After a long wait, today I purchased Scrivener 3 and installed it in both my PC and laptop.
I keep an Excel file with my work’s progress, and thus every day I register there my word count. Today, after installing the new software and upgrading all my 4 projects, I find that there is a mismatch in word counts–in all of them.
Without exception, all projects register around 10% less words. This means over 3,000 words in one of the cases.
I don’t know what could be happening. Yes, I have tried “saving and rebuilding indexes” many times.
I opened my backup files–which I had made just prior to the upgrade–and they show the correct word counts I’ve been using for days.
Any insight would be valuable. This is driving me crazy.
Other than the word count, what is causing you to believe that you lost actual content? It’s more likely that the word count algorithm changed, or that Scrivener 3’s default options are excluding some files.
(With a uniform 10% reduction, my first guess would be an algorithm change. FWIW, in a recent dispute with an editor I found that four different word processors gave four different word counts for the same text.)
After our very extended beta period, we saw no evidence of Scrivener 3-related data loss.
Thank you for your reply. At the moment I cannot be sure if there was any content loss, but I agree that it would be extremely unlikely. Most probably my information is safe. Regarding the algorithm, I believe it should be very similar, if not the same.
I tried with different combinations of checkboxes (include only compile, etc.) and noticed no difference. I work with four different files, and all of them have the same configuration (include all documents in the word count). From the largest one at over 450,000 words to the smallest, a little over 12,500, the apparent losses–or count errors–range from 10% to 20% of the previous count totals in Scrivener 1.
I just updated the files and changed nothing afterwards.
To convince yourself that it’s a change in the count, not a change in the content, find a reasonably small document that shows a word count difference between the two versions and do a line-by-line comparison.
It sounds like you have both Scriv 1.9 and Scriv 3 installed.
If that’s the case, why not compile your entire project using 1.9 and 3.0 to docx, and the open the results in an impartial 3rd party tool, say Word, That will negate possible word count methodology differences between Scriv versions.
The counts of the 1.9 and 3.0 docx docs should match. or at least be very close (in the event the compile formats have differences in headings or chapter titles or whatever).
What concerns me is that these Scrivener documents have been built over the course of 3 years, and I don’t always write directly while using the software–I mostly write notes on the go and then paste them, one by one, at the end of the day. This means that each one has its own word count, and it has been registered in said Excel file. Theoretically, the sum of all these numbers should amount to the total in Scrivener, and yet it does not–it used to, but only in Scrivener 1, and not Scrivener 3.
Sometimes I do professional translations and I am well aware of the different word counts depending on the application used. Only this time the difference seems to be too significative.
Just for kicks, I tried the experiment myself with a large project that I converted from v1.9 to v3.0 for this purpose.
In Scrivenings mode over the Draft folder, v1.9 showed 247,468 words and v3.0 showed 246,963 = 505 “missing” words.
I compiled both projects to docx.
Word’s compare tool (thanks Katherine for the suggestion!) showed that the only differences between compiled docx files were due to chapter & scene titles, which I had in the v1.9 version compile format but hadn’t yet set up in the v3.0 version.
The important point being that the word count for content was exactly the same in both versions.
So the word count algorithms in the Scrivener versions are definitely different.
Supermarine, that may explain what you’re seeing. Then again, it’s possible you might have some conversion issue in your project. Perhaps there’s a doc with some hidden error that v3.0 dropped or something like that. (I haven’t experienced anything like that myself and can’t recall other posters mentioning it, but it’s possible.)
Hopefully after you’ve run the above process yourself, you’ll either identify where the problems are or find some peace of mind that, even if the counts are off, all your words did convert correctly.
If I recall correctly, there was a word count algorithm change between Mac Scrivener 2.x.x and 3.x.x. It had to do with how hypenated words were counted—2.x,x would count hyphenated words as two (or more) words, while 3.x.x would count a hyphenated word as one word. Might this difference also apply to Windows Scrivener 1.9.x vs 3.0?
