Skip to main content
Powered By Data

Data Pitfalls Detector

Check data work, whether done by humans or AI, for common pitfalls before you ship it. Paste a written analysis, drop in analysis code, or upload chart images, and get specific, actionable findings.

Avoiding Data Pitfalls book cover

Powered by datapitfalls, the open-source detector based on Ben Jones’s Avoiding Data Pitfalls. It checks against a 75-rule taxonomy spanning eight pitfall domains.

Ben Jones

Created by Ben Jones, author of Avoiding Data Pitfalls.

Mistaking the data for reality — overgeneralizing from limited evidence, or treating numbers as more certain than they are.Mishandling the data itself — dropped rows, broken joins, mismatched keys, or inconsistent categories.Calculation slip-ups — mixing units, averaging rates without weighting, or aggregating across uneven periods.Comparison and inference errors — treating an average as typical, ignoring uncertainty, or over-reading small samples.Flawed analysis — extrapolating a trend too far, overfitting a model, or choosing a metric that flatters the story.Charts that mislead — truncated axes, the wrong chart type, cherry-picked windows, or missing context.Design choices that trip up the reader — colorblind-unsafe palettes, inconsistent colors, or unclear interactions.Who or what the data leaves out — unheard voices and missing groups that skew the whole picture.

Hover or tap a domain to see what it covers.

Drag & drop chart images here

or paste from your clipboard, or

PNG, JPEG, GIF, or WebP · up to 4 images · 3 MB each

Your input is sent to Anthropic’s Claude API to perform the analysis and is not stored by this site. Findings are AI-generated guidance — use your own judgment and verify against your underlying data.

Prefer to run it yourself?

The detector is open source and runs from your terminal — scan charts, code, and documents with no limits.

datapitfalls — terminal
$ npx datapitfalls scan ./your-chart.png

Or install it globally: npm install -g datapitfalls

View on npmStar on GitHub

Found this useful, or spotted a problem?

Share feedbackSponsor on GitHubDonate via PayPal

This tool calls the Claude API on every scan. Donations help offset token costs and keep it free to use — thank you.

Want to use this with your organization?

If this is useful for your own work, it can do more for your whole team. We help organizations build pitfall-checking into their workflow — and the data literacy to back it up. Let's talk about what that looks like for you.

Talk to us