Analyzing Insights

Analyzing Insights

Quick Start

When analyzing an insight:

1. Call view_insight for saved insights, or use provided chart data directly
2. Identify the chart type and what it measures
3. Look for patterns (trends, seasonality, anomalies)
4. Quantify observations with specific numbers
5. Provide actionable interpretation

When to Use This Skill

• User asks "what does this insight show?"
• Analyzing visualization results
• Identifying trends or anomalies
• Explaining patterns in data
• Generating insights from visual data

Chart Types

| Type | Best For | Look For | |------|----------|----------| | Line | Time series trends | Direction, inflection points | | Bar | Category comparison | Relative sizes, outliers | | Area | Volume over time | Growth, composition | | Pie | Distribution | Proportions, dominance | | Table | Detailed data | Patterns, sorting | | Metric | Single values | Change from baseline | | BarList | Ranked items | Top performers, long tail |

Analysis Framework

1. Describe What You See

Start with objective observations:

• What is being measured?
• What is the time range?
• What are the key dimensions?

2. Identify Patterns

Look for:

• Trends: Upward, downward, flat
• Seasonality: Weekly, monthly, yearly cycles
• Anomalies: Spikes, drops, outliers
• Inflection points: Where direction changes

3. Quantify Observations

Always include numbers:

• Absolute values
• Percentage changes
• Comparisons to baselines

4. Provide Interpretation

Explain significance:

• Why might this be happening?
• What are the implications?
• What actions should be considered?

Pattern Recognition

Trend Patterns

Upward Trend

• Consistent growth over time
• Look for: slope, acceleration/deceleration
• Note: sustainability, growth rate

Downward Trend

• Consistent decline over time
• Look for: rate of decline, stabilization
• Note: severity, projected impact

Flat/Stable

• No significant change
• Look for: volatility within range
• Note: whether stability is expected

Seasonality Patterns

Weekly Cycles

• Weekday vs weekend differences
• Monday dips, Friday spikes
• Note: business day patterns

Monthly Cycles

• Beginning/end of month patterns
• Billing cycles, payroll effects
• Note: calendar effects

Yearly Cycles

• Holiday impacts
• Seasonal business patterns
• Note: YoY comparisons

Anomaly Patterns

Spikes

• Sudden increase
• Look for: magnitude, duration
• Consider: campaigns, events, bugs

Drops

• Sudden decrease
• Look for: recovery pattern
• Consider: outages, issues, seasonality

Outliers

• Values far from normal range
• Look for: explanation
• Consider: data quality, real events

Analysis by Chart Type

Line Charts

Focus on:

• Overall trend direction
• Volatility/smoothness
• Inflection points
• Comparisons between lines

Questions to answer:

• Is the metric growing or declining?
• Are there regular patterns?
• Where are the peaks and troughs?

Bar Charts

Focus on:

• Relative bar heights
• Ordering (if applicable)
• Gaps between categories
• Outlier categories

Questions to answer:

• Which category leads/lags?
• Is distribution expected?
• Are there surprising values?

Pie Charts

Focus on:

• Dominant segments
• Small segments
• Unexpected proportions

Questions to answer:

• Is any segment too dominant?
• Are proportions as expected?
• Has composition changed?

Tables

Focus on:

• Sorting patterns
• Extreme values
• Null/missing data
• Relationships between columns

Questions to answer:

• What patterns emerge?
• Are there data quality issues?
• What correlations exist?

Quantification Guidelines

Describing Changes

| Change | Description | |--------|-------------| | +/-5% | Slight change | | +/-10-20% | Moderate change | | +/-20-50% | Significant change | | +/-50%+ | Dramatic change | | 2x | Doubled | | 3x | Tripled |

Time Comparisons

• WoW: Week-over-week
• MoM: Month-over-month
• QoQ: Quarter-over-quarter
• YoY: Year-over-year

Statistical Context

• Compare to historical average
• Note standard deviation if known
• Reference typical ranges

Communication Patterns

Good Insight Format

[What]: Revenue increased 23% this week
[Context]: From $45,000 to $55,350
[Comparison]: This is 15% above the 4-week average
[Interpretation]: Likely driven by the holiday promotion
[Recommendation]: Consider extending the campaign

Avoid Vague Statements

| Bad | Good | |-----|------| | "Revenue went up" | "Revenue increased 23% to $55,350" | | "There's a trend" | "Daily active users grew 5% WoW for 6 consecutive weeks" | | "Something changed" | "Conversion dropped from 3.2% to 2.1% on March 15" |

Common Pitfalls

• Making claims without numbers
• Ignoring context (seasonality, events)
• Confusing correlation with causation
• Over-interpreting normal variance
• Missing obvious anomalies
• Not considering data quality issues

Reference Files

• [Insight types detail](references/insight-types.md)
• [Visualizations guide](references/visualizations.md)
• [Anomaly patterns](references/anomaly-patterns.md)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: altertable-ai
• Source: altertable-ai/skills
• License: MIT
• Homepage: https://altertable.ai

Install and usage instructions live in the source repository linked above.