I have not yet tried to compare the complied documents’ word count, but what I did was open both Scrivener 1 and 3 and go side by side comparing each document. What I immediately found was that some of them had a different word count, but not all. Interestingly, what did always vary was character count.
My first discovery was what Silverdragon (thank you!!!) mentioned: hyphenated words are counted as a single one in Scrivener 3, while they are divided in Scrivener 3. This means that “well-being” would be one word for Scrivener 3, and two words for Scrivener 1. “Larger-than-life” would count as three separate words for S1, but only one for S3. This detail alone is enough to account for some 3% (average) of variation, however, there were still documents without dashes (or em dashes) that had different word count. I am a big fan of em dashes, but thankfully these are still treated like spaces in S3, so they made no difference. Apostrophes are OK too.
This meant that the variance was partly caused by hyphens, but not entirely. The weird character count still bugged me; visually, both documents were the same. No, there were no extra spaces or lines–remember that these are “upgraded” documents, unmodified before or after the procedure.
So, as Jim said, it could have something to do with titles or associated comments or footnotes. Or perhaps URLs or such long, weird strings of characters, but those are seldom present in my work. I was also suspicious of parentheses, underscores, numbers, and even ellipses. These suspicions are fueled by the always differing char count; it is always ~5% off between versions, hyphens or not. Sometimes more, and closer to 10%. There are some documents–with no hyphens at all–that still show 20-30 characters less in S3.
I guess I don´t yet know Scrivener with enough depth. I have yet to find such options as “include titles in word count”, but I have looked for them and can’t find them anywhere. For the purpose of the experiment, I changed all settings to count in all documents and notes: totals versus totals.
After doing this, I still have got no definite answer. I can safely say that hyphenated words alter significantly the word count as explained above. Character count still puzzles me.
It appears we’ll have to adapt to this counting decision made by the devs. We can’t really say they’re wrong. Myself, I don’t know if compound things like “Spider-man”, “anti-multiple-hyphen”, “twenty-one-year-old” or “life-threatening” should be counted as a single word or several. I am neither a linguist nor a purist of language, but that was the standard I was accustomed to use for years. Today it is gone, but the consequences have accumulated.
As a writer, I’ve come to feel that I live and die by the word (count). As I’ve told you before, I register and measure everything and monitor my daily progress. To suddenly have 7,000 less words in my books not only makes me sad, but unsure of how to deal with it. I cannot make up numbers, and this looks a lot like an “accounting adjustment” .
Now I am asking for a different kind of help, and your opinion. What should I do? What would you do?
I can only tell you what I did. When I switched from Mac Scrivener 2.x.x to 3.0, I found the hyphenation difference (which, BTW, became more in accordance with publishers’ practices), did a compile comparison such as JimRac suggests, found that no words were missing, and shrugged.
I like to kill myself with numbers. I’m trained as an engineer, and have an emotional reaction to numbers that many lack. But having convinced myself that all the words I wrote were still there and it was only a difference in tally method, I comforted myself with the famous Mr Spock quote: “A difference that makes no difference is no difference.” I adjusted my personal output expectations to be in line with the new algorithm and moved on.
While I hope this helps, I realise that your emotional reaction may differ. Good luck.
Well, the good news is–so far at least–you’ve only lost ‘accounting’ words, no actual ones!
Thank you for reporting the results of your research. Good to know. Like you and Silverdragon, I’m a word counter. (Retired project manager) I too have a spreadsheet that tracks word output. When the time comes and I need to transition a draft-in-progress to from v1.9 to v3, I’ll figure out a way to make a one-time accounting adjustment in the spreadsheet so that the overall count aligns with v3, and then move on.
Just wanted to say thanks for confirming my suspicions. I just upgraded to v3, and saw the same “loss” of words as others have described, which made my daily spreadsheet tally a bit sad… I’d seen something similar happen with word counts when I compiled v1 documents into Word, but it was reassuring to see that others have experienced the same thing.
I feel that the hyphen should be treated just like a space when doing the word count, but that doesn’t seem to be an option…
That will negate possible word count methodology differences between Scriv versions.
