AO3 News

Site Stats: A Look at 2013 and Beyond, Part 2

Published: 2014-02-14 22:16:59 -0500

In our last post, we talked about trends in posting, commenting, and bookmarking activity, as well as page view numbers in 2013. We now want to look at some geographical stats and browser usage. As before, raw data is available in spreadsheet form:

» AO3 Stats - 2013 (on Google Drive)

 

More international usage!

Sadly, it requires a bit of work to generate usable chunks of data from our massive server logs, so we only started keeping track of some international stats last year - with the help of our aforementioned server monitoring tool, New Relic. We aim to improve our data this year, and will be able to provide a closer look at trends in international usage farther into 2014.

One metric we can use to compare site usage between different countries is the number of "average sessions" over a period of time. A session starts when someone (i.e. an IP address) opens an Archive page, and ends after five minutes (unless that same IP address keeps browsing the Archive). If several people visit the Archive at the same time, that's a number of "simultaneous sessions", and we can track of how many of these we have, on average, at any given time.

Partial screenshot showing world map interface in New Relic. The US are tinted a dark blue, indicating heaviest site usage. Almost all other countries across the world are a lighter blue. Only a few countries in Africa are white, indicating no data.
Image: World map showing countries with recorded site usage in blue. Screenshot was taken on July 17, 2013.

At the end of July, we stacked up our Top 20 countries by comparing how many average simultaneous sessions we had from each country during the preceding three months.

Country Average Sessions
United States 6,510.00
United Kingdom 1,130.00
Canada 809.00
Australia 556.00
Germany 348.00
France 171.00
Sweden 134.00
Italy 123.00
Russian Federation 98.60
Finland 97.10
Netherlands 93.00
Philippines 92.50
Singapore 86.80
New Zealand 85.80
Brazil 78.20
Poland 77.30
Mexico 70.30
Spain 69.90
Ireland 64.80
Denmark 58.20

On average, roughly 6,500 people from the United States were browsing the Archive at the same time, and roughly 90 people from the Philippines. As our traffic increased during the year, we noted growing numbers for each country as well. All data is available in the 'countries' tab in the AO3 Stats spreadsheet.

We've also noticed a slight uptick in Support tickets in languages other than English. With the help of our Translation volunteers, we've replied to requests in Chinese, Russian, Indonesian, Spanish, German, Portuguese, and French last year. You can always find our list of officially available Support languages at the top of our contact form!

One of our next code deploys will include changes to the FAQ system that will make it possible to offer translated versions of our FAQs. We're very excited to roll out this feature, and will post more information about language features on the Archive closer to the deploy date.

 

More mobile browsing!

Chrome on Windows remains the most popular browser among our users, by a wide margin. However, it's immediately followed by Safari on iPhone, which is in turn followed by a number of other mobile platforms among the usual suspects, Firefox and Internet Explorer. From user feedback (and our own Archive use!) we know that browsing the AO3 on phones or tablets is becoming increasingly popular. (Some numbers can be found in the 'browsers' tab in the AO3 Stats spreadsheet.)

While we do our best to make the Archive accessible and easy on the eyes on smaller screens, we frequently receive requests for a dedicated AO3 app. We've explained why we won't be able to offer iOS and Android apps anytime soon and continue to follow up on the topic in relevant posts on our Tumblr. In short: designing, coding, and maintaining even one app (and providing ongoing tech support for that app) is an incredible time commitment, and we've already got our hands quite full with the Archive website.

We have several improvements to our download feature coming up, and will continue to provide EPUB and MOBI versions of all fanworks on the Archive. Just use an e-reader app (such as iBooks or Aldiko) to organize and access your downloaded files, and you'll always have things to read even when you're offline!

 

More changes under the hood!

2013 was the year of many "invisible" efforts to keep the Archive stable and easier to maintain going forward. We spent a lot of time getting all our code adjusted for our big Ruby on Rails upgrade, and even more time updating, fixing, and improving our automated tests for use with Travis-CI over the year. All in all, we closed 300 issues in our public bug tracker. Together with our Systems team, we installed new servers and made changes to our server architecture in preparation for bigger and better things in 2014.

