Start with the destination rule
Confirm whether the form means 100,000 bytes or 100 KiB, which file formats it accepts, and whether it also limits width or height. Use the stricter number when the wording is unclear so the upload has a safety margin.
Change the least visible variable first
For photographs, try JPEG or WebP quality before reducing dimensions. For logos or flat graphics, PNG may remain clearer but can miss a small target; WebP often provides a useful alternative when the destination accepts it.
Know when the target is unrealistic
A detailed multi-megapixel photo cannot always look clean at 100KB. If quality reaches its lower bound, reduce pixel dimensions gradually and inspect text, faces, and edges at actual size before accepting the file.
Keep a reproducible record
Save the achieved byte count, dimensions, output format, and chosen target with the downloaded file. Those facts are more useful than a vague compression percentage when the same portal must be used again.
Inspect the file that will actually be submitted
Open the downloaded result rather than relying on the browser preview alone. Zoom into small text, hair, product edges, signatures, and smooth gradients, because those regions reveal compression damage first. Confirm the filename extension matches the receiving form, check the exact byte count in your file manager, and keep the original beside the derivative. If the image is still above the limit, set a smaller width cap before accepting very low lossy quality. A modest reduction in dimensions often looks cleaner than severe artifacts across a larger frame.