What is Screen-Space Error

Hey guys, I have a question regarding screen-space error. There are a couple of resources out there that explain the topic, however I feel a little dumb as I can't understand exactly how it works. Perhaps somebody can help me clarify this issue?

Let's say I have a procedurally generated plane from a height map. It has a few heights in it. This plane will subdivide into 4 children when the camera is too close based on screen space error calculations. When I want to calculate the screen space error, do I calculate on the existing vertices? For example: On a tile, I have 4 vertices, two triangles. Vertices from the bottom left are a, b, c and d. What I understand is that I must project the vertices to the screen and calculate their screen counterparts in pairs. a-d and b-c. What i get is a pixel "distance". Is that what needs to be taken under consideration as pixel error? Do I need to do this for each tile in my mesh? And if so, wouldn't that cost a lot on CPU?

Can somebody please tell me if I am on the right course? I have been at it for a few weeks, however I can't seem to wrap my head around it. Can anybody help me shed some light into it?

Thanks, Billy