How usage and reliability are measured

Site decisions combine anonymous product events, search performance, automated tests, and manual browser checks. No single metric is treated as proof that a tool is useful or reliable.

Usage definitions

• View: a public page loaded in a browser.
• Attempt: a visitor begins a tool operation.
• OK: the tool produced its expected output.
• Fail: validation or processing ended without the expected output.
• Download: a generated file link or text download was activated.

Attempts per view can exceed 100% because one person may retry or run a tool multiple times. Success rate is successful outcomes divided by all recorded outcomes. Counts are directional, not unique people or audited market-share figures.

Search demand

Search Console impressions, clicks, click-through rate, average position, countries, devices, queries, and landing pages are reviewed together. Low-volume query exports are incomplete because search engines omit some data for privacy, so visible query totals do not always equal page totals.

Verification

Automated checks cover conversion logic, browser interactions, accessibility, routes, canonical URLs, and representative real-file downloads. A passing fixture proves that supported example; it does not prove every codec, malformed file, browser, or device combination.

Privacy wording

“No upload” means the selected file and generated result are not sent to a conversion server. The page still downloads its HTML, scripts, fonts or images, and—when needed—a WebAssembly processing engine. Anonymous application events and normal hosting requests are described in the privacy policy.