Build a Claude.ai Data Dashboard

Turn a CSV or Excel file into a validated interactive dashboard in Claude.ai.

Introduction

30 Second Summary

Spreadsheets can hold the answer to an important question while still making that answer hard to see. A useful dashboard brings the signal forward without losing the trail back to the source data.

In this project, you will use Claude.ai to turn one approved CSV or XLSX file into an interactive Claude Artifact. You will make its metrics traceable to the source while keeping its sharing intentional.

What You'll Build

Picture selecting a category in a dashboard beside your Claude conversation and watching every KPI, chart, and insight update to explain that segment.

By the end of this project, you'll have:

A visible data profile that shows which measures, categories, dates, missing values, and duplicates shape the dashboard.
An interactive dashboard where one filter updates three KPI cards, two charts, and a concise insights panel before reset restores the complete view.
A defensible final artifact whose metric definitions reconcile to the source and whose access setting fits its intended audience.
An optional Secret Mission: add a comparison mode that places two segments side by side and calculates their differences.

Are there any prerequisites?

You need a Windows computer with a web browser plus a paid Claude account.

You also need permission to use one approved CSV or XLSX file within the 30MB per file limit.

Before We Start

Before any hands-on work, this step gives your dashboard a clear purpose. That purpose rests on one approved CSV or XLSX file, with a redacted working copy protecting anything the intended audience should not see.

Profile Your Data in Claude

Your dashboard purpose statement gives this project a clear destination. Claude needs access to the spreadsheet before it can produce an honest visualization.

Data profiling exposes the dataset's structure before any chart is built. It also reveals quality issues that could distort the dashboard.

In this step, get ready to:

Confirm Claude's file capability is enabled.
Attach your approved CSV or XLSX spreadsheet to a new Claude chat.
Generate a complete profile of the approved spreadsheet.

Enable file creation in Claude

Claude uses its file capability to inspect spreadsheets. The capability must be enabled before Claude can analyze your approved source.

Go to Claude.ai in your web browser.
Sign in with your paid Claude account.

You'll see Claude's chat workspace after your account opens.

Use Claude's account menu to open Settings.
Select Capabilities.

You'll see the capabilities available for your account.

Confirm Create and edit files is on.
Confirm Code execution and file creation is on if your account shows the older label.

That setup hurdle is cleared. Claude can now inspect the spreadsheet you attach.

Can't find the file setting?

Confirm you signed in with the paid Claude account you prepared for this project.
Refresh Settings before checking Capabilities again.

Help me find the file capability in Claude.

Attach your approved spreadsheet

Claude analyzes the source attached to the conversation. The approved spreadsheet from earlier is already the right working copy for your intended audience.

Uploading business data can feel sensitive. Your approved or redacted working copy keeps the analysis limited to information that audience can see.

Start a new chat in Claude.

You'll see an empty conversation with a message box ready for your request.

Select + in the lower-left corner of the chat box.
Choose Add files or photos.

You'll see the file picker for your Windows computer.

Select the approved CSV or XLSX file from your computer.
Click Open.

Good. You'll see the spreadsheet name attached to the new chat.

Spreadsheet not attaching?

Confirm the source uses the CSV or XLSX format.
Confirm the file is no larger than 30 MB.
Try attaching the approved working copy again from the new chat.

Help me troubleshoot my Claude file upload.

Request the source-data profile

The profiling request separates inspection from dashboard creation. This gives you evidence about the source before Claude chooses any metrics.

Paste this profiling prompt into the message box:

Inspect the attached file before building anything. Report the file or sheet used, row count, column names, inferred data type for each column, missing values, duplicate rows, valid date fields, numeric measures, and useful category fields. Flag ambiguous columns or values. Do not build the dashboard yet.

What does this prompt check?

The source check establishes which file or sheet Claude inspected.
The structure check maps each column to an inferred data type.
The quality check exposes missing values plus duplicate rows.
The field review identifies candidates for dates, measures, or categories.
The ambiguity check surfaces meanings that need clarification before metric selection.

Before you send it, which columns do you expect Claude to identify as useful measures or categories?

Send the profiling prompt.

