> AI agents: this is one page from Mammoth Analytics documentation. The index of all pages as Markdown is https://docs.mammoth.io/llms.txt. Append `.md` to any docs URL, or send `Accept: text/markdown`, to get Markdown.

# View Insights

> View Insights uses AI to turn your View's column statistics into 3-7 plain-language observations. Learn how to generate them, when to regenerate, and limits.

View Insights turns the statistical profile of your data into short business observations. You click a button and Mammoth's AI writes the most significant observations in plain language. Use it to orient yourself in an unfamiliar dataset, prepare talking points for a briefing, or find areas to investigate further.

![Animation of the cursor clicking View Insights in the toolbar above the data grid](https://docs.mammoth.io/api/v1/images/20260421_170001_View_insights1.gif)

## What View Insights are

View Insights generates 3-7 business-relevant observations about your data. The AI works from column statistics for the View, such as distributions, date ranges, concentrations, and gaps, and surfaces the most significant findings in natural language.

### How it works

When you click the **View Insights** button in the action bar, Mammoth collects column statistics for the View and asks an AI model to turn them into business observations. The AI prioritizes:

1. **Revenue risks**: Concentration, declining trends, and underperforming segments
2. **Growth opportunities**: High-value segments and untapped patterns
3. **Decision blockers**: Missing data that prevents segmentation, forecasting, or targeting

![Menu opened from the 6 Key Insights button, listing six one-sentence insights about transactions, revenue and stores](https://docs.mammoth.io/api/v1/images/20260421_170613_Howitworks.jpg)

### What you get

The AI generates 3-7 insights (more when several patterns or risks are present). Each is one short sentence with key numbers and terms in bold. Insights may include:

- **Revenue concentration**: "67% of revenue is concentrated in Q3-Q4 period, with significant seasonal variation pattern detected"
- **Regional disparities**: "East region accounts for 45% of sales volume but only 32% of revenue, suggesting lower average order values"
- **Temporal patterns**: "Order processing time increased 23% in recent data compared to historical average"
- **Data gaps**: "Customer contact details are missing for 15% of records, blocking personalized outreach"
- **Risk signals**: "Shipments span from near-zero to 20.99K kg suggesting extreme order size variance that may indicate a fulfillment concentration risk"

Insights are business-oriented—the AI provides observations and context to help you understand what the data means, not just raw statistics. They are not instructions to change your data—that judgment remains with you.

## Key distinctions

### Insights versus the Metrics panel

The two features serve different purposes:

| Feature | Key Insights | Metrics panel |
| --- | --- | --- |
| **Purpose** | AI-generated observations | User-defined calculations |
| **Trigger** | Manual click | Open from the action bar |
| **Output** | 3-7 business patterns | Custom Count, Sum, Average, etc. |
| **Content** | "What AI noticed" | "What you want to measure" |
| **Language** | Business observations | Numbers and formulas |

### Insights versus the Data Quality Report

Both features analyze your data, but with different focus:

| Aspect | View Insights | Data Quality Report |
| --- | --- | --- |
| **Framework** | AI pattern discovery | DAMA quality standards |
| **Purpose** | Business intelligence | Quality assessment |
| **Output** | Business observations | Issues + recommended pipeline tasks |
| **Result** | "Here's what's interesting" | "Here's what needs fixing" |

> **Tip**: Use both together. The Data Quality Report identifies issues to address, and View Insights reveals business patterns to explore.

## Generate insights

### Prerequisites

- The View must contain data
- No specific column types required
- Works with raw uploads or transformed data
- The pipeline must not be running. The button is unavailable in preview or edit mode and in View-only workspaces

### Steps

**Step 1—Open your View**

Navigate to any View containing the data you want to analyze. This can be freshly uploaded data or a View with pipeline transformations already applied.

**Step 2—Find the View Insights button**

Look for the **View Insights** button on the right side of the action bar, to the left of the Export button. The button appears only when View Insights is enabled for your workspace. Before you generate insights, the button reads "View Insights." Once insights have been generated, the label updates to show the count—for example, "6 Key Insights."

**Step 3—Click the button**

Click the button. This starts the AI analysis—insights do not generate automatically when you open a View.

**Step 4—Wait for the analysis**

The button changes to show "Analyzing Insights" and "This will only take a moment."

