Word count versus Microsoft Word

I know this issue has been discussed on forums before but can anyone explain the difference in word counts I’m getting? Two in Scrivenver and one in Word.

Project statistics: 88,041

Project target: 87,783

Compiled Word document: 86,873

These look like acceptable differences to me. There are going to be minor variations in how most programs count words. Whether they consider hyphenated compound words to be one or several, for example, is a common point of difference.

Part of the Session Target vs. Project Statistics discrpency is probably due to some of these factors, assuming I’m correct that the Statistics does a background compile:

  • Automatic chapter numbering - Every chapter that gets the automatically generated “CHAPTER ONE” is going to increase your word count by 2, but only at compile time or when checking Project Statitics (PS)
  • Document titles that are included in the compile will also add some words to the output, reflected in the PS, but not the Targets window.
  • Include in Compile checkbox - If you de-select this box on any of your documents, and have the Project Targets “Documents included in compile only” box also un-checked, then this will increase the discrepancy. You can quickly check for the state of the “include in compile” checkbox on all of your documents by going to Compile->Contents; the Include column indicates if the document will be compiled, and therefore if it is counted in PS.
  • Replacements (in the Compile window), could also expand the number of words in your document

Related question — it seems like Scrivener itself counts differently in different situations (forgive me if this has been covered elsewhere).

Looking at the bottom of Scrivener’s main screen, it says my word count is 149.058
Exporting it to PDF, the <$WC100> string is converted to 148.400
And exporting it to EPUB, the same string is converted to 149.300

This can’t really be because different program counts words differently, since the number is “written” by Scrivener into the document at compile time.
How come Scrivener counts the words differently depending on export format?

And most importantly: Should I worry? It’s obviously a bit discomforting to see the number of words mismatch between exports — have something been lost? Has some hidden annotation become visible? Is there in fact a discrepancy between the documents or is it just the numbers…?

Love to hear your thoughts!



I think Robert summed up a number of the things which can cause discrepancies, well. The compile can make drastic modifications to the word count that is seen in the editor, but it is the minor ones like chapter names and whether or not comments are included that can make subtle differences. The main issue is that some areas of the interface must compromise accuracy for speed. The footer bar when you are typing in a document (or a sequence of them in Scrivenings) must be real-time, and for that reason it does not use compile settings to determine if various texts are actually a part of the final draft or not. Sections that have been excluded from compilation will still be counted, inline notation of all sorts will be counted, and so on. The same is true for the Project Targets tool, which also counts your typing in real-time, just on a larger scale. Project Statistics will in general be the most accurate count, because it does a background compile to calculate those numbers (and that is why it can take longer to pull up when working in a large project).

I get the real-time discrepancy, that makes sense.

My worry comes from using the same compile settings, just choosing different types in the bottom dropdown menu: Compile for Print, Epub, Word.

For some reason, with the exact same compile settings, changing only the target format will change the word count.Since it’s scrivener converting the string to numbers, that puzzles me — all other settings but format being the same…

E-books in particular will modify the total word count, depending upon the length of the work, as they generate a table of contents automatically, unlike the standard document formats.

That begs the question then; should automatically generated pages like a ToC be counted for that token? For that matter, should anything in the Front Matter folder count? What about page headers… do they add words to the total for <$WC>?

It’s a valid point, the ToC generally wouldn’t be considered countable, and perhaps that should be included in the (not yet for Windows) “Do not include front matter in word and character counts” option in the Statistics compile option pane.

Headers should not be counted under any circumstances.

Guys, thanks so much for your thoughts, especially Robert.

This issue still worries me. I have 35 chapters, which means the two words from the chapter headings isn’t material. I have checked and re-checked the compile list to make sure all the content is being compiled.

That still leaves me 1000 words short in the version I have to send to my publisher. I may be wrong here but that seems too much to be due to hyphenated words.

It worries me greatly that somehow some content isn’t ending up in Word…