

Investigating quality

Inspect the conversations and judgments behind an aggregate score.

A custom Evaluation’s Results combines a summary with individual scored conversations. Open it from Quality, or use All evaluations to move to the full Evaluation area.

Check the filters and scored population before interpreting an Evaluation. Photo:

Check the result population

  • Independent scope: Results remembers its own date, Environment and Result filters.
  • Category entry: clicking an Options category sets the Result filter to that category.
  • Reset: click Clear filters, then select the dates and Environment for this investigation.

Check missing or failed results

  • No judgments: an outcome rate is blank when no sessions in the selected scope have a verdict. Check scoring and the selected dates.
  • Card error: click Try again on the failed card to reload it.
  • Page error: use the page’s Try again button when the dashboard fails to load.

For the built-in outcome definitions, see Resolution categories.

Continue in Evaluations or Transcripts

  • Evaluations: edit or test the criteria shown in the Prompt tab.
  • Transcripts: investigate the conversation and its execution details.

Investigate a scored conversation

Trace an aggregate score back to its judgment, reasoning and Transcript.
1

Open the Evaluation results

Open a custom Evaluation from Quality to inspect its Results. Confirm the date, Environment and Result filters for the investigation.

2

Inspect individual results

Use the results table’s sorting and navigation controls to inspect additional results. Open a conversation result to read its score and reasoning.

3

Read the conversation and criteria

Read the Transcript alongside the result’s reasoning. Open Prompt to inspect the criteria, or edit and test the Evaluation.

Customer feedback

Review customer survey ratings and written feedback.

Customer survey CSAT lets customers rate chat conversations from 1 to 5 stars. It is separate from a satisfaction score generated by an Evaluation.

Read written feedback and narrow it by rating or search. Photo:

Enable the customer survey

To enable a disabled survey, click Turn on on the CSAT card. This enables the survey in the chat widget settings.

Interpret ratings and comments

  • Response coverage: the average uses submitted ratings. The answered-survey percentage compares submitted ratings with surveys shown.
  • Search scope: search matches only comments loaded in the dialog. Clear the search, scroll to load older comments, then search again.
  • Written feedback: a customer can submit a star rating without a comment.

Read and filter customer feedback

Find comments by rating or search, then inspect their conversations.
1

Open the comments

Use Read comments to open written feedback. To start with one rating, select its star-distribution row.

2

Filter and search

In the dialog, choose a star filter or All. Search the comments and clear the search to broaden the list again. More comments load as you scroll.

3

Read the conversation

Open a comment to review its conversation.

4

Return to Analytics

Close the dialog to return to the dashboard.

Usage and resources

Investigate conversation activity and the resources the agent uses.

The Usage group shows activity and resource rankings for the selected review scope.

Read resource activity alongside the volume of conversations that used it. Photo:

Read activity and resource rankings

  • Conversation activity: Unique users counts distinct user identifiers; User messages counts messages. Voice agents also show call minutes and average call duration.
  • Resource activity: Playbooks, Workflows, tools and Knowledge documents show which resources contributed to that activity.
  • Deleted resources: a deleted resource cannot be opened from a historical ranking.

Understand response latency

Average latency is shown in milliseconds. The range average is weighted by message count; call duration is measured separately.

Find and inspect a resource

Use resource rankings to open the relevant configuration or document.
1

Choose a resource ranking

Use the cards for Playbooks, Workflows, tools or Knowledge. In Tools, switch between All, Integrations, APIs, Functions and MCPs. Scroll a populated ranking to see additional returned rows.

2

Open a resource row

Select a resource row to open its editor or document:

3

Check the current resource

The row opens the resource’s current configuration. Changes made after the selected Analytics period are reflected in that editor.

Spend analysis

Examine spending by category, model and conversation volume.

The Spend group breaks down operating cost by category and model for the selected Environment and period.

Inspect model spending alongside activity in a configured Analytics view. Photo:

Available tabs and units

  • Breakdowns: By category shows the spending mix; LLM shows the model breakdown. Provider breakdowns appear for text-to-speech and speech-to-text spending in the selected range.
  • Display unit: the account determines whether operating spend appears in credits or dollars. This is not a dashboard currency switch.

Read total and per-conversation cost

  • Total spend: the sum of operating spend in the selected range.
  • Per-conversation spend: total spend divided by the number of conversations or calls in the range.
  • Other models: combines smaller contributors into one group.

Explore a spend breakdown

Choose a breakdown, inspect its series and read individual chart values.
1

Choose the breakdown

In Spend, choose By category for the spending mix or LLM for the model breakdown. Use the speech-provider tabs when the selected activity includes the relevant text-to-speech or speech-to-text spending.

2

Inspect or toggle a series

Hover a legend entry to emphasize its series. Select an entry to hide or show that series in the chart; at least one remains visible. This changes the chart display, not the underlying usage.

3

Read a bucket or the range summary

Inspect a chart bucket to read its value, then leave the chart to return to the range summary. Check the active tab and visible series when interpreting the result.

Estimated value

Apply your own value assumptions to resolutions or conversions.

The Saved metric uses your value-per-resolution assumption. For Conversion, Revenue uses the value assigned to one conversion.

The Saved metric applies the configured value per resolution. Photo:

Interpret the estimate

  • Configured value: the dollar amount you enter is an assumption about one resolution or conversion, not measured revenue or staffing cost.
  • Calculation: for each chart bucket, Analytics multiplies conversation volume by the Resolution or Conversion rate, then by your dollar assumption. It adds those bucket estimates for the selected range.
  • Outcome coverage: the rate uses judged sessions, while the estimate applies that rate to conversation volume. With incomplete scoring, the estimate can differ from multiplying only positively judged sessions by your assumption.
  • Operating-spend currency: editing a business-value assumption does not change the account’s operating-spend currency.

Set or change a value estimate

Enter an assumption for one resolution or conversion, then check the saved value.
1

Open the value control

For Saved, select Add estimate in an unset metric to enter the value for one resolution. For Conversion, Revenue uses Add a value for the value assigned to one conversion. To change an existing value, select Saved or Revenue, then use Edit.

2

Enter a positive dollar value

Enter a dollar amount greater than zero.

Enter a positive dollar assumption for one resolution; Enter saves and Escape cancels. Photo:
3

Save or cancel the draft

Press Enter or leave the input to save. Escape cancels the draft.

Calculate an estimated value

See how conversation volume, scoring coverage and a dollar assumption affect Saved.

Using illustrative data for one chart bucket:

  • Volume: 100 conversations.
  • Resolution scoring: 8 of 10 judged conversations resolved, giving an 80% rate.
  • Value assumption: $5 per resolution, entered by the team.
  • Estimated value: 100 × 0.8 × $5 = $400.

The estimate applies the judged rate to all 100 conversations. It does not calculate 8 × $5 from only the positively judged sessions. Operating spend is displayed separately and is not subtracted from Saved. The $5 assumption is an example, not a Voiceflow default.

Use the up and down arrow keys to select a result, Enter to open it, and Escape to close the search.