![Toolbar with an Analyzing Insights spinner in place of the View Insights button](https://docs.mammoth.io/api/v1/images/20260421_174825_Toptoolbarshowing_AnalyzingInsights1.jpg)

**Step 5—Review the insights**

Click the button again to open the list of 3-7 insights below it. The list scrolls when it is long. Each insight is a single sentence that includes:

- Key terms and numbers in bold: highlights the primary data element or business concept
- Business framing: numbers presented with business implications where relevant

![List of five key insights, such as negative quantity values and a six-day transaction window](https://docs.mammoth.io/api/v1/images/20260421_175229_View_insights1.jpg)

### Use insights effectively

**For executive communication:** Copy insights into presentations or briefs. The AI uses business-friendly language with specific values and comparisons you can use as talking points.

**For pipeline development:** Use discovered patterns to guide transformation tasks. If the AI flags inconsistent formatting or negative values, investigate those areas before building your pipeline.

**For dashboard creation:** Convert significant patterns into dashboard KPIs. Revenue concentration insights become revenue trend charts; customer behavior patterns become segmentation views.

**For further investigation:** Use Explore Cards to investigate specific patterns the AI surfaced. Apply filters to analyze the subsets or segments mentioned in insights.

## When to regenerate insights

### Regeneration triggers

View Insights reflects the View's data at the time it was generated. When the View's data changes, the saved insights are cleared and the button returns to "View Insights". Regenerate when:

- **After pipeline changes**: insights are not updated automatically when you add or run tasks
- **New data added**: when the Dataset receives fresh batches
- **Schema changes**: when columns are added or data types change

### How to regenerate

Click the View Insights button after it has returned to its initial state. Previous insights are replaced, not saved alongside. If you need to keep the current insights for comparison, copy the text or take a screenshot before regenerating.

Insights are saved with the View until its data changes, so reopening the View does not generate them again.

## Real-world use cases

### Use case 1: Executive briefing preparation

**Situation**: A manager needs to brief the CEO on Q4 customer engagement before a 2 PM meeting.

**Workflow**:

1. Opens customer_engagement View with 280,000 interaction records
2. Clicks the View Insights button
3. Receives business-critical patterns:
   - "Email engagement rates dropped 34% in November compared to September baseline"
   - "Mobile app users show 3.2x higher retention rates than web-only users"
   - "Premium tier customers interact 5x more frequently than basic tier"
4. Copies insights directly into executive brief

**Outcome**: Ready-made talking points. CEO immediately asks for mobile app investment analysis based on the 3.2x retention insight.

### Use case 2: Initial data exploration

**Situation**: A business analyst receives a new quarterly sales dataset from 3 regional databases—150,000 rows across 23 columns—with pressure to deliver insights for Friday's executive meeting.

**Workflow**:

1. Uploads consolidated sales data to new View
2. Generates Key Insights before building any transformations
3. AI surfaces critical patterns:
   - "West region contributes 62% of total revenue despite 34% of transaction volume"
   - "Average order value declined 18% in March compared to February"
   - "Top 15 customers account for 41% of quarterly revenue"
4. Uses insights to guide dashboard creation and pipeline development

**Outcome**: Five ready-made talking points for Friday's meeting, plus strategic focus areas identified for further investigation.

### Use case 3: Quality validation before production

**Situation**: A technical user completed an transformation pipeline of 8 tasks and needs quality validation before publishing a dashboard to 40 stakeholders.

**Workflow**:

1. Opens final transformed View (post-pipeline)
2. Runs Data Quality Report (shows "Good" 82% score)
3. Generates Key Insights for additional validation
4. AI reveals unexpected pattern: "Transaction amounts show unusual concentration at exactly $1,000 intervals (23% of records)"
5. Investigation reveals a data truncation issue in the source system
6. Fixes pipeline before dashboard publication

**Outcome**: Caught a production issue before 40 stakeholders saw bad data. The combination of the Data Quality Report and Key Insights caught a problem that statistical summaries alone had missed.

### Use case 4: Trend discovery for budget planning

**Situation**: A business analyst analyzes a 2-year sales history (2.3M rows) for budget planning with no specific hypothesis.

**Workflow**:

1. Opens sales_history View with historical data
2. Generates Key Insights
3. AI reveals unexpected correlations:
   - "Sales to repeat customers increased 156% year-over-year while new customer acquisition declined 23%"
   - "Average days between purchases decreased from 47 to 31 days in Q3-Q4"
   - "Customers purchasing Product Line C show 4.1x higher repeat purchase rates"

**Outcome**: Strategic insight about the business shifting from an acquisition to a retention model. Revised budget model based on repeat customer trajectory.

### Use case 5: Vendor data validation

**Situation**: A business analyst just uploaded vendor-supplied inventory data and wants to validate before investing time in pipeline development.

