Claude Dashboards + Motion: From KPIs to Video

Back

Kevin Riedl

13 min read · 9 Oct 2026
Last reviewed October 9, 2026

Next

A product lead asks why the share of new customers reaching their first useful result has fallen. Someone checks the data, prepares a chart for the team meeting and writes an explanation for colleagues who miss it. Claude Dashboards and Claude Motion bring those outputs into the same working environment.

The useful combination is a dashboard people can inspect and a short animation that explains one finding. For a founder or product team, that could make the route from a question to a shared decision easier to manage. The value depends on carrying the correct meaning from the analysis into the video.

This guide works through a hypothetical SaaS onboarding report, including the metric definition, two reusable prompts and the handoff between the tools.

Documentation review: 9 October 2026. The figures, prompts and pilot are original illustrative examples, not results from a Wavect client or a hands-on product benchmark.

What are Claude Dashboards and Claude Motion?

Claude Dashboards builds explorable views of connected business data. Claude Motion turns supplied content into editable animations that can be exported as MP4 video. Anthropic introduced both in beta on 8 October 2026. Its launch announcement demonstrates a dashboard built from bike-share data and then animated with Motion.

Motion's code-based approach matters when a report changes: a number, label or scene timing can be edited without treating the whole animation as a single piece of footage. Anthropic explicitly says Motion does not use a video-generation model to create realistic footage or people. For this use case, think animated evidence and explanation.

Dashboards can use documented data-platform examples such as BigQuery, Databricks and Snowflake, as well as connected apps such as Salesforce. The warehouse example below is about product events. Adapt the metric to the source you actually have.

Which Claude plans include both features?

The current Artifacts plan matrix, checked 9 October 2026, distinguishes the two betas:

A Team or Enterprise plan is currently needed for the complete workflow. The Artifacts admin guide requires cloud code execution and file creation. Both templates are off by default on Enterprise; Motion is on by default on Team, subject to organization settings. An owner can configure the available templates and connectors.

Both features count toward plan usage. Longer or more complex animations consume more usage than a typical message. Budget for the investigation, revisions and export workflow, and check any separate data-platform costs. A short first pilot gives you a better estimate than assuming every weekly update will require the same amount of work.

Worked example: more activations, a lower activation rate

Suppose your SaaS product creates reports from a customer's connected data. You define activation as a new account connecting a source and then completing its first report within seven days of signup.

Use two complete signup cohorts: 14–20 September and 21–27 September 2026, with UTC boundaries. The data cutoff is 5 October 2026 at 00:00 UTC. Each account therefore has its full seven-day observation window. All numbers below are hypothetical.

A video saying “activations increased by 5%” would describe the count correctly: 420 is 5% above 400. It would still omit a decision-relevant fact. Eligible accounts increased by 20%, while the share reaching activation fell from 40% to 35%.

The rate change is minus five percentage points, or a 12.5% relative decline from the previous rate. Those are different ways to express the same change. Put the denominator beside the result so viewers know which one you mean.

A useful headline is: “More accounts activated, but seven-day activation fell from 40% to 35% as the signup cohort grew.” That observation supports an investigation. It does not establish that a new onboarding screen caused the decline. Acquisition mix, plan mix, instrumentation changes or other differences could also matter.

If your team has not agreed on what activation means or which decision the metric should inform, start with our guide to product-market-fit evidence and cohort decisions. Here, the focus is carrying that agreed definition through Claude's reporting tools.

Build the Claude dashboard around a checkable metric

Start from the Dashboards template in the Artifacts tab, or choose Output → Dashboards in a chat. Use the approved data connection for this reporting question.

Give Claude a reporting specification before asking for a chart. This proposed brief is an example of your team's own metric definition, not a Claude API schema:

{
  "metric_id": "activation_7d_v1",
  "entity": "unique_account",
  "cohort_timestamp": "account_created_at",
  "window": "[created_at, created_at + 7 days)",
  "sequence": ["source_connected", "first_report_completed"],
  "timezone": "UTC",
  "data_cutoff": "2026-10-05T00:00:00Z",
  "excluded_account_types": ["staff", "test", "migration"]
}

Map those illustrative event names to your actual schema. Count each eligible account once. For the numerator, require a source connection on or after signup and a completed first report on or after that connection, both before the account's seven-day window closes. Keep accounts that never activate in the denominator.

The cohort boundaries are separate from the event window. The first cohort includes signups from 14 September at 00:00 UTC up to, but not including, 21 September at 00:00 UTC. The second covers the next seven days. Do not compare a fully observed cohort with this week's signups before their activation windows have finished.

Use an appropriately restricted data connection. Claude's connector documentation explains source permissions and action restrictions. A custom connector using shared credentials reaches what those credentials can access. For a reporting pilot, configure read access to the relevant views and enforce restrictions in the source or connector; a sentence in a prompt is not an access control.

Prompt for the dashboard

Build a product activation dashboard using the authorized reporting views.
First map activation_7d_v1 to the available tables and event fields.
If a required field or permission is missing, explain the gap before proceeding.
Compare signup cohorts [2026-09-14, 2026-09-21) and [2026-09-21, 2026-09-28), in UTC.
Use data available as of 2026-10-05T00:00:00Z.
Count unique eligible accounts and ordered activation within seven days.
Keep non-activating accounts in the denominator; apply the stated exclusions.
Show the cohort sizes, stage counts and activation rates together.
Expose the query, cohort boundaries, event window and refresh timestamp.
Report absolute counts, percentage-point change and relative rate change separately.
Flag missing, duplicated or late-arriving events and any schema assumptions.
Describe the observation without claiming what caused it.