What should I see?

You'll see a visible table or structured summary of the attached source.

The file or sheet Claude used.
The source row count.
Column names with an inferred data type for each column.
Missing values plus the duplicate-row count.
Valid date fields, numeric measures, and useful category fields.
Any ambiguous columns or values that need clarification.

Confirm the response identifies the file or sheet used.
Confirm the response states the row count.
Confirm every column has an inferred data type.
Confirm the quality summary covers missing values plus duplicate rows.
Confirm the response identifies candidate field roles for the dashboard.
Confirm the response flags ambiguous columns or values.

Profile missing a required detail?

Resend the same profiling prompt if Claude skipped part of the requested profile.
Confirm the approved spreadsheet is still attached to the current conversation.

Help me complete Claude's data profile.

That is the data groundwork done. Claude has exposed the spreadsheet's structure before any dashboard calculations begin.

Your approved source now has a visible profile you can defend. Next, you'll use those findings to build the first dashboard view.

Build Your First Dashboard

Your approved source file is attached. Claude.ai has mapped its structure through a visible data profile.

Now you can turn that evidence into a custom Claude Artifact for your chosen audience. Can a polished overview answer the focused questions behind the decision you defined?

In this step, get ready to:

Resolve any ambiguous field meanings before building the dashboard.
Build a source-based dashboard with three KPI cards, two charts, visible assumptions, and concise insights.
Test whether the overview can isolate a useful segment of the source data.

Confirm your field choices

Charts become trustworthy when every measure has a clear meaning. The data profile gives you the evidence needed to catch ambiguous fields before they shape the dashboard.

Return to the data profile in the Claude chat from earlier.
Review the valid date fields against the decision in your dashboard purpose statement.
Review the numeric measures against the same decision.
Review the useful category fields against the same decision.
If Claude flagged an ambiguous field, type its meaning into the chat message box.
Send each required field clarification before continuing.

The conversation now contains any clarification Claude needs. Every candidate field has a meaning you can defend.

Build the first artifact

The build request constrains the artifact to source columns. Visible assumptions keep the resulting overview traceable to the spreadsheet.

Prepare Claude to build the first dashboard by copying this prompt:

Using only the attached data, build a custom artifact named Data Insights Dashboard. Include three decision-useful KPI cards; one time-trend chart if a valid date field exists, otherwise a numeric distribution; one category comparison chart; and a short insights panel. Use only source columns, state assumptions visibly, do not invent values, and do not add filters yet. Make it readable on a laptop.

How does this prompt shape the artifact?

The artifact name makes the dashboard easy to identify beside the conversation.
The three KPI cards create a quick summary for the decision you defined.
The first chart adapts to the fields Claude found during profiling.
The source-column rule prevents Claude from filling gaps with invented values.
The visible assumptions make each analytical choice easier to review.

Before you send it, which three measures do you expect Claude to choose for the headline cards?

Paste the prompt into the message box in the Claude chat from earlier.
Send the prompt.

Claude processes the attached data. You will see the custom artifact open beside the conversation.

Confirm the artifact title reads Data Insights Dashboard.
Count three decision-useful KPI cards.
Confirm the first chart matches the profile. It should show a time trend when a valid date field exists. It should show a numeric distribution when no valid date field exists.
Confirm the second chart compares values across a source category.
Read the short insights panel.
Locate the visible assumptions inside the artifact.
Check that every displayed field appears in the data profile.
Check that the labels remain readable within your laptop window.

That is your first visible win. Your spreadsheet now has a decision-focused dashboard that you can inspect beside the conversation.

Artifact not opening?

Confirm you are still using the Claude chat that contains the attached source file.
Resend the unchanged build prompt if Claude returned only a written description.
Ask Claude to build the result as a custom artifact if it did not open beside the conversation.

Help me troubleshoot why Claude is not opening my spreadsheet dashboard as a custom artifact.

Test a focused question

An overview earns its place by supporting real questions. Test whether this first version can move from the overall picture to one specific slice of the data.

Before you test it, do you think this version can isolate one category, period, or numeric range?

