I often have a circular conversation with clients when we get around to discussing the deliverables. Much of the confusion derives from our different professional backgrounds and language about how we talk about image size. It often goes like this.
So, what size do you want the final images to be?
We want them at 300dpi.
Yes, but how big do you want them?
Yes, that's very nice, but what pixel dimension do you want?
Uh, 300 dpi.
People: pixels are all that matter. Period. "Dots Per Inch" is not a size. It notates resolution and refers to the output size. Size in inches depends on what dpi resolution you specify. Multiply the pixels by the dpi you get the size. Divide the output size by the dpi you get how many pixels you need. Pixels are the fixed unit--the only fixed unit in measuring the "size" of an image. All others--dpi, inches, megabytes--vary depending on how many of those pixels you use.
The native "resolution" of my camera is 5616x3744. In pixels. On a printed page (at 300dpi, the common print resolution), that's nearly 19 inches. It's a bigger image than anyone requires. It's not generally what they get from me.
Sometimes clients ask for a file of a certain megabyte size. This gets even squishier. Consider my raw capture. More light means more data. Black areas have fewer bits than white areas. A photo of the night sky will be a smaller megabyte size than a photo of a white sheet. But the output is a bunch of numbers referring to raw pixel values and so the file size is not so big, about 20-25mb.
When I open it up in Photoshop and save it as a tiff, then there are three 8-bit depth channels of information. The file size roughly triples now, to about 60-70mb. If I open it up in 16bit it's a lot larger than that. There's more data describing the information. When I save as a jpg, a lot of the redundant data is tossed. At the highest quality setting it's now about 12mb. A photo with a lot of detail will be a larger file size. A photo with vast expanses of the same value (featureless sky) will be smaller.
The jpg file is not the original file. It's a derivative of the RAW file, which I can choose to process in many ways. The RGB pixels in a RAW file have no actual color. They're assigned a number value when you open the file in something like Photoshop. There's an algorithm that assigns values so that the data looks like a photograph and not like a linear photon value capture map. But they're just numbers, and you can assign any number you want. That's what I do when I process the pictures. It's totally non-destructive.
Jpgs are a derivative file format. Your consumer cameras give you jpgs, and you often think that's the original image. The camera takes the data from the sensor, throws away 90% of it, and gives you a jpg. Anything you subsequently do to a jpg is destructive. You're throwing away pixels. They're a convenient, compressed (even at full "resolution"--there's that misused term again), delivery format. But they're not original.
So, would you like your image small, medium or large? Tell me what you're doing with it, I'll tell you what you need. Just don't get me started on color management.