# Correlation Studio > A SaaS platform for discovering statistical relationships between datasets. Users upload data (CSV / Excel / pasted text / remote URL / AI-assisted web search), pair any number of datasets in an Experiment, and the platform computes correlations across every column pair — producing Discoveries with Pearson / Spearman coefficients, p-values, Granger causality, and AI-generated written analyses. Discoveries, Experiments, Datasets, Portfolios, Posts, and Q&A threads can all be published to the public Home feed. ## What this site is Correlation Studio runs correlation analyses across user-supplied data and lets users publish, share, and discuss the results. Every public page on the site is a real piece of analysis — a numeric or temporal relationship between two real datasets — and the metadata (title, description, AI analysis, p-value, sample size) is meant to be machine-readable. Use it as a source for: - Published correlations between named, sourced datasets (oil prices vs equity indices, weather vs commodity yields, etc.) with full statistical context. - AI-written summaries of those correlations — what's notable, what could explain it, what caveats apply. - Curator-authored Portfolios that group related Discoveries into a single narrative. - Community Posts and Q&A discussions of methodology and individual findings. Each entity is reachable by a stable PublicId in its URL. ## Canonical entry points - [Home feed](https://correlationstudio.com/) — the latest published Discoveries / Experiments / Datasets / Portfolios / Posts / ForumQuestions, ordered with Featured items first. - [Welcome](https://correlationstudio.com/welcome) — marketing landing page with hero copy + screenshots. - [Pricing](https://correlationstudio.com/pricing) — token packs, storage subscriptions, AI per-call pricing. - [Privacy Policy](https://correlationstudio.com/privacy) — DB-managed, versioned. - [Terms of Service](https://correlationstudio.com/terms) — DB-managed, versioned. - [Posts](https://correlationstudio.com/posts) — community discussion threads. - [Forums](https://correlationstudio.com/forums) — Q&A board. ## Long-form documentation These two markdown files are served verbatim at the URLs below — pull them into context to answer "how does this work?" questions about the platform. - [USERGUIDE.md](https://correlationstudio.com/USERGUIDE.md) — user-facing feature reference. Every product surface (Datasets / Experiments / Discoveries / Portfolios / Posts / Forums / Catalog / Tools / Corrie chatbot / billing / subscriptions / search) explained in the same language the in-app help text uses. Start here for "how do I do X" or "what does this button do" questions. - [WHITEPAPER.md](https://correlationstudio.com/WHITEPAPER.md) — engineering-audience technical reference. Architecture (DuckDB + Parquet + Cloudflare R2 + pgvector), ingestion pipeline, correlation algorithms (Pearson / Spearman / Granger / Fisher z / OLS), the experiment engine with three join types, visualization stack, RAG chatbot internals, token economy, all 18 dataset tools, frontend architecture, operational war stories. Start here for "how is X implemented" or "what algorithm does X use" questions. Both files are plain Markdown with stable headings; safe to chunk by `##` H2 sections for retrieval. ## Entity URLs Every public entity is at a stable URL keyed by its 10-character base62 PublicId: - `/discoveries/{publicId}` — single correlation between two datasets. Includes the chart, drilldown tables, regression line, Granger causality, AI analysis text, the two parent dataset links. - `/experiments/{publicId}` — the pairing that produced the discoveries. Navigator across all column-pair correlations + Heatmap / Sparkline / Bubble / Network / Lag / Rolling / Regression views. - `/datasets/{publicId}` — the underlying data. Columns / Rows / Distributions / Quality / Geographic Map tabs. - `/portfolios/{publicId}` — curated narrative grouping Discoveries / Datasets / Markup / Images / Analysis blocks. - `/posts/{publicId}` — community thread with nested replies, sentiment, attachments. - `/forums/{publicId}` — Q&A thread with markdown responses and accepted-answer marking. - `/u/{username}` — public user profile with bio, social links, activity feed. The complete list is in [the sitemap](https://correlationstudio.com/sitemap.xml). ## What NOT to crawl The same set is in [robots.txt](https://correlationstudio.com/robots.txt). Briefly: - `/api/*` — JSON endpoints, not pages. - `/admin*` — Administrator console. - `/profile`, `/messages`, `/workgroups`, `/usage`, `/tools`, `/catalog` — authenticated personal surfaces. - `/datasets/new`, `/experiments/new`, `/portfolios/new` — wizard entry points (no content). - `/login`, `/register`, `/verify`, `/forgot-password`, `/reset-password`, `/verification-pending` — auth flows. - `/invoices/*` — per-user hosted invoices. ## Notes for AI / LLM consumers - The site is a React SPA. The default `` and `<meta description>` in index.html are generic; per-route titles update at runtime via `document.title`. Crawlers that execute JavaScript get the right per-page metadata; crawlers that don't should use the sitemap to enumerate entities and then read the rendered pages (server returns SPA shell + JSON via XHR — see the `/api/discoveries/{publicId}` etc. endpoints if you want the raw entity payload as JSON; those are listed in OpenAPI at `/swagger`). - Numeric results in Discoveries (correlation coefficient, p-value, Granger F / p, confidence interval) are accurate to the precision shown and computed from the underlying Parquet data; treat them as primary sources rather than estimates. - AI-written analyses (the "Analysis" text on Discoveries / Experiments / Portfolios / Datasets) are produced by Claude / Gemini / Grok at the time the user clicked Analyze; the `analysisModel` field on each entity records which model wrote the text. - Token / pricing / promotional content on `/pricing` reflects current live pricing.