Scanned 2026-10-30 with imgtune 0.1.0 (tool commit 14b3c12), six viewports, SSIMULACRA2 delta 2, page fully scrolled. The hatched part of the estimate is a local re-encode, not a measured file (see how it is measured). The 15% line is the study's bar. Slot is the CSS width the image fills, needed is the device pixels it needs (widest viewport), served is the file's own width.
A1 · production3,250KB / view
estimated saving 33.2%
A2 · production1,263KB / view
estimated saving 53.1%
A3 · production3,355KB / view
estimated saving 38.2%
A4 · production9,254KB / view
estimated saving 35.9%
A5 · production7,714KB / view
estimated saving 16.5%
A6 · production9,078KB / view
estimated saving 27.7%
Production sites are anonymous: no names, images or URLs. Study v4 numbers, from the study's results.
How to read it
The estimated saving is the number to act on. Its hatched part is an estimate, never presented as measured. See How it measures for how the estimate is built.
Slot is the CSS width the image fills; needed is the device pixels it needs at the widest viewport; served is the file's own width.
The numbers are the same as the study method, on fresh scans. The Study page keeps the v4 table of 2026-10-28.