Run the full test-validation stack on your data: reliability, factor analysis, IRT, measurement invariance and norms, plus adverse-impact and item-level DIF checks. Then let Claude draft the results section in APA 7 or plain language. No install, no code required, and R when you want it.
One workbench for test development, fairness evidence, written reports and data that lives in a database.
Everything a technical manual or validation paper asks for.
The adverse-impact evidence stack, layered the way reviewers read it.
Every analysis has an Add to report button. Claude drafts the narrative.
Upload a file, or connect the database your responses already sit in.
Plus the everyday toolkit: ANOVA, regression, correlation, chi-square, mediation and moderation, path analysis, cluster analysis, power analysis, time series and text analysis.
A read-only workbench on your live response database. Pick a test and a date range, every item arrives labelled with its question text, and the technical report is drafted for you.
The complete adverse-impact evidence stack for NYC Local Law 144, California ADS and EU AI Act work: four-fifths, effect sizes, item-level DIF and invariance, with a client-ready summary and an APA appendix.
CFA, graded-response IRT, DIF by language or sex, measurement invariance, cut-offs and norms without writing R, and the R script downloadable when a reviewer asks for it.
A browser lab students open from one link. Nothing to install, APA output scaffolded, and a sandboxed R console for the students who want to see the code.
Their research lead was iterating scales against hundreds of completions a day, and every analysis started with an engineer export, a clean-up and a codebook. We put a branded instance of DataWeaver directly on their production database.
Ask for the referenceUpload CSV or Excel, connect a database, or start with a built-in demo dataset.
Pick from 24 modules. The AI advisor suggests one from your data and research question.
Add results to the report queue, let Claude draft the narrative, export to Word or HTML.
Self-serve plans below. Organisation plans are quoted, because they come with setup.
Explore with built-in datasets
For students with an academic email
Faculty and clinical researchers: ask about Academic Annual at $600 per year.
Or $1,200 per year, two months free
For consultancies and research groups
Team plus the fairness pack for firms delivering adverse-impact and bias-audit work: four-fifths with z-tests and Holm, Cohen's d, item-level DIF, invariance, counsel-ready report styles, JSONL audit trail for record retention, and two hours of methods consultation a quarter.
Talk to usYour brand, your domain, a read-only connector to your response database with question text as labels, the full stack, MCP access for your analysts. Hosted on our EU infrastructure or deployed on yours. Setup is a quoted two-to-four-week project.
Book a schema-fit reviewCourse workspace, deep-linked assignments, term-limited student accounts without individual billing, free instructor seats. Department licences from $6,000 a year.
Ask about a pilotNeed it done for you? Scale validation reports, adverse-impact and DIF evidence packs, and technical report sections, run in DataWeaver and reviewed by a psychometrician, from $1,500. Ask us.
The app runs on servers in the EU (Hetzner, Germany) over HTTPS. Uploaded files are processed in your session. Database credentials are held server-side as environment variables and never reach the browser. Report drafting sends the statistical output, not your raw data, to the Anthropic API; on Team plans you can use your own key.
Not yet. DataWeaver is a small company and we would rather say so than imply otherwise. We will sign your data-processing agreement, we recommend starting with de-identified data, and for organisations that need it we deploy the embedded edition on your own infrastructure.
Yes. The engine is standard R: lavaan, mirt, lordif, difR, pROC and psych, with package versions locked. Paid plans include a sandboxed R console over your loaded dataset, and you can download the script and run it in plain R.
No. Claude drafts the narrative from the actual output (fit indices, loadings, flags, effect sizes). Every figure in the text comes from the tables above it, the draft is editable, and a named person should still sign it.
CSV and Excel (.xlsx, .xls), or a direct database connection. SPSS .sav files are not supported yet; export to CSV from SPSS and the variable names carry over.
Yes. Every tab is deep-linkable, buttons have stable labels, and the MCP endpoints let Claude Code or Cowork list datasets, read schemas and run R scripts in the same sandbox the app uses.
Start with the demo, or send us a de-identified extract and we will run it through and send you the report.