I know that ID questions are very contentious on different sites - some sites love them and others hate them.
I believe that they can be successful here if we set some strict ground rules to make the questions less of a "guessing game" and more of an actual, answerable question.
So, to help them be successful here, what requirements should we have for these questions?
There are some really great examples of ID questions that have been posted already, even in our short time here... and there are a couple that I don't really think work.
The two existing questions here are both great examples:
What makes them great?
Without an image, there's little chance we could ID this item. Even in the case of the item that the OP didn't own, they did research to find an image of exactly what they were talking about so that it could be identified.
This question doesn't have an ID tag... though it probably should... maybe technique-identification?
But, again, this fits the basic needs to make this an answerable question...
The question has an image of the exact technique they are talking about. It specifies the type of craft it's used for, and it's obvious that the question is art/craft-related.
Now, a question I think is problematic for our ID requirements:
What are these balls of wool bound together with a rubber band?
This question does have an image but the OP makes it clear that this image is not of the actual item being described, it's only an approximation:
Note: this was the closest thing I could find
So we have no way of knowing what the actual item looks like. Many items look similar but have vastly different uses.
The OP also gives no indication that this item is related to arts or crafts in the first place. It wasn't part of an art kit, it wasn't sold in a crafts store... it's a generic "what is this thing" question.
So, now that we've looked at some examples, what guidelines should we use:
There may be some cases where images are unnecessary but I think that we should consider requiring them initially and make exceptions if it becomes clear that good-quality questions can be asked without them.
ID questions can be very fun for the site but they can also get to be extremely problematic if not properly managed, so I want to encourage us to limit what ID questions we accept to reduce the chances of an influx of "guessing game" style questions.
The criteria laid out in Catija's answer are great, and I already upvoted it. But one thing she doesn't address is what we should do with ID questions that don't match these conditions.
My suggestion is: don't crack down hard on these questions unless they become a problem.
Some sites (like Movies & TV, with which Catija is familiar) have a long-standing problem with identification questions, to the extent that they quickly vote to close any which don't meet their standards; this is a necessary measure to take just in order to try and stem the tide of daily terrible ID questions.
Other sites (like Puzzling, with which I'm familiar) don't get many identification questions and haven't needed to impose any particular conditions on them; none have had to be closed so far.
Science Fiction & Fantasy (where I'm a moderator) does get a ton of identification questions, but is lucky enough not to have a serious quality problem. We have a meta post to describe what the ideal story-ID question should look like, but we don't need to crack down on all the questions that don't satisfy this ideal. We downvote and sometimes close those that are way too broad, but this is a relatively rare occurrence compared to at e.g. Movies & TV.
Until we know which of these categories Arts & Crafts is going to fall into, I suggest that we encourage all ID questions to satisfy Catija's conditions, but not move to close all those that don't. If someone posts a question without a clear enough picture or description of what they need identified, we leave a comment asking them to add this, and maybe downvote the question; but we don't need to close such questions except in very clear cases, e.g. if the item described seems to have nothing to do with arts and crafts or is so poorly described that it could be almost anything.
Let's err on the liberal side until we know what we're going to be dealing with. If we end up getting lots of bad ID questions, we can revisit this issue and decide on a more draconian policy then.
To complement Catija's excellent concrete example-based answer, here's a more abstract and theoretical (but also much shorter) answer.
Since my opinion on id questions in general has changed very little over time, I'd like to quote a portion of one of my SFF meta answers.
In my experience, the acceptability of ID questions on a given site comes from the natural quality level that ID questions have there when they're allowed without restriction, and this is a product of 1) how their enthusiasts stumble across things they want identified, and 2) whether their enthusiasts remember the kinds of details that help distinguish one work from another.
I then went on to argue that this helps explain why Anime & Manga gets such bad id questions, SFF.SE naturally gets mostly decent ones and Arqade was able to restrict id questions in such a way that now they're mostly decent.
Of course there's no way to know for certain whether this site will naturally get mostly good or mostly bad id questions. My instinct is that "tool identification" will do fairly well without much help, since the people who need tools identified are far more likely to be self-motivated experts or enthusiasts with a modicum of prior knowledge rather than someone rushing to complete a homework assignment. For more or less the same reason, I suspect "arts-y/craft-y object identification" will do far worse if left alone, but if we require images it should be okay since this is a subject where most useful objects are going to be visually identifiable.
However, until we know for sure how good or bad "natural" id questions are, taking the middle ground of requiring all types of id questions to have images seems like the safest bet. By allowing id questions at all, we get some meaningful data to make better decisions later. By formally requiring at least a little upfront effort to get the picture, we're making it easy to quickly nuke them if they do turn out to be bad. And by using the same rule for all id questions, we prevent any confusion should the categories of id questions that get asked in practice change drastically as the site grows.
