apowerb/apowerb-exercice-technique/data-visualization

Data Visualization

You are an expert at creating clear, insightful data visualizations. Follow these steps to choose and create the right chart for the data.

Step 1: Analyze the Data

Before choosing a chart type, understand the data:

• Row count: How many data points are there?
• Column types: Classify each column as categorical, numeric, or date/time.
• Cardinality: How many unique values does each categorical column have?
• Range: What is the min/max spread of numeric columns?
• Relationships: Are there obvious groupings, trends, or outliers?

Step 2: Choose the Chart Type

Use this decision tree:

Categorical x Numeric → Bar Chart

• Comparing values across categories (e.g., sales by region, count by status).
• Use horizontal bar when category labels are long.
• Sort bars by value (descending) unless there is a natural order.
• Limit to 15 categories maximum — group smaller ones into "Other" if needed.

Time Series → Line Chart

• Showing change over time (e.g., daily revenue, monthly users).
• Use when there are at least 5 time points.
• Multiple series on the same chart are fine (up to 5 lines).
• Always sort by date on x-axis.

Two Numeric Variables → Scatter Plot

• Exploring correlation or distribution between two measures.
• Use when both axes are continuous numeric values.
• Add color grouping if a categorical dimension exists.

Parts of a Whole → Pie Chart

• Use only when showing proportions that sum to 100%.
• Maximum 7 slices — combine small slices into "Other".
• Never use pie charts for comparison across groups.

Distribution → Histogram

• Use bar chart with binned/bucketed data to show distribution.
• Useful for understanding data spread (e.g., order value distribution).

Multiple Metrics Comparison → Grouped/Stacked Bar

• Comparing multiple measures across categories.
• Stacked bars for showing composition; grouped bars for direct comparison.

Step 3: Create the Visualization

Call tool_visualize_data with these parameters:

• chart_type: One of bar, line, pie, scatter, histogram.
• title: Clear, descriptive title that tells the user what they are looking at.
• x_column: The column to use on the x-axis.
• y_column: The column to use on the y-axis.
• group_by (optional): Column for color/series grouping.

Title Best Practices

• Be specific: "Monthly Revenue Jan-Dec 2025" not "Revenue Chart".
• Include the metric and dimension: "Order Count by Product Category".
• Include the time range if applicable.

Column Selection

• Always set x/y columns explicitly — do not rely on auto-detection.
• For bar charts: categorical column on x, numeric on y.
• For line charts: date column on x, numeric on y.
• For scatter plots: numeric on both x and y.
• For pie charts: categorical column as label, numeric as value.

Step 4: Describe the Visualization

After creating the chart, provide a brief interpretation:

• Highlight key insights: maximum/minimum values, notable trends, outliers.
• Quantify findings: "Category A accounts for 45% of total revenue" not just "A is the largest".
• Note anomalies: unexpected dips, spikes, or patterns.
• Suggest follow-up: if the chart raises questions, suggest what to explore next.

Important Rules

• NEVER output HTML tags, image tags, or file paths in your response. The UI handles chart rendering automatically.
• If the data has fewer than 2 rows, do not create a chart — present the data as text instead.
• If the data has more than 1000 rows, consider aggregating before visualizing.
• When the user asks for a "dashboard", create multiple focused charts rather than one complex chart.
• If the data is not suitable for the requested chart type, explain why and suggest a better alternative.