Choose one category, period, or numeric range that matters to the dashboard decision.
Scan the artifact for a control that can isolate your chosen slice.
Attempt to show only that slice in the dashboard.

The overall KPI cards, charts, and insights remain visible. You cannot isolate your chosen slice because no interactive filter control is present.

What did this test prove?

The dashboard provides a useful source-based overview. Its static view cannot answer a segment-specific question.

This shortfall is intentional. You now have a concrete behavior to improve while preserving the current metric definitions.

Seeing a filter anyway?

Ask Claude to remove every filter control from this first version.
Repeat the focused-question test after Claude updates the artifact.

Help me remove the unexpected filter without changing the dashboard metrics.

Your first dashboard delivers a readable overview from the approved source. Next, you will add one synchronized filter so the same view can answer a focused question.

Add a Synchronized Filter

Your Claude Artifact already turns the approved spreadsheet into a useful overview. Its source-based metrics give you a strong starting point.

A static overview cannot answer a focused question about one segment. In this step, you will add a synchronized filter that applies one selection across the complete dashboard.

In this step, get ready to:

Choose a decision-useful field from the data profile.
Add one synchronized filter to Data Insights Dashboard.
Verify two filter selections plus the reset state.

Choose the filter field

The filter field determines which focused questions the dashboard can answer. A clear category usually creates the most useful control because each value represents a distinct segment.

Return to the visible data profile in the current Claude conversation.
Review the useful category fields listed in the profile.
Choose a category field with a manageable number of values when one is available.
Choose a valid date field for a range filter when no suitable category exists.
Choose a numeric measure for a range filter when neither option fits.

What makes a useful filter?

A useful filter divides the source data into groups that matter to the dashboard audience. The selected field should support the decision in your dashboard purpose statement.

A field with too many unique values creates a cumbersome control. The data profile helps you choose a field whose values are clear enough to test.

Add the filter to the artifact

The dashboard needs one shared filtered dataset behind every component. This keeps its KPI cards, charts, and insights aligned with the same selection.

Paste this prompt into the message box in the current Claude conversation:

Add one filter using the most decision-useful existing field in the source data. Prefer a categorical field with a manageable number of values; otherwise use a date or numeric range. Make the filter update every KPI, chart, and insight. Add a clear reset control and show the active filter state. Do not change the metric definitions.

What does this prompt do?

The source-field constraint keeps the filter grounded in the attached spreadsheet.
The shared scope keeps every dashboard component synchronized.
The active state shows viewers which subset they are seeing.
The reset control provides a direct path back to the unfiltered view.
The final constraint protects the existing metric definitions.

Send the prompt.
Wait for Claude to finish updating Data Insights Dashboard.
Inspect the updated artifact for one filter control.
Locate the visible active-filter state.
Locate the reset control.
Compare the three KPI labels with the original dashboard.
Confirm each KPI keeps its existing definition.

You should now see one filter built from an existing source field. The same three KPI definitions should remain in place.

That control closes the gap you found in the static version. The next check proves whether its reach covers the complete dashboard.

Filter missing from the artifact?

Resend the same prompt in the conversation containing the approved source file.
Keep Data Insights Dashboard active beside the conversation while Claude applies the update.

If the artifact still remains static, ask for help diagnosing why the synchronized filter was not added.

Verify synchronized filtering

A visible control only proves that the filter exists. Two different selections prove that the complete dashboard responds to it.

Before you choose the first value, which dashboard elements do you expect to change together?

Choose one KPI card to track across every filter state.
Record its label here: KPI label you are tracking.
Select the first filter value.
Confirm the active-filter state shows that value.
Inspect all three KPI cards.
Inspect both charts.
Read the insights panel.

Every KPI card should now reflect the selected subset. Both charts should show the same filtered scope.

The insights panel should describe that subset. The active-filter state should match your selection.

Record the first selection here: first filter value.
Record the tracked KPI value here: KPI value for first selection.

Dashboard components out of sync?

Check whether the active-filter state matches the value you selected.
Resend the synchronized-filter prompt if any KPI card remains unchanged.
Resend the synchronized-filter prompt if either chart ignores the selected value.

