Public validation experiment
Privacy notice
FeedGuardrail is a provisional, no-charge product-research experiment. It is designed so product-feed data stays in your browser.
Effective: August 29, 2026 · Version 1.6
Browser-level choice
Anonymous product-research measurement
Pausing stops future experiment events and research reservations on this browser. Feed scanning, comparison, reports, and evidence verification continue to work locally.
FeedGuardrail also honors enabled Global Privacy Control and browser Do Not Track signals. A browser-level signal can keep measurement paused after this preference is resumed. Pausing does not delete earlier records; use the deletion control below for that.
What never leaves your device
Feed files, evidence HTML or JSON files, filenames, file hashes, product IDs, titles, descriptions, prices, URLs, images, validation rows, comparison and verification details, and downloaded reports are processed in browser memory. FeedGuardrail does not upload or store them.
What the experiment records
The application creates a random browser identifier in local storage. It records page and troubleshooting-guide visits, an allowlisted referral label, scan starts and completions, whether a run used a built-in sample or local files, coarse file-size and row-count ranges, aggregate error and warning counts, correction and client-summary downloads, Regression Guard starts and completions, aggregate resolved/introduced/persistent counts, a fixed comparison verdict, evidence-download format, evidence-verification starts and a fixed verification outcome, actionable/not-actionable feedback, pricing and plan clicks, and optional reservation answers. It never records the compared filenames, product IDs, hashes, evidence contents, feed values, or file contents. It does not store a referring URL or arbitrary query parameters. Reservation answers are limited to fixed choices for plan, work role, catalog-size range, and primary monitoring need; there is no free-text or contact-information field.
Reports you download
Correction CSVs, client summaries, and Regression Guard evidence packs are created on your device and are not uploaded to FeedGuardrail. Evidence packs include both local file hashes and the ruleset version but exclude filenames and product IDs by default; you can deliberately include those identifiers. Feed rows and attribute values are never embedded in an evidence pack. HTML evidence contains a public FeedGuardrail link so an intended recipient can locally verify the accompanying JSON and exact feed files. A client summary can contain the selected filename, product IDs, issue groups, and source-row references. Review any download and share it only with intended recipients.
Why and how long
These records are used only to decide whether this product should be built and which problems matter. Application records are automatically deleted after 90 days. The hosting provider may maintain short-lived security and operational logs under its own policies.
Sharing and selling
Experiment records are not sold, used for advertising, or shared with data brokers. They are stored with the site’s infrastructure provider solely to operate and evaluate this experiment.
Local storage and cookies
FeedGuardrail uses local storage for the random experiment identifier, scan count, and an optional measurement-pause preference. It does not set application advertising or cross-site tracking cookies.
Your control
Pause future anonymous measurement with the control above. Use the button below from the same browser to delete the experiment events and reservation associated with its existing random identifier. Deletion does not create a new identifier. A new identifier is created only after measurement is active and you use a measured product flow again.
Delete this browser’s experiment record
This is immediate and cannot be undone. It does not affect files or reports because FeedGuardrail never stored them.