COUNT ignores your text, so a column full of names comes back as zero. The COUNTA function is the one that counts everything a cell might hold.
Three functions sit in this family. COUNT takes numbers only, COUNTA takes anything that is not empty, and COUNTBLANK takes the gaps. Between them they ought to add up to the row count.
Response rates, stock takes and import checks are all the same question underneath: how many rows actually have something in them. One formula answers it, so long as you know what the function calls something.
The COUNTA function.
Point it at a range and read the count back.
=COUNTA(A2:A20)
Text, numbers, dates, errors and logical values all count. In short, the COUNTA function asks only whether the cell is empty, never what kind of thing is sitting in it.
The contrast is one letter away. =COUNT(A2:A20) counts numbers alone, so a column mixing names with quantities gives two very different answers depending on which you typed.
An empty-looking cell is not empty.
Here is where the arithmetic stops matching the screen. A formula that returns "" leaves a cell that looks blank and is not.
=COUNTBLANK(A2:A6)
COUNTA counts that cell because it holds a formula. COUNTBLANK skips it for the same reason. Consequently the two functions agree with each other and both disagree with your eyes, since two rows look empty and only one of them is counted as a gap.
This is exactly what IFERROR leaves behind when you hand it "" as the fallback. So a cleaned-up column and a genuinely sparse one look identical while counting quite differently.
Result of the COUNTA function.
Before. Five rows, two of which look empty, and nothing counted yet.
After. COUNTA says four and COUNTBLANK says one, over those same five rows.
A2 -> Widget A3 -> 42 A4 -> (genuinely empty) A5 -> ="" (a formula returning nothing) A6 -> Gadget =COUNTA(A2:A6) -> 4 =COUNTBLANK(A2:A6) -> 1 =COUNT(A2:A6) -> 1
Four plus one is five, which matches the row count and is still not what the sheet shows. In fact two of those rows read as empty, and only one of them ever was.
The cell that fooled the count usually arrives from IFERROR, so that is where to look first.