For help tracing an inconsistent component, ask why one part of the dashboard is ignoring the shared filter.

The first selection proves that the filter can narrow the dashboard. A different value tests whether the synchronization holds across another subset.

Select a different filter value.
Confirm the active-filter state shows the second value.
Inspect all three KPI cards.
Inspect both charts.
Read the insights panel.

You should see a new set of KPI values for the second subset. The charts should also move to that subset.

The insights should now describe the second selection. This confirms the control affects more than the headline cards.

Record the second selection here: second filter value.
Record the tracked KPI value here: KPI value for second selection.

Before you use the reset control, do you expect any filtered value to remain in the dashboard?

Use the reset control.
Inspect the active-filter state.
Compare the tracked KPI with the original unfiltered view.
Inspect both charts for the complete data range.
Read the insights panel for the unfiltered overview.

You should see the unfiltered KPI totals again. Both charts should return to their complete source-based views.

The insights panel should return to the overall dataset. The active-filter state should show that no segment remains selected.

Record the reset KPI value here: KPI value after reset.

Reset not restoring the dashboard?

Confirm that the filter selection cleared after you used the reset control.
Compare each dashboard component with its original unfiltered state.
Resend the synchronized-filter prompt if any component stays filtered.

If the reset behavior remains inconsistent, ask for help restoring every component to the same unfiltered dataset.

You have turned a static overview into a dashboard that can answer focused questions. Every KPI, chart, and insight now follows the same filter state.

Your synchronized filter now works across the complete dashboard. Next, you will audit the displayed numbers against the source file.

Reconcile Your Metrics

Your filter already keeps every KPI, chart, and insight in sync. The next job is to prove that each headline number agrees with the source file.

A polished dashboard can hide plausible calculations that use the wrong rows or aggregation. Metric reconciliation gives Claude a traceable check for every displayed KPI.

In this step, get ready to:

Audit every displayed KPI against the attached source file.
Correct mismatches while documenting metric definitions and missing-data treatment.
Rerun the reconciliation for unfiltered totals and a tested filtered state.

Audit every KPI

The first audit separates trustworthy calculations from numbers that only look reasonable. Claude will trace every KPI back to its source columns before changing the dashboard.

In the Claude conversation from earlier, paste this audit prompt into the message box:

Audit the current dashboard against the attached source file. Return a reconciliation table for every KPI with these columns: displayed label, exact source column or columns, calculation, included rows, excluded rows, recomputed value, displayed value, and match status. Explain how blanks, duplicates, and invalid dates were handled. Do not change the dashboard yet.

What does this audit check?

The exact source column or columns entry makes every KPI traceable to the spreadsheet.
The included rows and excluded rows entries expose the data used by each calculation.
The recomputed value and displayed value entries reveal whether the dashboard agrees with a fresh calculation.
The missing-data explanation shows how blanks, duplicates, and invalid dates affect the results.

Send the prompt.
Review every row in the returned reconciliation table.
Identify each row with a nonmatching or ambiguous status.

Claude returns a structured table for every KPI without changing the dashboard. You now have a visible baseline for finding calculation gaps.

Missing details from the audit?

Ask Claude to revise the table if any requested column is missing.
Confirm that the approved spreadsheet remains attached to this conversation.
Use this if the response still lacks a complete audit: Help me get a complete KPI reconciliation table from my attached spreadsheet.

Correct mismatches and document definitions

A nonmatching row points to a calculation that needs correction. An ambiguous row points to a definition that needs to become explicit.

Review every nonmatching or ambiguous row in the audit.
Use the data profile from earlier to confirm each affected source column.
Use your agreed metric definitions to resolve each ambiguity.
Paste this correction prompt into the same message box:

Correct every mismatch using the source file and the agreed definitions. Add concise metric definitions and a missing-data note inside the dashboard. Preserve the working filter and reset behavior. Do not introduce any field or value that is absent from the source.

What does this correction protect?

The source-file requirement keeps each correction grounded in the attached spreadsheet.
Visible metric definitions explain what each KPI measures.
The missing-data note documents how the dashboard treats incomplete or unusable rows.
The preservation instruction protects the synchronized filter and reset behavior.

