I believe there are two criteria that are central to judging the quantity and quality of information in a grid.
First would be the amount of value and interest in each grid entry. This is tricky to define in a purely objective way. One approach would be to imagine a scale with "Commonplace & Simple" at the low end and "Unique & Complex" at the high end.
Consider, for example, two 8-letter slots in the grid, one filled with SINECURE and the other with HE ATE ONE: Same amount of matrix space filled but with very different ratings on the value/interest scale. You could write an essay on the meaning of the first entry, while the second entry is as simple, commonplace and uninteresting as it gets.
Second, an ideal grid would have only base words and phrases . These are words or phrases as they first appear alphabetically in dictionaries and lexicons They cannot be reduced further (have any letters removed) without losing their meaning, having their meaning completely changed, or losing their status as a stand alone words/phrases, e.g., becoming an abbreviation.
BATTLE would be an example. Remove any letters and the meaning is gone. Add letters, however, and it is no longer a base word. BATTLES, BATTLED, BATTLER, BATTLERS or BATTLING all fill up more of the limited matrix space without adding much of value or interest.
This is the hallmark of Letter Count Inflation (LCI, discussed elsewhere on this blog). It's largely wasted space, wasted opportunity. It's like padding an essay to meet a minimum word count requirement, kind of like adding non-nutritive filler, so to speak, to increase volume without increasing value. The more LCI in a grid, the farther away it gets from the ideal crossword puzzle.
Crossword construction constraints such as symmetry, no two-letter entries, black square maximums, theme entries, etc., make it challenging to achieve the ideal grid fill. The quality of a grid hinges on how close or far away it is from this ideal. One that comes as close as possible would be a grid with the least amount of wasted space and the most interesting fill, one that makes optimum use of the limited space information display matrix.