Unseen to AO3 users, we spent several weeks reviewing our internal knowledge base (removing nearly half of the 350 wiki pages for being outdated), organizing pertinent documentation and training materials, and making sure all our shared spaces for discussion and collaboration were in order and archived items were easy to find. Yay!

 

This concludes our look back; more posts about exciting celebrations will follow shortly! Many, many heartfelt thanks to everyone who has left comments of encouragement and support either here or on Tumblr or Twitter. As always, if you have any questions or comments, just let us know!

Comment

Site Stats: A Look at 2013 and Beyond, Part 1

Published: 2014-01-27 18:38:38 -0500

2013 was another record year for the Archive of Our Own in terms of site growth and traffic increases, and we'd like to highlight some figures in this post (and the following). All raw data used for these charts is available in spreadsheet form, if you want to crunch the numbers or create some graphs yourself.

» AO3 Stats - 2013 (on Google Drive)

 

More works! More kudos! More everything!

We started the year 2009 with 182 users and a little over 2,000 works, a few of which probably said something like, "dskjdlj test test," or variations thereof. The Archive code was still being written by a small group of volunteers, who created a framework for posting, browsing, and commenting on fannish works entirely from scratch. It was considered stable enough for the general public in November 2009, when we kicked off our Open Beta phase. By January 2010, we had over 4,200 users and 36,500 works.

In 2013, we went from about 103,000 accounts to 247,000, and we passed the quarter-million mark earlier this month. (Excitement!) Roughly 416,000 new works were posted to the Archive last year, which is almost half of all our current works. (And for those reaching for the calculator right now, that's about 1,140 new works per day, on average.)

The really staggering increases, however, happened where users were interacting with each other: leaving kudos, talking in the comments, adding works to their bookmarks and recs.

 

Month-by-month growth of kudos, bookmarks, comments, works, and users for the year 2013. The high numbers and steeper increases for kudos, bookmarks, and comments dominate the chart, with the lines for works and users seemingly crawling along the bottom. All numbers can be found in the 'works & users' tab in the linked Google Drive spreadsheet.

 

We went from roughly 2 million comments in our database to almost 5.5 million, and from 10 million kudos to 28 million. (That number is one of the reasons we can't offer a feature that would give you a list of all works you've left kudos on: there's just too much data to make those calculations for each individual user while keeping the servers happy.)

What really took off in 2013 were bookmarks: 1.8 million at the beginning of 2013, almost 7 million by the end.

As it turned out, our search index code was not prepared for this amount of data at this level of user activity, and slowly broke down towards the end of last year. We are currently rewriting the relevant part of our code, making use of new tools at our disposal, for a more streamlined indexing process.

To better illustrate trends in all these numbers, here's a chart showing relative bookmark, kudos, comment, user, and work growth in 2013. (Increases are shown in percent: for example, going from 50 works to 100 would be a 100% increase, and going from 100 comments to 150 would be a 50% increase.)

 

Month-by-month relative growth of bookmarks, kudos, comments, works, and users for the year 2013 (in percent, starting on January 1, 2013). Bookmarks show the steepest increase (277%), users the lowest (77%). Kudos and comments show an almost identical growth, even though they differ in absolute numbers. All numbers can be found in the 'works & users' tab in the linked Google Drive spreadsheet.

 

Over the last twelve months, kudos and comment numbers almost doubled (both increasing by roughly 190 percent), while the number of bookmarks almost tripled (280% percent increase).

One last stat we looked at were average feedback numbers per work, given the total number of works.

 

Month-by-month growth in the average number of kudos/comments/bookmarks per work. The numbers were calculated for every month, based on data in the 'works & users' tab in the linked Google Drive spreadsheet.

 

As you can see, these numbers also went up over the year. We started out with 3.6 comments, 3.4 bookmarks, and 18 kudos per work, on average. This was followed by a relatively slow increase in works, and a much steeper increase in feedback numbers. By the end of 2013, we registered 5.8 comments, 7.3 bookmarks, and 30 kudos per work. (Of course, all these are averages across almost a million works, so a relative handful of very widely-read works in popular fandoms will drive up the numbers by quite a bit.)

 

More traffic!

One way to measure site activity is by looking at works and comments, another is through reports from our server monitoring tool, New Relic. Among many other things, it keeps track of how many pages were generated and served to our users at any given time.

