A table outperforms prose when the information is multi-dimensional and comparative: several items measured across the same set of shared attributes. The moment a reader has to compare three plans across price, storage, and support, prose forces them to hold every variable in their head and flip back and forth between sentences, while a table lets them scan a column and read the answer off a grid. That is the line. The pivot is not whether the content is “scannable” in the abstract, it is whether the information is genuinely comparative-structured.
Prose wins on the other side of that line, and the cases are just as clear. Narrative, causal, and single-thread explanation belong in sentences, because their value is in the connective tissue between ideas. Explaining why a redirect can dilute a signal, walking through how a process unfolds, or arguing a judgment call all depend on reasoning that a table strips out. Forcing that material into rows and columns destroys the logic that made it useful. A table is a comparison instrument, not a default layout.
This is why the blanket advice to “use tables, they’re scannable” misleads. Scannability is only a virtue when the reader’s task is to compare values across a shared dimension. Drop a paragraph of explanation into a two-column table and you have not made it scannable, you have fragmented an argument into cells that no longer connect. The test is the structure of the information itself: does it consist of items measured on common attributes, or does it depend on a thread of reasoning?
So before you reformat anything, ask whether the data is truly comparative. If you can name the items and the shared attributes you are measuring them on, build the table, because it will outperform prose for exactly that job. If the value lives in the explanation, the cause, or the story, keep it in prose and resist the urge to gridify it. Convert to a table only when the information is genuinely comparative, never as a default formatting habit.