Send the prompt.
Switch back to the Claude Artifact from earlier.
Confirm that every KPI now has a concise visible definition.
Confirm that the missing-data note covers blanks, duplicates, and invalid dates.
Select one filter value that you tested earlier.
Use the dashboard's reset control.

The filter should still update every KPI, chart, and insight. The reset control should restore the unfiltered totals.

That closes the risky part of the audit. Your dashboard now explains its metrics without sacrificing the interaction you already tested.

Did the filter behavior change?

Tell Claude which KPI, chart, or insight stopped responding to the filter.
Ask Claude to restore the previous synchronized behavior without changing the reconciled definitions.
Use this if a correction disrupted the dashboard: Help me restore my synchronized filter without undoing the reconciled KPI calculations.

Rerun the reconciliation

A correction becomes trustworthy when a fresh audit reaches the same answer as the dashboard. You will check the unfiltered totals first.

Use the dashboard's reset control to restore the unfiltered totals.

You should see the original unfiltered state across every KPI, chart, and insight.

Before you rerun the audit, do you expect every recomputed value to match its displayed value?

Paste the same audit prompt into the Claude conversation:

Audit the current dashboard against the attached source file. Return a reconciliation table for every KPI with these columns: displayed label, exact source column or columns, calculation, included rows, excluded rows, recomputed value, displayed value, and match status. Explain how blanks, duplicates, and invalid dates were handled. Do not change the dashboard yet.

What does this rerun prove?

This rerun uses the same audit structure against the corrected dashboard. Matching values show that the dashboard and source-based calculation now agree.

Send the prompt.
Compare each recomputed value with its displayed value.
Compare one unfiltered KPI with the reset value in your filter test record.

Every KPI row should now show matching recomputed and displayed values. The checked unfiltered KPI should also agree with the value you recorded earlier.

Still seeing a mismatch?

Check whether the mismatched row uses the agreed source column.
Check whether its included and excluded rows follow the documented missing-data treatment.
Use this to investigate the remaining gap: Help me trace why this KPI still has different recomputed and displayed values.

The unfiltered check proves the dashboard's overall totals. One filtered check now confirms that the same definitions hold when the viewer isolates a segment.

Select one filter value from your earlier filter test record.

The active filter state should become visible. Every KPI, chart, and insight should update together.

Before you run the final audit, do you expect the filtered KPI to match both the fresh calculation and your recorded value?

Paste the same audit prompt into the Claude conversation:

Audit the current dashboard against the attached source file. Return a reconciliation table for every KPI with these columns: displayed label, exact source column or columns, calculation, included rows, excluded rows, recomputed value, displayed value, and match status. Explain how blanks, duplicates, and invalid dates were handled. Do not change the dashboard yet.

Why test a filtered state?

The filtered audit checks whether every KPI uses the same subset of source rows. It also confirms that the corrected definitions survive interaction with the dashboard.

Send the prompt.
Review the match status for every filtered KPI row.
Compare the selected KPI with the value in your filter test record.
Use the dashboard's reset control.

Every filtered KPI row should match. The selected KPI should agree with your recorded value, while reset should restore the reconciled unfiltered totals.

Does the recorded value differ?

Confirm that you selected the same filter value used in your earlier test.
Confirm that the active filter state shows that value.
Use this to compare both states carefully: Help me reconcile my recorded filtered KPI value with the current dashboard and source file.

Your dashboard now has numbers you can defend with source columns, formulas, row treatment, and matching checks. Next, you will polish the presentation while protecting the data and its access setting.

Polish and Control Access

Your reconciliation proves that every displayed KPI matches the source. The validated dashboard in Claude Artifacts is ready for its final presentation pass.

Correct numbers still fall short when viewers struggle to scan the dashboard. Broad access or unnecessary row-level data can expose more than the intended audience needs.

In this step, get ready to:

Create a polished duplicate of the validated dashboard.
Remove unnecessary personal or row-level details from the duplicate.
Verify the finished artifact with an intentional access setting.

Create a polished duplicate

A duplicate protects the reconciled original while you refine the presentation. The visual changes should improve readability without altering any validated calculation.

