Resizing an image sounds like the most basic thing in the world. Drag a corner, type in a number, done. But spend five minutes zoomed in on a badly resized image and you'll see the problem: blurry edges, weird color fringing around high-contrast areas, distorted proportions. These don't come from bad luck. They come from specific choices the software makes during resizing, most of which happen invisibly.
You don't need to understand the math to get better results. You just need to know which choices matter.
What actually happens when you resize
An image is a grid of pixels. Resizing changes the size of that grid. When you make an image smaller, you're removing pixels - the software has to decide which information to keep and which to throw away. When you make it larger, you're adding pixels that didn't exist before - the software has to invent values based on what surrounds them.
The algorithm used for this matters a lot. The simple ones are fast but produce blurry or blocky results. The better ones take more processing time but preserve edge sharpness and color fidelity much more cleanly. For most online tools, you don't get to choose directly - you just get the tool's default, which is why results vary so much between different services.
Always maintain the aspect ratio
This is the one that causes the most obvious problems. If you specify both width and height for a resize and those dimensions don't match the original proportions, something has to give. Either the image gets stretched, which makes people look weirdly wide or tall, or the tool adds padding to fill the gap.
The fix is simple: lock the aspect ratio when resizing. Specify one dimension and let the other calculate automatically. The image will shrink or grow proportionally and nothing will look distorted.
The exception is when you genuinely need specific dimensions that don't match the original ratio - like a social media profile photo that needs to be exactly 400x400 and your source image is 800x600. In that case, the right approach is to crop first, then resize. Crop to a square, then resize to 400x400. Don't just stretch it.
Target sizes that actually make sense
For website hero images and wide banners, somewhere between 1200 and 1920 pixels wide covers most screens without producing an unnecessarily heavy file. At that width, a well-compressed JPEG typically stays under 200KB.
For social profile photos, each platform has recommended dimensions and it's worth using them. When you upload an image that's larger than needed, the platform recompresses it - usually with more aggression than you'd choose yourself. Sizing to their recommendations first gives you more control over how it looks.
Email attachments you want people to open without their inbox exploding: keep image dimensions under 1200px on the long side and you'll rarely have size issues.
Saving after resizing
The format and settings you save with matter as much as the resize itself. JPEG applies lossy compression every time you save - so if you resize and re-save a JPEG multiple times, each round degrades it a little more. If you're doing any iterative editing, save as PNG between steps and only export to JPEG at the end.
For web use, WebP is worth considering. Files are 25-30% smaller than equivalent JPEGs with no visible quality difference. Browser support is solid now - above 97% globally. The main downside is that people who download the image and open it in older desktop software might have issues, so if your images are meant to be downloaded and used in other contexts, JPEG is still the safer choice.