You have data. You have a point to make. You open your dashboard builder, stare at 40 chart types, and pick... a pie chart. Because it looks nice.

That's the wrong approach — and most people do it.

The chart type you choose is not an aesthetic decision. It's a communication decision. The wrong chart doesn't just look bad, it actively misleads your audience and buries the insight you're trying to share.

This guide will give you a framework for choosing the right visualization every time, walk you through the most important chart families, and show you how to use them to tell a story that lands.

Before you open any dashboard tool, ask yourself:

"What relationship am I trying to show?"

Every chart type is designed to show one of these six things:

Get this right and the rest becomes easy. Get it wrong and no amount of color-coding will save you.

Use comparison charts when you want to show how values rank against each other across categories.

The most versatile chart in existence. Use a column chart (vertical bars) when you have fewer than 8 categories and the category labels are short. Use a bar chart (horizontal bars) when you have many categories or long labels — it's much easier to read.

Common mistake: Using 3D bars. They distort values and make the chart harder to read. Never do it.

When to use: Comparing sales by region, tracking KPIs across teams, ranking features by importance.

When NOT to use: Showing change over time across many time points (use a line chart instead), or showing proportions of a whole (use a treemap).

If you have subcategories, you can either group bars side-by-side or stack them.

• Grouped bars: Better for comparing individual subcategory values
• Stacked bars: Better for comparing totals while still showing composition

Distribution charts answer the question: "Is my data concentrated around one value, or is it spread out? Are there outliers?"

A histogram splits your data into bins and counts how many values fall into each bin. It's the first chart you should make when you get a new numeric column.

What to look for:

• Symmetric bell curve: Normal distribution, most values cluster around the mean
• Right-skewed (long tail to the right): A few very large values are pulling the average up — the median is a better summary than the mean
• Bimodal (two peaks): Two distinct groups in your data — worth investigating why

Box plots pack a lot of statistical information into a small space: median, quartiles, and outliers all at once. They're especially useful when comparing distributions across multiple groups.

Read a box plot: The line in the middle is the median. The box spans from the 25th to the 75th percentile (the interquartile range). Whiskers extend to 1.5× the IQR. Any dots beyond the whiskers are outliers.

A violin plot is a box plot with a twist: it shows the full distribution shape, not just the summary statistics. Use it when the shape of the distribution matters (e.g., you want to know if it's bimodal, or heavily skewed).

The scatter plot is your go-to for exploring relationships between two continuous variables. Each dot is one data point. If the dots form an upward trend from left to right, there's a positive correlation. Downward trend: negative correlation. Shotgun pattern: no relationship.

Pro tip: Add a trend line (linear regression) to make the relationship immediately obvious. AICU adds this automatically when you hover over the plot settings.

When you have many numeric columns and want to understand how all of them relate to each other at once, the correlation heatmap is the answer. Each cell shows the correlation coefficient between two variables, colored from red (negative) to green (positive).

Read it quickly: Ignore the diagonal (a variable always correlates perfectly with itself). Look for strong colors off the diagonal. Here, Revenue and Users have a strong positive correlation (0.82), while Churn is negatively correlated with Revenue (-0.31) — makes sense.

Pie charts get a bad reputation — often deservedly. The human eye is terrible at comparing angles and arc lengths. If you have more than 4-5 slices, or if the slices are similar in size, a bar chart will communicate the same information much more clearly.

Use a pie chart only when: You have 2-4 categories and you specifically want to emphasize that they add up to 100%.

For hierarchical data (categories with subcategories), a treemap or sunburst chart is far more powerful than a nested pie. Area is easier to compare than angles, and hierarchies are immediately visible.

Use a treemap for flat comparisons of many items. Use a sunburst when the hierarchical structure is the point of the story.

The line chart is the default choice for data over time. It works because our brains naturally read left-to-right as the passage of time, and lines make trends and turning points obvious.

Tip: If you're showing multiple lines, limit yourself to 4-5 at most. More than that and the chart becomes a spaghetti mess.

Waterfall charts are underused and underrated. They're perfect for showing how an initial value builds up (or erodes) through a series of positive and negative contributions.

Classic use cases: showing how revenue minus costs equals profit, how a budget was spent, or how a metric changed from one period to the next.

Sankey diagrams visualize flow — how things move from one state to another. The width of each band is proportional to the volume flowing through. Use them for user journeys, process flows, budget allocations, or energy flows.

Choosing the right chart is step one. Turning it into a story is step two.

Most people build a chart and then let the audience figure out what it means. Flip it. Put your conclusion in the title.

"Revenue by Region, Q3 2025" — tells nothing.

"DACH Revenue Up 34% — Outpacing Every Other Region" — tells the story.

The chart proves the title. The title tells the story.

Every element on a chart that doesn't carry information is stealing attention from the elements that do. Ask yourself:

• Does this gridline help? (Usually no — make them lighter or remove them)
• Does this legend label add information? (Only if there are multiple series)
• Does this decimal place matter? (Rarely — round to 1 or 2 significant figures)

The goal is to make the insight impossible to miss.

Color should communicate information, not decoration. Use one highlight color to draw attention to the most important thing, and use muted tones for everything else.

See how the single highlighted bar makes the story obvious in under a second? That's intentional design.

Numbers tell you what happened. Annotations tell you why. Add a text annotation at the turning point in your time series to mark the product launch, the campaign, or the outage. Context transforms a chart from an observation into an explanation.

Your data has memory. A weekly view can look volatile; the same data on a monthly view shows a clear upward trend. A monthly view can hide a sudden drop that a daily view reveals. Always ask: what time window best represents the reality of what happened?

Great data visualization isn't about making things look pretty. It's about removing friction between your insight and your audience's understanding.

The formula is simple:

1. Ask what relationship you're showing
2. Pick the chart designed for that relationship
3. Remove everything that doesn't help
4. Lead with the conclusion, not the data

The charts above were built interactively inside AICU's plot dashboard — no code, just data and a few clicks. If you want to explore your own dataset visually, try loading it and asking the AI to suggest plots. It will analyze your columns and propose relevant visualizations as questions, so you can build the story without having to know the chart names first.

[1] Tufte, Edward R. The Visual Display of Quantitative Information. Graphics Press, 2001.

[2] Cairo, Alberto. How Charts Lie: Getting Smarter about Visual Information. W. W. Norton, 2019.

[3] Plotly Documentation. https://plotly.com/python/

[4] AICU Plot Types Documentation. https://aicuflow.com/docs/tool/plots