We started the last year with 27.6 million page views in the first week of January 2013, which comes out to roughly 3.9 million page views a day, or 2,700 page views a minute. That's on average, of course. Sundays will be much busier than any weekday, and there are peak times and slower periods throughout the day.

In the first week of January 2014, we counted 49.5 million page views. At this rate, a Sunday with 7 million page views isn't a rarity anymore. That's an average of 4,800 pages served to users every minute.

The following graph shows this increase in page views, focusing on 2013 specifically: The first week of January represents our zero point, and the growth is charted from there.

 

Increase in weekly page views during 2013, starting at the first week of January 2013, and ending at the first week of January 2014. Every Monday-Sunday period is represented by a dot, and the dots gently meander upwards, with a steeper increase towards the end of the year. Numbers are available in the linked Google Drive spreadsheet.

 

The pattern we witnessed in 2012/2013 - increased traffic in December, with a considerable peak in the first week of January, followed by a brief lull - also held in 2013/2014.

Even though we're only hosting text-based works right now, this much site usage still generates a fair amount of traffic to and from the database. By the end of 2013, we were moving 10 Terabytes of data every month.

As mentioned before, all these numbers (and more!) are available in spreadsheet form:

» AO3 Stats - 2013 (on Google Drive)

 

This concludes our first post! The next one will follow at the beginning of February and include looks at international site usage (spoiler: we are everywhere!) and browser preferences (spoiler: a lot of people browse from phones and tablets). If you have any questions or have created any additional charts, let us know!

Comment

AO3 Passes 200,000 User Accounts

Published: 2013-08-28 11:46:35 -0400

At the start of 2013 we posted about the AO3's growth in 2012. Those numbers showed new milestones in total works, total fandoms, total collections, total Support tickets answered, and total account users.

At the time we said, "While we still have a large body of users awaiting accounts, currently around 10,000, this is down significantly from the 30,000+ that we saw through the latter half of 2012, and we hope to decrease the wait further in 2013. It seems likely that Archive use will continue growing strongly this year, possibly even repeating 2012's feat of more than doubling its user base."

And it has! As of yesterday the Archive passed 200,000 users. By March the wait for an account dropped to 24-48 hours and we've been averaging about 500 new users a day.

For those who are feeling nostalgic, you may want to take a look back at a post recapping the AO3's first year in open beta.

When we entered Open Beta on 15 November 2009 we had:

  • 347 users
  • 668 fandoms
  • 6565 works

The site went through a 3000% increase in users that first year and an equally phenomenal jump in content. Yet today we gain more users each day than we started with in 2009.

Visitors from Everywhere!

While account holder statistics are good to know, there are many more people using AO3 than currently have accounts. Whether this is because they're random visitors, casual users, or don't find the benefits of an account appealing, we still have some clues that they're stopping by. For example, in the month of July the AO3 served pages to 3,663,572 unique IP addresses. So even if everyone with an account was using 5 different locations to access us, that's an awful lot of non-account visitors!

Another clue comes from the amount of bandwidth we're using. From just over 5 terabytes served in January we used 7.9 in July. If that pace keeps up we'll have doubled our use this year.

We've also been able to see the many countries that people are clicking in from. We've counted over 60 countries with simultaneous user activity over the past several months (including Denmark, Kenya, Jordan, Cambodia, Bosnia and Herzegovina, and Uruguay, just to name a few at random), and many more with at least a handful of visitors this year.

Technology Old and New

We also took note of browser usage among our users and have shown the top 20 in the graph below. In the month of July this ranged from 43.5 pages per minute being served to a Windows IE 8 browser to 533 per minute being served to Windows Chrome 28.

Pie chart of browsers that accessed AO3 in July 2013

Users Doing Stuff!

Here's a quick look at the growth in certain actions on the site.

The number of users we've watched grow over the years is dwarfed by the much larger amount of commenting and bookmarking activity. In the chart below the red line below barely seems to increase compared to the growth in commenting activity this year (green line).

Graph showing growth in comments vs. user accounts

Because the scale of growth is so different for these two actions we can also look at actual numbers taken from the graph below. Comment growth has been somewhat less than that of bookmarking (yellow line). In July there were 299,849 comments left and 438,746 bookmarks added. But both are far less than the growth of kudos (orange line) which is often used by people who don't have accounts or aren't logged in. There were 1,544,028 kudos given in July 2013 alone.

