Data Pitfalls

Data Pitfalls gives Claude the full taxonomy from Avoiding Data Pitfalls by Ben Jones (Wiley, 2019): 78 rules across eight domains, from technical trespasses and mathematical miscues to statistical slipups, graphical gaffes and design dangers.

Ask Claude to review your work for data pitfalls and it pulls the checklist that fits the work in front of it, whether that's a SQL query, a chart image, a dashboard, a written analysis or a full report. It checks every item against your actual work and reports each finding with its rule, the evidence, why it matters and how to fix it. Findings are marked active (evident in the work) or latent (dependent on data Claude can't see), and rated by whether they change the takeaway, weaken its support, or are polish.

It's especially useful right after Claude builds something for you. Ask it to review the dashboard it just made, and it checks every chart's query and design, then fixes what it finds.

You can also look up any pitfall by name for its definition, detection strategy, a bad and a good example and the fix, or search the rules by topic.

The connector is read-only and needs no account. It serves the taxonomy and nothing else: your work stays in your conversation with Claude and is never sent to the connector, which stores nothing and makes no AI calls of its own. Open source at github.com/bjonesdataliteracy/datapitfalls.