name: data-analyzer description: Expert data analysis and insights generation using Python Pandas, visualization, and statistical methods

Data Analyzer

Overview

Specialized skill for comprehensive data analysis, statistical insights, and business intelligence reporting. Use when users need data-driven decisions, trend analysis, or actionable insights from structured data.

When to Use

Activate this skill when:

• Analyzing business data (sales, revenue, customer metrics)
• Generating reports with visualizations and statistical summaries
• Performing exploratory data analysis (EDA)
• Detecting anomalies, outliers, or patterns in data
• Creating dashboards or executive summaries
• Time series analysis and forecasting preparation
• A/B test analysis and experiment interpretation
• Customer segmentation and cohort analysis

Use especially for:

• Multi-source data aggregation and comparison
• Month-over-month or year-over-year analysis
• Performance metrics and KPI tracking
• Data quality issues and validation

Core Capabilities

1. Data Loading & Processing

import pandas as pd
import numpy as np

# Load multiple formats
df = pd.read_csv('data.csv')
df = pd.read_excel('report.xlsx')
df = pd.read_sql(query, connection)

# Data cleaning
df.dropna(subset=['key_column'])
df['date'] = pd.to_datetime(df['date'])

2. Statistical Analysis

• Descriptive statistics (mean, median, std, percentiles)
• Distribution analysis (histograms, box plots)
• Correlation matrices and heatmaps
• Group-by aggregations and pivot tables
• Rolling averages and trend calculations

3. Visualization

import matplotlib.pyplot as plt
import seaborn as sns

# Time series
plt.figure(figsize=(12, 6))
sns.lineplot(data=df, x='date', y='revenue')

# Distributions
sns.histplot(data=df, x='sales', bins=30)

# Correlations
sns.heatmap(df.corr(), annot=True)

4. Business Intelligence

• Same-store growth calculations
• Year-over-year and month-over-month comparisons
• Customer segmentation (RFM analysis)
• Churn prediction indicators
• Product performance rankings
• Channel attribution analysis

Analysis Workflow

Phase 1: Data Understanding

1. Load and inspect data structure
2. Identify data quality issues (nulls, duplicates, outliers)
3. Understand business context and metrics definitions
4. Validate data sources and calculations

Phase 2: Exploratory Analysis

1. Univariate analysis (distributions, summary stats)
2. Bivariate analysis (correlations, relationships)
3. Time series patterns and trends
4. Segment-based analysis (by channel, product, region)

Phase 3: Insight Generation

1. Identify key drivers and performance indicators
2. Detect anomalies or unexpected patterns
3. Compare against benchmarks or targets
4. Generate actionable recommendations

Phase 4: Reporting

1. Create executive summary with key findings
2. Visualize critical insights (charts, heatmaps)
3. Provide detailed methodology and assumptions
4. Suggest next steps or areas for deeper analysis

Common Analysis Patterns

Month-Over-Month Growth

def calculate_mom_growth(df, value_col, date_col='date'):
    df = df.sort_values(date_col)
    monthly = df.set_index(date_col)[value_col].resample('M').sum()
    growth = monthly.pct_change() * 100
    return growth

Customer Segmentation

# RFM Analysis
recency = df.groupby('customer')['date'].max()
frequency = df.groupby('customer')['order_id'].count()
monetary = df.groupby('customer')['revenue'].sum()

# Create segments
segments = pd.concat([recency, frequency, monetary], axis=1)
segments['segment'] = pd.qcut(segments['monetary'], q=4, labels=['D','C','B','A'])

Anomaly Detection

from scipy import stats
z_scores = np.abs(stats.zscore(df['revenue']))
anomalies = df[z_scores > 3]

Best Practices

1. Always validate data quality before analysis
2. Document assumptions and calculation methods
3. Use appropriate visualizations for the data type
4. Provide context for statistical significance
5. Include business implications, not just numbers
6. Sanity check results against domain knowledge
7. Handle time zones correctly in time series data
8. Preserve raw data and work on copies

Common Pitfalls to Avoid

• ❌ Mixing week-based and date-based aggregation without alignment
• ❌ Calculating growth rates on single data points instead of periods
• ❌ Using sum() for rate metrics (use weighted average instead)
• ❌ Ignoring data quality issues in source files
• ❌ Comparing across different time zones without normalization
• ❌ Treating correlation as causation
• ❌ Overfitting patterns to noise
• ❌ Ignoring seasonality in time series

Output Templates

Executive Summary Format

## Key Findings
- [Most critical insight with supporting metric]
- [Second most important finding]
- [Unexpected pattern or anomaly]

## Performance Overview
- Total Revenue: [formatted value] ([change%] vs last period)
- Key Metric: [value] (target: [target], achievement: [%])
- Top Driver: [driver name] contributed [contribution%]

## Actionable Recommendations
1. [Specific action with expected impact]
2. [Second recommendation]
3. [Third recommendation]

## Detailed Analysis
[Full analysis with visualizations and methodology]

Dependencies

# Core requirements
pandas>=2.0.0
numpy>=1.24.0
matplotlib>=3.7.0
seaborn>=0.12.0
scipy>=1.10.0

# Optional but recommended
scikit-learn>=1.2.0  # For clustering and classification
plotly>=5.14.0       # Interactive visualizations
openpyxl>=3.1.0      # Excel file support

Example Usage

User: "Analyze this week's sales data and identify trends"
Agent: [Activates data-analyzer skill]
- Loads sales data from CSV/Excel
- Performs time series decomposition
- Identifies top products and channels
- Detects anomalies and patterns
- Generates summary with visualizations
- Provides actionable insights

User: "Compare customer retention across channels"
Agent: [Activates data-analyzer skill]
- Calculates retention metrics by channel
- Performs statistical significance testing
- Visualizes retention curves
- Identifies best practices from high-retention channels

Related Skills

• systematic-debugging - For investigating data quality issues
• python-testing-patterns - For validating analysis code
• csv-data-summarizer - For quick CSV overview