Switch back to the Data Insights Dashboard artifact beside the current conversation.
Open the artifact's title menu.
Choose Duplicate.

Good call. You'll see a separate artifact where you can refine the presentation while the validated original stays untouched.

Paste this polishing prompt into the duplicated artifact's chat box:

Polish this dashboard without changing any calculations. Use a clear visual hierarchy, descriptive chart titles, readable labels, a restrained color palette, and a visible active-filter state. Avoid using color as the only signal. Keep all metric definitions and the missing-data note.

What does this prompt protect?

The opening sentence freezes the reconciled calculations while Claude changes the presentation.
The visual hierarchy gives viewers a clear reading order.
The visible active-filter state makes the selected view easy to identify.
The final sentence preserves the definitions that make each metric traceable.

Send the prompt.
Read each chart title on your laptop screen.
Read each KPI label.
Read each chart label.
Hover over every chart tooltip.
Confirm that the palette communicates meaning without relying on color alone.
Confirm that the metric definitions remain visible.
Confirm that the missing-data note remains visible.

You should see a clear reading order across the dashboard. Every title, label, tooltip, insight, and table should remain readable on your laptop screen.

Did the polish change a calculation?

Compare the reset-state KPI values with the matching rows in your rerun reconciliation table.
Ask Claude to restore the reconciled definition for any KPI whose value changed.

If the cause is unclear, help me identify why the visual polish changed a reconciled KPI.

Remove unnecessary data exposure

A privacy review checks every surface where source details might appear. The finished dashboard should retain only the aggregates needed for its stated decision.

Inspect every KPI card for personal identifiers.
Read every chart label for unnecessary row-level details.
Hover over every tooltip to check its displayed fields.
Review each insight for identifying details.
Review each table for unnecessary source rows.
Continue to the next substep if every surface contains only the required aggregates.
Remove any exposed details by sending this prompt if your review finds them:

Remove any personal identifiers or unnecessary row-level data from this dashboard. Retain only the aggregates required for the dashboard's stated decision. Preserve the reconciled calculations, working filter, active-filter state, reset behavior, metric definitions, and missing-data note.

What does this prompt preserve?

The dashboard keeps only the aggregates required for its stated decision.
The reconciled KPI calculations remain unchanged.
The existing filter behavior remains available.
The metric definitions keep the numbers traceable.

Reinspect every surface that Claude changed.
Confirm that each changed surface now shows only required aggregates.

Your polished artifact should now contain no unnecessary personal identifiers. Any row-level detail that remains should be essential to the dashboard's stated decision.

Still seeing exposed details?

Name the exact card, chart, tooltip, insight, or table that still exposes the detail. Ask Claude to replace that detail with the corresponding aggregate while preserving the reconciled metric definition.

For a second review, help me find personal identifiers or unnecessary row-level data still exposed in my dashboard.

Reopen, test, and set access

Reopening the artifact proves that the finished version was saved outside the immediate conversation view. Repeating the filter test confirms that visual and privacy changes preserved its behavior.

Select Artifacts in the Claude sidebar.
Open the polished duplicate from the artifact list.

You'll see the finished dashboard reopen with its polished layout, metric definitions, missing-data note, filter, and reset control.

Before you test it, do you expect the polished copy to reproduce the KPI values in your filter test record?

Select the first filter value from your earlier test record.
Compare its recorded KPI value with the reopened dashboard.
Select the second filter value from your earlier test record.
Compare its recorded KPI value with the reopened dashboard.
Choose the dashboard's reset control.

Each selection should update the complete dashboard while the active-filter state remains visible. Reset should restore the unfiltered totals from your record.

Does the filter test fail?

Return to the duplicate's conversation if a recorded KPI value no longer matches.
Ask Claude to restore the reconciled calculation while preserving the polished design.
Check that reset clears the active-filter state if the unfiltered totals do not return.

If the failure continues, help me troubleshoot the polished dashboard's filter and reset behavior.

You stay in control here because Claude keeps artifacts private until you choose to share them. Your final setting should expose the dashboard only to its intended audience.

