Data Analyzer

Interpret CSV, JSON, and structured data, extract insights, identify patterns, and recommend appropriate visualizations.

Description

This skill provides a systematic approach to data analysis for tabular and structured data. It covers data profiling, statistical summary, trend identification, anomaly detection, and visualization recommendations. Designed for users who have data but need help understanding what it means and how to present it effectively.

When to Use

• Analyzing CSV or JSON data files
• Understanding trends in sales, traffic, survey, or any time-series data
• Comparing groups, segments, or categories
• Identifying outliers or anomalies in datasets
• Preparing data insights for reports or presentations
• Recommending chart types for specific data stories

Instructions

Step 1: Data Profiling

Before analysis, profile the data:

Dataset Profile:
- Rows: [count]
- Columns: [count]
- Column types: [list each column with type: numeric, categorical, date, text]
- Missing values: [percentage per column]
- Date range: [if applicable]
- Key metrics summary: [mean, median, min, max for numeric columns]

Identify data quality issues:

• Missing values (>5% in any column needs attention)
• Duplicates
• Inconsistent formats (dates, categories, units)
• Outliers that might be errors vs. genuine extreme values

Step 2: Analysis Framework

Apply the appropriate analysis based on data type and question:

For Time-Series Data (dates + values)

1. Trend Analysis: Is the metric growing, declining, or stable?

   • Calculate period-over-period change (MoM, YoY)
   • Identify inflection points (where trend direction changes)

2. Seasonality: Are there recurring patterns?

   • Weekly, monthly, or quarterly cycles
   • Compare same period across different years

3. Anomaly Detection: Any unexpected spikes or drops?

   • Flag values >2 standard deviations from the mean
   • Check if anomalies correlate with known events

For Categorical Comparisons

1. Ranking: Which categories lead/lag?

   • Sort by value
   • Calculate percentage of total for each category

2. Distribution: How are values spread?

   • Identify concentration (is 80% of value in 20% of categories?)

3. Correlation: Do categories relate?

   • Cross-tabulation between two categorical variables

For Numeric Relationships

1. Correlation: Do two metrics move together?

   • Note correlation direction and approximate strength
   • Caution: correlation ≠ causation — always state this

2. Segmentation: How do metrics differ across groups?

   • Compare averages/medians across segments

Step 3: Insight Extraction

Structure findings as:

## Key Findings

1. **[Finding title]**
   - What: Specific observation with numbers
   - So what: Business impact or implication
   - Action: Recommended next step

2. **[Finding title]**
   - ...