THE LAB
Small tests. Explicit limits.
Each note states the input, environment, method and result so a reader can reproduce it. A controlled fixture or geometry calculation can explain one behavior; it cannot predict every seller photo, browser or marketplace pipeline.
What is measured
The format test records the bytes produced from the same deterministic sample. The crop test calculates which centered pixels a frame removes. The memory note estimates one decoded RGBA buffer from pixel count.
What the results do not prove
The format sample is not a photograph and its byte counts are not a universal format ranking. Geometry does not recreate an Etsy or Shopify theme. Memory arithmetic is not a device limit. Use the local tools to test your own files and inspect the actual destination preview.
JPG, PNG or WebP for Product Photos? A Reference Encoder Benchmark
A reproducible reference-encoder comparison of image formats, with a clear record of what varies by encoder and source image.
Rechecked 2026-09-22Framing experimentCrop Geometry: How Much of a Product Image Can a Thumbnail Remove?
Use transparent calculations and controlled frames to see what centered 1:1, 4:3 and 3:4 crops remove from the same source.
Rechecked 2026-09-22Browser experimentBrowser Image Limits: Why a Small Photo File Can Use a Lot of Memory
A reproducible explanation of pixel buffers, browser limits and why local processing still needs conservative batch sizes.
Rechecked 2026-09-22