Leave the artifact private if nobody else needs access.
Click Share if the privacy-reviewed dashboard is safe to distribute.
Choose the least permissive option under Who has access that reaches the intended audience.
Confirm that the displayed access setting matches your dashboard purpose statement.

You'll see either the unchanged private state or the audience you intentionally selected under Who has access.

Can't find the access choice you expected?

Available access choices depend on your Claude plan. Use the least permissive choice shown under Who has access that still reaches the intended audience.

If you are unsure which choice fits, help me choose the safest available artifact access setting.

That completes the final quality gate. Your dashboard is readable, tested, privacy-reviewed, and available only as widely as needed.

Secret mission

Compare Two Segments

Your validated dashboard can focus on one segment at a time. In this mission, you will add two selectors for side-by-side KPI comparison. The completed mode also compares the chosen segments in one chart. Reset still restores the original dashboard.

Clean Up Your Resources

Choose whether to keep your Claude.ai work available, restrict access while you pause, or clean up the conversation history. There are no new metered cloud resources to charge you after you stop because this work uses your existing paid Claude plan.

Resources you used:

The Claude.ai conversation with your approved source spreadsheet plus the full profiling through comparison audit trail.
The original Data Insights Dashboard Claude Artifact with its reconciled single-filter view.
The polished Data Insights Dashboard duplicate with its privacy review plus intentional access setting.
The comparison-enabled dashboard duplicate with its tested side-by-side segment mode.

Keep everything running

No action needed. Choose this option if you plan to keep using the dashboards or their audit history.

Keep the Claude.ai conversation when you want to revisit the source profile or reconciliation history.
Leave the original Data Insights Dashboard saved in the Artifacts tab.
Leave the polished dashboard duplicate saved in the Artifacts tab.
Leave the comparison-enabled dashboard saved in the Artifacts tab.
Retain each current access setting only while it still matches the intended audience.

Pause - I'll come back to this later

This option keeps your conversation and artifacts available for later. It limits who can open each artifact while you are away.

Return to the comparison-enabled dashboard from the Artifacts tab.
Click Share in the artifact header.
Choose Only you under Who has access on Pro, Max, or Ultra.
Choose Only people invited under Who has access on Team or Enterprise.
Remove each individual invitee when nobody else needs access.
Repeat these access changes for the polished dashboard duplicate.
Repeat these access changes for the original Data Insights Dashboard.

Delete - I don't want to use this again

This option revokes access to each dashboard before deleting the source conversation. The final confirmation removes the conversation from your history immediately.

Artifact access comes first because the conversation is the final cleanup target.

Return to the comparison-enabled dashboard from the Artifacts tab.
Click Share in the artifact header.
Choose Only you under Who has access on Pro, Max, or Ultra.
Choose Only people invited under Who has access on Team or Enterprise.
Remove every individual invitee.
Repeat these access changes for the polished dashboard duplicate.
Repeat these access changes for the original Data Insights Dashboard.

The source conversation holds the attached spreadsheet plus your audit history. Deleting it closes the documented cleanup path for that project history.

Return to the source conversation in the left sidebar or Chats and tasks.
Select ⋮ for the conversation.
Choose Delete.
Confirm Delete.

You have closed the loop on the sensitive part of this project. You'll see the conversation disappear from history immediately.

Claude schedules its deletion from back-end systems within 30 days.

Nice Work!

You did it! You turned an approved spreadsheet into a trustworthy interactive dashboard with Claude Artifacts.

You've learned how to:

Used data profiling to map the spreadsheet's structure. Flagged data-quality issues before choosing metrics or charts.
Built a custom dashboard that presents source-based KPIs through a clear visual hierarchy. Added interactive filtering across the complete dashboard. Made the reset control restore the unfiltered view.
Completed metric reconciliation against the source spreadsheet. Added visible metric definitions. Documented missing-data treatment. Completed a privacy review before choosing an intentional access setting.
Secret Mission: Added a two-segment comparison mode using an existing reconciled KPI. Displayed the absolute difference. Showed a percentage difference only when the denominator was nonzero. Preserved the original single-filter mode through reset.

Ready to quiz yourself?