A word search fails in one of two ways. It is too easy, which is a construction failure and can be fixed with a script. Or it is unpleasant to look at for twenty minutes, which is a typographic failure and cannot.
We spent the better part of a year on the second problem. The finding, stated plainly: letter-spacing in a grid does more for solvability than grid size, entry count, or difficulty rating combined.
The tracking threshold
Below roughly 0.08em, adjacent letters in a row begin to read as words that are not there. The eye assembles false positives and the solver loses confidence. Above 0.18em, the grid stops reading as a field and starts reading as a table — rows separate, and diagonal entries become nearly invisible.
The grid has to feel like a surface, not a spreadsheet. Everything else is downstream of that.
Our production setting is 0.12em on a 20 x 20 field, at 16pt. It is not a universal answer. It is the answer for a cream stock read at arm’s length under a kitchen light, which is the condition we design for.
What we abandoned
Shaded grids. Alternating row tints. Any form of border between cells. All three were tested and all three measurably slowed solvers down while making the page look busier — a rare case where the aesthetic argument and the functional argument agreed.