#cross platform comparison

Update: 

A clarifying note from @elizabethminkel about @fansplaining‘s comments:

“When we discussed Wattpad turning away from fanfiction, we were specifically talking about the platform, not its users. (I don’t think you made that assumption having listened, but I don’t know if it’s super clear from your description.) Wattpad has said things along the lines of, ‘Fanfiction has only ever been about 20% of our platform”—which it looks like it continues to be! So the interesting question here is how a platform continues to scale with user-generated content when a solid portion of that content has kind of maxed out monetarily. Like, more and more fic will be published on the site, but that doesn’t mean more and more profit for Wattpad or its users—which is a problem within the tech industry’s current models.“  

She also has some excellent observations about a bunch of Wattpad content not being very fic-like, which I’ll hopefully dive into more soon -- thanks, Elizabeth! :)

/Update

Which site is growing fastest?

FFN is the oldest site, but growth on the platform slowed and then started decreasing somewhere around 2012 (based on my samples and Finn’s work).  I.e., there are still works being posted each year, but at a decreasing rate.

Meanwhile, both AO3 (founded 2007) and the “Fanfiction” genre on Wattpad (founded 2006) are growing fast; posting rates surpassed that of FFN in 2015 and 2014, respectively.  Wattpad appears to be growing fastest.  As of November: 

Wattpad’s “Fanfiction” genre received at estimated 1.6M new works so far in 2018, which means I’d estimate (very) roughly that there could be more like 2.5M new fanworks total, including ones in other genres. But that’s based on my hand labeling a sample of 100 works from various years, and it assumes that the ratio of unlabeled:labeled fanfic on the site has stayed roughly the same over the years.  That assumption may not be right.

AO3 has 0.9M new works so far in 2018.  

FFN has 0.3M new works so far in 2018.  

Based on past patterns, I predict there will be big surges in production rate in AO3 December due to annual holiday breaks and gift exchanges -- it looks like this seasonal surge occurs more dramatically on AO3 than on FFN, though it happens on both.  I don’t know about seasonal production patterns on Wattpad.

Which site ends up with the most deleted or unpublished works?

Update: With huge thanks to @zz9pzza for clarifications about AO3 and examples -- I’ve rephrased most of the following from them:  As of a system update around 4 years ago, AO3 only assigns 1/3 of possible story IDs to actual stories (to avoid numbering collisions), which means that most of the “missing” story IDs were never actually assigned. For instance, the following sequence of story IDs would be assigned: [16686808, 16686811, 16686814, 16686817, 16686820, 16686823].  Some quick estimates lead to thinking AO3 may have more like a 14% deletion rate -- far lower than what I show in terms of missing story IDs.

Do the other platforms also have similar explanations for their large missing ID rates?  Not necessarily -- FFN does not appear to be doing a similar thing currently (Nov 2018); there are recent sequences of several ID numbers in a row corresponding to actual stories (e.g.: 3119884, 3119885, 3119886, 3119887).  Wattpad also does not appear to be skipping IDs, as I found some pairs of IDs in a row, and recent sequences of story IDs with very few missing (e.g., 159620101, 159620102, 159620104, 159620108, 159620110).  So while these platforms may not have assigned all possible IDs, but I can’t detect any regular pattern to what’s missing, as is the case for AO3.

I updated the slide up at the top, but in case you clicked through from an older reblog, here it is again:

 /Update

AO3 and Wattpad both assign a new story ID to each new draft of a work, meaning that all unpublished works have URLs that don’t correspond to stories (edit: but as mentioned in the update above, AO3 doesn’t assign all possible IDs to stories).  FFN only assigns a new story ID at the time that a new work is published. All platforms also end up with some published works being deleted, either by the author or the platform (works can be deleted for being spam or violating TOS -- e.g., being explicit on FFN).  Thanks to Finn for this info -- see their post for more details.

As expected from the fact that it doesn’t assign story IDs for unpublished works, FFN has the lowest missing work rate fewer missing story IDs than Wattpad -- but it’s all due to deletions of previously published works (some done by FFN, which has done a number of mass deletions due to TOS changes, and some presumably by the works’ authors)  Wattpad has the highest missing story ID rate, but it’s unclear what that indicates. Many of these Wattpad works could be drafts that haven’t made it out of draft form yet. And some (maybe a lot) of the IDs may never have been assigned to a story in the first place; that is a side effect of some methods of database construction.   It also seems (based on notes in the summaries and titles) that on Wattpad it’s very common for authors to revise works that have already been published, and it’s possible that many authors use Wattpad’s “Unpublish” option to temporarily revert existing works to drafts.  I also found a few cases of spam/advertisements in the sample of 100 fanworks I hand classified, so possibly Wattpad has such a high missing works rate in part due to spam takedowns.

Detailed methods

I used AO3 Work Search to determine the exact numbers for AO3.  

All sites assign higher story IDs (the numbers found in URLs) to more recent works, so on the remaining two sites, I found the highest newly published work ID I could and used it as a maximum.  For Wattpad, the max story ID was about 160M.  For FFN, it was about 13M.  I sampled 9000 URLs on Wattpad, of which 1176 had stories.  I sampled 3500 URLs on Fanfiction.net, of which 2048 had stories.  (Feel free to use the data for your own analyses.)  

(Aside:  more recently, I’ve seen a few MUCH higher story IDs on Wattpad -- closer to 650M.  So I did several samples of 1000-2000 URLs using that higher number, and I couldn’t find any stories with IDs over 167M; it seems like there are very few with the much higher numbers.  I also double checked my belief that story IDs were assigned in order that drafts were created by graphing date published against story ID and found it to be accurate -- there were a few stories published long after their story IDs were assigned, but generally there was an increase in story ID by date published.)

In both cases, I drew the samples in batches of 500-2000 fanworks at a time, and I averaged the estimates I got from each subsample.  For Wattpad, I got a mean estimate of 20.83M fanworks overall (stdev = 1.42M; stderr = 0.50M) and 5.11M works in the “Fanfiction” genre (stdev = 0.60M; stderr = 0.19M).  For FFN, I got a mean estimate of 7.65M fanworks (stdev = 0.19M; stderr = 0.08M).  

I found a kludgey way to search FFN that I *think* returned most/all of the works on the site.  The results contained 7.60M works (close to my estimate of 7.65M), so that also strengthens my confidence in my estimates.