The prompt describes desired work; it does not add undocumented product capabilities. If the connected system cannot reconstruct the requested historical data state, say so. An event-time filter alone does not reproduce what the database knew at a past cutoff.

Before moving to Motion, compare the returned counts with a trusted query or existing report using the same definitions. Inspect a few accounts near the time boundaries, one that never activates, a duplicate event and a report completed after day seven. Check that a join has not multiplied accounts and that exclusions apply consistently.

The dashboard exposes a query so a reviewer can inspect the calculation. It still needs that review. Once the numbers agree, record a result version, the metric version, cohort dates, data cutoff, refresh time and the person who approved the interpretation.

Turn the approved result into a short Motion update

Choose Output → Motion in a chat, or start from the Motion template in Artifacts. Give it the reviewed table and a focused brief. For the example, the audience is the product team, the finding is the decline in activation rate, and the next action is to inspect the onboarding steps and acquisition segments behind it.

A 30-second update is a useful editorial target for this example. It is not a documented maximum duration. A simple sequence keeps the denominator and qualification visible:

Prompt for the animation

Create a 30-second product-team update from the approved table and brief.
Use only these figures: cohorts 1,000 and 1,200; connected accounts 600 and 660;
activated accounts 400 and 420; activation rates 40% and 35%.
The finding is a five-percentage-point decline in seven-day activation.
Do not claim that any product change caused it.
Use the same scale for comparable charts and keep denominators visible.
Keep values and labels legible after the animation settles.
Show the cohort dates, UTC timezone, data cutoff and result version.
End with: inspect onboarding-step and acquisition-segment differences.
Use the provided brand assets and a calm pace suited to a team meeting.
Return an editable animation plus a plain-text explanation of every figure.
Prepare an MP4 for review; do not publish or distribute it.

Review the finished sequence, not just the script. Make sure the scene containing 420 activations also shows the 1,200-account denominator. A chart should not animate an increase in the count while presenting it as an improvement in the rate. Check cropped labels, rounding, color meaning and whether the final frame stays readable long enough.

Keep an equivalent text explanation alongside the video. If your production process adds speech or other meaningful audio, provide captions as well. The W3C's media accessibility guidance covers descriptions of visual information, captions and descriptive transcripts. Do not assume that an MP4 export provides those automatically.

Does the Motion video update when the dashboard refreshes?

Treat an exported MP4 as a dated snapshot. The launch demonstrates using the tools together, but the reviewed documentation does not establish an automatic link that regenerates exported videos whenever dashboard data changes.

A dashboard's latest refresh and the figures approved for a particular update can represent different moments. Keep both timestamps. The reviewed Dashboards guide shows refresh information but does not specify a universal refresh interval or latency guarantee.

For a recurring report, create a small release record: metric version, query reference, reviewed result table, cohort dates, data cutoff, review time and video filename. For example, activation_7d_2026-10-05_v1.mp4 identifies this illustrative update. If newly arrived events change the figures, advance the data cutoff to reflect their inclusion, preserve the original cohort windows, review the revised result and export a new version. Mark the earlier one as superseded wherever you control its distribution.

This also gives the team a useful audit trail: someone watching an older video can identify the data it described instead of assuming that every clip shows the latest state.

Who can see the dashboard, and who can watch the video?

The artifact-sharing rules distinguish access to an artifact from access to its connected data. A connected artifact uses the viewer's own connections. Missing source access can produce an error. Connected artifacts cannot use “Anyone with the link,” and connector-dependent parts do not work for invitees outside the organization.

Test a shared dashboard with the intended colleague's account. Check both that the page opens and that the permitted figures load. Also review any stored information in the artifact, because stored shared data has a different visibility boundary from a live connector call.

An exported video already contains rendered figures. Its viewers do not make the original warehouse query when they play it. Choose the approved audience before export and avoid including customer names or account-level details that the audience should not receive. Revoking a dashboard link does not recall copies of a downloaded MP4.

For an internal update, distribute the approved clip with its text summary and a link to the dashboard for colleagues who have source access. For an external update, prepare a separately approved version with suitable aggregated or anonymized information. Treat that as a publishing decision, not a shortcut around dashboard permissions.

When should the workflow move into your existing BI or product stack?

Keep a one-off question in the exploratory workflow when its scope and audience are clear. A repeated operational report needs maintained metric definitions, ownership and a dependable review process. Customer-facing analytics may also need a product-specific permission model, availability expectations and support.

A documented onward route already exists. Sigma's launch-partner guide describes sending a Claude dashboard into a Sigma workbook with its queries and filters. It requires a Sigma Build license and the relevant MCP OAuth permission; Sigma permissions apply there. Evaluate the actual handoff with your existing tools instead of assuming every integration has identical behavior.

If the goal is an interactive experience embedded in your own application, our generative UI architecture comparison covers that separate product decision. The Dashboards-and-Motion workflow here serves the reporting and explanation step.

A focused first pilot for a product team

Choose one already understood metric, one authorized data source and one audience. Have the metric owner review the dashboard and a second person compare the animation against the approved results.

1. Establish a baseline. Record how the team currently produces the update, including query work, checking, writing, revisions and distribution.
2. Run the example workflow. Use a complete reporting period, verify the calculation and prepare one short animation with its text equivalent.
3. Test a revision. Change one approved number in a test copy and check every dependent percentage, label and scene. Verify that the data cutoff stays visible.
4. Test the audience. Confirm what an intended dashboard viewer can access and what the exported clip actually discloses.
5. Decide whether to repeat it. Compare total preparation and review effort, number of corrections and whether viewers understand the finding and next action.

The combination is promising because exploration and explanation can share a clear set of inputs. A useful pilot proves that your team can preserve their meaning, revise the update and deliver it to the right audience with less friction. That is a more actionable success measure than the polish of the first animation.