Imagine there are two transit services in an area, a low quality system (L) that is pervasive (everyone is within 400 m of a low quality stop) and a high quality system (H) that is skeletal (only a small fraction are within 400 m of a high quality stop).
Imagine there are two classes of potential users, poor people (P) who will use either system, and rich people (R) who will use only the H routes.
Poor people perceive the system as larger (both L and H) and get more network externalities from the system. They can go anywhere in town on transit. Rich people see a small system, and perceive few network externalities. They can only go places on the H system.
As a consequence, poor people are more likely to use the system than rich people.
There are several solutions to this problem. The expensive solution is to build high quality services everywhere to attract the fraction of R that would not otherwise take transit. The less expensive solution is to change the perception (and reality) of the low quality system so it appears higher quality. Give it as many of the same features of H as possible, starting with information (e.g. what bus stops at the bus stop, when does it stop, what hours does it operate, where does it go, what does the local neighborhood look like, is the bus ontime, how much does it cost) and navigability.
Why do we never consider the less expensive solution?