One source of the derailment of eLearning has been the prevalence of "solutions looking for problems", i.e. vendor and committee products designed to capture the market. The eLearning decision makers have been driven into purchasing expensive LMS systems without a clear concept of what they are to achieve and without a single course to deliver. I see it over and over where eLearning consumers tell us that they're not yet interested in purchasing an authoring system because they are implementing their LMS (usually a 2-3 year process). This is like building a highway system without ever having owned a car or truck. (A simple web site could actually meet all their initial needs.)
Many corporate issues can be solved by basic Web 1.0 techniques. Course content should be reachable at any time. It should be searchable. This way, when a worker needs to look up some definition, they can go to the "training" web site, do a search on the term, and see the multiple courses that relate to that content. Then, they can jump to the page(s) with the content. All this in a couple of minutes. If the course authors want tracking, that can easily be done without any fancy expensive LMS system.
From what I have seen, the approach that the analysts and much of the community encourage is what we could call a "Web 0.5" or a pre-web approach. This involves simulating the face-to-face training as closely as possible:
1. Content consists of presentation slides converted to some web-deliverable format without any consideration of instructional effectiveness.
2. Student sessions have a fixed start and a fixed finish, with a single test session (one question per page) at the end that can be passed or failed. Random access to the middle of the content is discouraged.
3. The objectives from the instructor's point of view are to deliver the material with the least possible change and to track (maybe) that the student has "completed" their training.
Until this paradigm can be replaced with a more web-like approach, we'll be stuck with federally mandated obsolescent specifications like SCORM, and ROI for eLearning will consist mostly of empire building by the managers who purchase these expensive systems.