1) A hook, preferably one with a sense of intrigue - some kind of mystery or exploration of unknown territory.  This is most important for me.  Something with page after page of blather about NPC plans and schemes but no real intrigue is going right back on the shelf.

2) Danger.  I prefer that this be unbalanced (i.e., some areas are more dangerous - possibly much more dangerous - than others).  Ideally smart and/or lucky players should be able to avoid the most dangerous areas through caution and good dice rolls.

3) Rewards.  Similarly to danger, I prefer that these be unbalanced - some areas offer no or paltry rewards, while others offer the mother lode.  Again, smart players should be able to locate some of the choice loot (or lucky players stumble across it).  There needn't always be a direct correlation between the level of danger and reward in every area.

4) Irony, specifically taking the form of being hoist by one's own petard.  I enjoy providing plenty of opportunity for players to cut their own throats (or see it happen to NPCs).

Some good suggestions in this thread, and I'd echo all of it.

The only thing i'd add is, a good adventure should transport the players away from the kitchen table and into the fantastic. Ideally it should have some really vivid imagination that open their eyes a little.

I normally judge published adventure modules by the following standards:

Atmosphere: It has to capture my imagination by having a promising setup, a clever hook, and/or inspired ideas. For instance, Anomalous Subsurface Environment has many cool ideas that look promising at the table, and it is dedicated to deliver its vision consistently. Another great module that sets out to present a particular atmosphere would be Barrowmaze, which is much more ordinary, nevertheless its vision is consistent.

Editing: The second most important thing for me is how easily usable the product is at the table. Dwimmermount is a prime example of superb execution regarding this; although sometimes the room descriptions may be just a bit too long (particularly for more complex areas), all the relevant information is there or referenced by page number or bookmarks in the PDF edition. Also, a few chapters are dedicated to collect generally useful information (such as how factions interact, or how the architecture differes from area to area and how they telegraph information about its denizens). Rappan Athuk comes to mind as a really poorly executed product concerning editing; it has really great ideas, but the big picture is only visible once you read through all of it. The lack of summarised information on various topics combined with layout issues made it really hard for me to run it effectively.

And... actually that's it. I also don't like railroads and whatnot, but that's rarely an issue in OSR modules anyway.