**Workflow**:

1. Opens inventory_raw View immediately after upload
2. Generates Key Insights before building transformations
3. AI surfaces critical risk patterns:
   - "A single SKU holds 12,500 units while most products sit at 125 creating severe inventory concentration risk"
   - "Expiry dates fall into only 2 distinct windows meaning a mass write-off event could hit with almost no warning"
   - "Last stock updates span 1,036 days with only 13 distinct dates leaving most SKUs with unknown freshness"
4. Contacts vendor with specific concerns before building any pipeline

**Outcome**: Critical inventory risks identified before a single pipeline task was built.

![List of five key insights about inventory, such as one SKU holding 12,500 units](https://docs.mammoth.io/api/v1/images/20260421_175249_Usecase5-viewinsights1.jpg)

## Best practices

### Before generating insights

**Prepare the View first.** Insights describe the View's data after its pipeline. To focus on a subset (such as Q4 data only), add a Conditional Filter task to the pipeline before generating. Filters applied in Explore Cards do not change insights.

**Run the Data Quality Report first.** Use both tools together—Data Quality Report identifies issues to fix, View Insights reveals business patterns to explore.

**Review Explore Cards.** Understanding basic distributions helps you interpret AI insights with proper context.

### While reviewing insights

**Look for the unexpected.** AI often surfaces patterns that humans miss in manual analysis. Pay attention to surprising correlations or concentrations.

**Cross-reference with business knowledge.** AI detects statistical patterns but lacks your business context. Some insights may not be relevant to your specific situation.

**Note specific numbers.** The AI provides exact values and comparisons suitable for stakeholder communication—copy these directly for presentations.

### After reviewing insights

**Document surprising patterns.** Share unexpected insights with domain experts who can provide business context for statistical observations.

**Prioritize further investigation.** Use insights to decide which areas warrant deeper analysis with Explore Cards or the Metrics panel.

**Create dashboard KPIs.** Convert significant patterns into ongoing monitoring through dashboard creation.

> **Tip**: Generate insights both before and after major pipeline changes. The "before" insights help prioritize which tasks to build. The "after" insights show whether the tasks changed the business patterns as intended.

## Common questions

### "Why do I need to click a button? why isn't it automatic?"

The manual trigger gives you control over analysis timing and prevents unnecessary processing. You choose when to run the analysis based on your workflow—after data changes, before stakeholder meetings, or when exploring new datasets.

### "The button shows a different number each time—is that normal?"

Yes. The button initially reads "View Insights" before any insights have been generated. After you click it and the AI completes analysis, the label updates to show the actual count for that session—for example, "6 Key Insights" or "7 Key Insights." The number reflects what the AI found in your data and cannot be set manually. It can differ between generations on the same dataset.

### "Can I customize which insights appear?"

No. The AI determines which patterns are statistically significant, so you can't choose which insights appear.

### "What if I disagree with an insight?"

That's expected. You bring business context that the AI lacks. Some insights might not be relevant to your specific situation, but they still represent valid statistical patterns. Use your expertise to determine which insights warrant action.

### "How do I save insights for later reference?"

Only the latest set of insights is kept. If you need to keep insights for comparison or documentation, copy the text or take a screenshot before regenerating. Each new generation replaces the previous set.

### "Can insights analyze data from multiple Views?"

No, insights analyze only the current View after all pipeline transformations are applied. If you need patterns across multiple datasets, combine the data in a single View first, or generate insights separately for each View and compare them manually.

## Troubleshooting

### Insights take a long time

**Problem**: Generation seems stuck or takes much longer than expected.

**Solutions**:

- Check your internet connection stability
- Refresh the browser and try again. If generation fails, Mammoth shows an error message and the button returns to "View Insights"
- Verify the View contains data (empty Views cannot generate insights)

### Insights seem irrelevant or obvious

**Problem**: AI observations don't provide value for your use case.

**Solutions**:

- AI lacks business context—some patterns may not be actionable in your situation
- Add a Conditional Filter task to focus on a specific subset before generating
- Combine insights with the Data Quality Report
- Use insights as starting points for deeper investigation, not final answers

### No insights generated

**Problem**: AI returns fewer insights than expected or the list shows "No data to show".

**Solutions**:

- Ensure the View contains sufficient data variation (very uniform data has fewer detectable patterns)
- Check that columns contain meaningful business data, not just IDs or constants
- Verify data has fully loaded before generating
- Try regenerating—occasional processing issues can be resolved with a second attempt

### Insights don't reflect pipeline changes

**Problem**: Insights still reflect old data after adding or running transformation tasks.

**Solutions**:

- This is expected behavior—insights do not auto-update when the pipeline changes
- Click the Key Insights button again to analyze the transformed data
- Wait for pipeline tasks to finish executing before regenerating

---
Source: https://docs.mammoth.io/learn/explore-discover/view-insights.md · Updated: 2026-10-13