What the file size is actually counting
File size is the number of bytes stored on disk. For a raster image those bytes describe the pixels: their color values, the structure around them, and a little information about the file itself.
That is why two pictures that look identical can weigh very different amounts. One may repeat smooth areas efficiently while the other stores every variation, and the difference shows up in the number you see in your file manager.
What quality loss looks like in practice
Lossy compression, which JPEG uses, removes information that the eye is less likely to notice. Pushed gently, the result looks the same at normal size. Pushed hard, it shows up as blocking around edges, smeared gradients, and mush around fine text.
Lossless compression, which PNG uses, removes nothing: the pixels that went in are the pixels that come out, and the file stays large as a result. There is no quality setting to tune, only the decision of how much detail you are willing to give up.
The trade-off is not a straight line
Early reductions usually cost very little you can see, because the first data removed is the least visible. Keep pushing and each further step buys fewer bytes while taking more noticeable detail with it, so the last stretch toward a tight limit is always the most expensive.
Resolution behaves the same way. Dropping from a camera-sized image to a screen-sized one is barely noticeable, while squeezing far below the display size starts to blur edges and thin lines that were once crisp.
Decide by what the image is for
A photo meant for print or for your own archive should stay large and lightly compressed, because you may need the detail later. A picture inside an email, a form, or a long web page has a job that a modest file does perfectly well.
When a destination sets a limit, treat that number as the constraint and aim for the largest file that still sits under it. Going far below the limit throws away detail nobody asked you to give up.
How this tool handles the balance
Enter a maximum in KB or MB and the compressor searches for a JPEG at or below that size, then verifies the finished file before reporting success. It targets a ceiling rather than an exact match, because encoding varies slightly between browsers.
For very tight targets it may reduce pixel dimensions to get under the limit, and a transparent PNG background comes back white in the JPEG output. The result screen shows the original and output sizes together, so you can judge the trade-off with numbers as well as your eyes.