Run core evaluation locally
Upload CSV data, map your columns, and inspect accuracy, precision, recall, MCC, ROC AUC, PR AUC, calibration, and more without shipping your dataset to a backend.
Browser-based · privacy-first · no data leaves your machine
Evaluate binary classifiers in the browser.
Upload a CSV with actual labels and either probabilities or predicted labels. Inspect 12+ metrics, ROC/PR, calibration, threshold tuning, and paid HTML report export without building a full SaaS backend.
Binary classification only in this starter. No multiclass magic, no regression fairy dust.
Upload CSV, map columns, and evaluate locally in the browser.
Core evaluation happens locally in the browser. Payment, licensing, and site hosting still involve third-party services.
What it does
EvalBench is a lightweight evaluator for binary classifiers. It stays honest about scope and keeps the core analysis in the browser.
Operationalize threshold choices
Use the threshold slider for live feedback, then unlock the Pro threshold sweep table when you need a stronger operational recommendation.
Generate artifacts that buyers can use
Pro exports a standalone HTML report and cohort audit so the output is more than a screenshot. It becomes something you can send, archive, or review.
How it works
No hidden SaaS backend. The site is static-plus-serverless: browser-side evaluation, Cloudflare Pages Functions for entitlement checks, and Lemon Squeezy for checkout and license keys.
1
Load data
Upload a CSV, paste rows, or start with the bundled demo dataset.
2
Inspect the model
Review metrics, charts, calibration, and the confusion matrix in the browser.
3
Unlock Pro if needed
Use hosted checkout, then return through the unlock flow to enable report export, threshold sweep, and cohort audit on this browser.
Free vs Pro
Keep the launch simple. One free tier, one paid tier, no subscription theater.
| Capability | Free | Pro |
|---|---|---|
| CSV upload, paste, and sample dataset | Included | Included |
| Core metrics + confusion matrix | Included | Included |
| ROC / PR / calibration / score distribution | Included | Included |
| Downloadable metrics CSV | Included | Included |
| Standalone HTML report export | — | Included |
| Threshold sweep table | — | Included |
| Cohort audit by segment | — | Included |
| Paid session recovery on the same browser | — | Included |
FAQ
The honest answers matter more than the shiny ones.
Does model data leave the browser?
The evaluator processes uploaded CSV data in the browser. Payment, licensing, and basic web hosting still rely on external services.
What does Pro unlock?
Pro unlocks the HTML report export, threshold sweep tables, and cohort-audit breakdowns. The free tier still includes the core evaluator.
What kind of data can this handle?
This version is intentionally scoped to binary classification CSVs. Multiclass and regression are not included in this starter build.
Need help after checkout?
The support page covers receipt recovery, My Orders, re-unlock, and switching browsers.