I find it useful to understand the underlying tools. These are cl.exe (compiler) and link.exe (linker). You need to tell the compiler the signatures of the functions you want to call in the dynamic library (by including the library's header) and you need to tell the linker what the library is called and how to call it (by including the "implib" or import library).
This is roughly the same process gcc uses for linking to dynamic libraries on *nix, only the library object file differs.
Knowing the underlying tools means you can more quickly find the appropriate settings in the IDE and allows you to check that the commandlines generated are correct.
Say A.exe depends B.dll. You need to include B's header in A.cpp (#include "B.h") then compile and link with B.lib:
cl A.cpp /c /EHsc
link A.obj B.lib
The first line generates A.obj, the second generates A.exe. The /c flag tells cl not to link and /EHsc specifies what kind of C++ exception handling the binary should use (there's no default, so you have to specify something).
If you don't specify /c cl will call link for you. You can use the /link flag to specify additional arguments to link and do it all at once if you like:
cl A.cpp /EHsc /link B.lib
If B.lib is not on the INCLUDE path you can give a relative or absolute path to it or add its parent directory to your include path with the /I flag.
If you're calling from cygwin (as I do) replace the forward slashes with dashes.
If you write #pragma comment(lib, "B.lib") in A.cpp you're just telling the compiler to leave a comment in A.obj telling the linker to link to B.lib. It's equivalent to specifying B.lib on the link commandline.