Graph showing growth in comments vs. bookmarks vs. kudos

We welcome all our new users, whether they use an account or not, and hope that they enjoy their time on the site! If you're hesitant to get an account because of long wait times, don't forget that our automated invites should arrive within two days of your request. If you don't see it, please check your spam folder and make sure that @archiveofourown.org is whitelisted.

Comment

2012 AO3 Milestones

Published: 2013-01-01 13:55:44 -0500

Happy New Year everyone! The OTW is looking forward to new developments this term and one thing we want to celebrate are some milestones that have been passed at the Archive of Our Own in the last few weeks.

Back in 2010, we wrote a post heralding our first major milestone when we reached 100,000 works less than a year into our beta. (We should point out that another OTW project, Fanlore, also hit the 100,000 edit mark that same year!)

As of November 25 we passed 500,000 works archived at AO3. We passed the 10,000 fandoms mark a week later, and on December 17, our Support team answered the AO3's 10,000th support ticket! Then by December 25 we passed the 100,000th user account. As many users discovered in June, the AO3 invite system was put in place to avoid having the site crash during a surge in enrollment. This practice paid off in 2012 when a large number of users migrated to the site in May and began adding works. During the following months new code was written for the site, primarily to redesign the way filters functioned, and new servers have been added. The site remained stable despite continued high demand for accounts, so the invite queue kept being increased every few months to get new users into the site more quickly. As of December 18 we also restored account users' ability to request invitations.

While we still have a large body of users awaiting accounts, currently around 10,000, this is down significantly from the 30,000+ that we saw through the latter half of 2012, and we hope to decrease the wait further in 2013. It seems likely that Archive use will continue growing strongly this year, possibly even repeating 2012's feat of more than doubling its user base.

Engagement With the AO3

As the following graph shows, there has been a distinct jump in various activities on the AO3 during the past eighteen months. User growth appears almost steady compared to reader activity, which can be seen clearly in the increase of bookmarking, commenting, and subscriptions.

However as this second graph shows, all of these numbers are eclipsed by the enormous jump in kudos activity. While the early years of the AO3 saw use primarily as a storage site for authors' writing history, current use is clearly favoring active searching, reccing and participation from readers.

Growth of the Site

While the Archive had to discontinue unique visitor counts in the spring due to the load on the site, we are averaging 80 million page views per month. Another way to look at the growth of the site is with the following statistics, comparing items from AO3's launch in September 2008 to September 2012.

  • Comments per Work 2008: .0075
  • Comments per Work 2012: 6.44
  • Bookmarks per Work 2008: .0382
  • Bookmarks per Work 2012: 7.17

Collections weren’t added until 2009. Their creation statistics are as follows:

2009: 93 new collections
2010: 478 new collections
2011: 771 new collections
2012: 1300 new collections

There are now over 1400 top level collections. Many of these have subcollections for yearly challenges/events, giving us over 2600 collections in total. Yuletide, for example, has 10 subcollections, but only the main Yuletide collection appears in that top-level listing.

We're looking forward to seeing what fan creators and AO3 users bring to the archive in 2013 and we'll keep working to improve your experience on the site!

Comment

So, About those Additional Tags...

Published: 2012-12-14 13:57:17 -0500

The following is a post created by a member of the Tag Wrangling Committee to address some ongoing questions and discussions involving freeform tags on the Archive of Our Own.

So.

Let's talk about those Additional Tags.

More specifically, let's talk about the long-form descriptive tags that are frequently being placed in the Additional Tags field. I want to get some facts on the table so our users - both consumers and creators - can have this important discussion properly. Any numbers cited are as of 0100UTC, 27 Oct 2012.

Full disclosure: Hi, I'm Sam J. I am a Wrangling staffer, a Wrangling volunteer, a Support staffer, and an Archive user. I have four horses in this race and, frankly, they're running in at least two different directions, leaving me with a varying opinion of these tags depending on when you ask me.

