What is the point in publishing a paper in a workshop rather than in a conference?

In my research area (software engineering), points why people submit to workshops include the following:

Lower barrier for acceptance: Some works are inherently hard to get accepted at top-notch venues, for example, because they are hard to evaluate properly. A paper that was rejected multiple times at higher-level venues might eventually end up at a workshop, so that the authors can still make a (albeit less valuable) publication out of it.

Publication strategies: Some researchers actively try to build a prolific track record of many small (including workshop-level) publications, rather than a few big ones. While it's debatable whether such strategy is best for advancing science, and even for convincing a selection committee, there are certainly some associated benefits, for example, plenty of opportunity for self-citation. Strategic publications may also be used to show an association between two authors (to convince a grant committee that they can work together well enough to write a paper), or to establish an association between the authors and the topic (to show a grant committee that they have some early work in the field).

Getting feedback: Sometimes, people have a nice idea, but are stuck in an early stage of implementing this idea. Workshops are a perfect opportunity for gathering feedback, since they (i) are designed to enable a focused discussion (small audience, specific focus), (ii) usually have a faster review cycle than conferences, and (iii) sometimes allow for short discussion or problem statement papers.

Opportunity to visit the host conference: Workshops are usually embedded into the ecosystem of a larger host conference or conference series. When a researcher wants to visit the host conference, but didn't get a paper accepted there, a workshop paper can be their entry ticket. (It's usually easier to get a conference trip funded if you have a paper at the conference or one of its workshops.)

Warm-up publishing: When joining a new field, a group of authors might want to get accustomed to the associated community to understand their culture, values and standards. Workshops are particularly suitable for this purpose due to high acceptance rates, interactive formats and the opportunity to visit the host conference.

Workshops are specialized and you can meet the "gurus" in your field. This is important for networking and also for very good feedback/discussions etc. Notice also that some workshops (at least in TCS) are at least as highly ranked as high ranked and prestigious conferences.

Basically, at least from the quality point of view, it does not mater a priori is a venue is classified as Symposium, Conference or Workshop. It depends on the quality of papers in attracts and what is your goal. Do you want your work to be visible in the specialized community? Or to be put in a wider spectrum?

Personally if given the choice between two equally ranked venues, one being specialized and the other being general, I would prefer the former for the reasons I outlined above. This also guarantees better visibility (because at the end of the day, only people close to your area would care about the result unless is really some huge breakthrough).

I'm sure this is extremely dependent on field and perhaps specialization within that field. To give you an example, in Computer Science, there are lots of workshops, of varying competitiveness. Within the systems discipline of CS, most major conferences have at least one workshop that can be viewed as a "feeder": people publish early work, which appear formally as proceedings, but which are much shorter than conference papers. For instance, most limit papers to 5 pages, while conferences might be 12-14. So people publish the early work in the workshop, then expand on it for the major conference. The workshops are smaller (usually no more than about 100 attendees) and more interactive, providing lots of useful feedback.

Over the years the set of workshops on Hot topics in XXX has jumped from a couple (like Hot Chips and then Hot Operating Systems - HotOS) to dozens. A few are standalone, like HotOS, which even limits the attendees to those who have position papers accepted. Many are tied to a conference. A good example is the USENIX Hot Topics on Storage (HotStorage), which has taken place before the broader Annual Technical Conference each summer and often leads to longer papers submitted to USENIX File and Storage Technologies in the fall.

In other words, in some disciplines it's expected and useful to publish early work in a workshop and later work in a conference. Judging from other comments, there may be other fields where it's only OK if the workshop has no formal proceedings.