Before you record your own actions, it's a good idea to write a rough outline of the steps you're going to record, and refer to the list as you record. Although you can edit actions later, recording too many mistakes makes it harder to edit the action because you'll have to figure out which steps to keep and which to delete or edit.
In my example, I want to:
Open a document.
Resize the image to my desired pixel dimensions for the Web. I don't use Image > Image Size to do this, because the Image Size command doesn't treat horizontal and vertical images the same way. However, the File > Automate > Fit Image command does, so that's my answer.
Sharpen the image at an amount that looks good on the Web. I use the Filter > Sharpen > Smart Sharpen command for that.
Save the final image as a JPEG document at specific optimization settings. I want the image to include metadata; File > Save for Web doesn't do that but File > Save As does, so I plan to use the Save As command to create the JPEG.
You can see why I'd like to create an action out of this sequence, and I use it as the example in the next section, "Creating Actions." Although I'm perfectly capable of performing all those steps, I may not remember to do all of them at the correct settings every single time. With an action, I know they will all be done, precisely, every time. Actions aren't just about time savings and convenience; they're also about precise repeatability.
Many actions are built to create effects, but an action may not always be the best way to create an effect. In some cases, your goal may be served better by using a feature such as a preset or a layer style. For example, a layer style is reversible.