  • At last count, there were around 160 Tag Wrangling Volunteers. There are 10,232 Fandoms on the Archive. Of those, roughly 5,300 do not have a wrangler listed, so they are not tightly monitored. Many of these unwatched fandoms are occasionally wrangled by volunteer teams, or are metatags containing fandoms that are tightly wrangled.
  • As per the precedent established in the AO3 Terms of Service, we consider the tags on a work to be part of the content of that work. As such, the Tag Wranglers do not—and cannot—change, add, or remove tags from a creator's work. Any such changes to tags have to be initiated by Abuse, who only act in cases of tags that are against policy and are handled according to their protocols and the Terms of Service.
  • In recent months, the Archive's seen an overall increase in the number of Additional Tags on works. From last October to November, the number of Additional Tags on the Archive increased by 2,535, while the number of total works increased by 7,046. From this September to this October, that number has increased by 12,920 while the number of total works has increased by 22,936. Neither increase is linear - the works-per-month growth has been roughly stable since April, and the Additional Tag growth has been consistent, plus or minus 10%, since July.
  • The rate of growth for canonical Additionals over the last year has remained fairly consistent, gaining a average of 220 a month. (Four months were aberrations: March increased by 388; May, 296; March, 288; and September, 147.)
  • The Additional Tags were not responsible for the Death of the Filters. The sheer number of works on the Archive are what stressed the old code, and the sudden spike in readers/viewers starting in May pushed it past its capacity to fulfill requests. Because the filters pulled and displayed the canonical forms of tags, there were often far fewer Additional Tags listed than in the actual search results.
  • Non-canonical tags with only a few uses put almost no strain on the servers. It's the popular canonical tags and metatags that put the most strain on the servers.
  • Additional Tags are not distributed evenly throughout the fandoms—the massive increases in Additional Tags are concentrated in a limited number of fandoms. Even fandoms of similar sizes can have wildly divergent Tags/Works ratios. Drawing from random fandoms:
    Fandom Tag Works using Fandom Tag All Additional Tags* Additional Tags per 1000 Works Canonical Additional Tags Canonical Additional Tags per 1000 Works
    Buffy the Vampire Slayer 10847 692 63.80 184 16.96
    Cats - Andrew Lloyd Webber 37 4 108.11 0 0
    Harry Potter - J. K. Rowling 19422 2391 123.11 344 17.71
    Hockey RPF 1381 179 129.62 82 59.38
    Homestuck 9990 2475 247.75 97 9.71
    Inception (2010) 3796 300 79.03 19 5.01
    Marvel Avengers Movies Universe 16442 3164 192.42 166 10.10
    Naruto 3167 281 88.73 19 6.00
    Sanctuary (TV) 1359 117 86.16 53 39.03
    Sherlock (TV) 18300 3981 217.54 60 3.28
    Xena: Warrior Princess 293 16 54.61 4 13.65
    *NB: These numbers do not include Additional Tags already wrangled into "No Fandom", as the system does not have a way to generate those numbers. However, the number of "No Fandom" tags tends to be proportional to the fandom-specific Additional Tags.
  • When users create new tags (be they Fandom, Character, Relationship, or Additional/Freeform), they automatically:
    • will not show up on that fandom's Show Tag page;
    • will not show in the Filter sidebar of Works pages (exception: your personal bookmark tags will show in your personal bookmarks filter), though they can be filtered on, to an extent;
    • will not show up in auto-complete fields.
    A wrangler has to manually add Fandom links (or toss the tag into No Fandom) by typing in the Fandom name(s), and/or mark it as Canonical (allows the tag to appear in the auto-complete and be filterable by anyone) via a checkbox. The Wrangling interface does allow for mass-wrangling tags into a fandom and mass-marking them as canonical. The guidelines for Additional Tags are very selective as to what should or should not be marked as canonical.
  • Users can search for works using unwrangled Additional Tags by either clicking on the tag where it appears or by using the Works Search. (The Works Search uses a string search for the text of the tag, in addition to searching via wrangled tags.)
  • Logged-in users have the options of a few skins that affect how Additional Tags display in search lists. This skin shortens the Additional Tags to around 15 characters. This one puts all tag fields over a certain length into a scrollbox so they take up less room on the works pages, and this one hides the appearance of Additional Tags in search lists completely. If you do not yet have an AO3 account, the CSS listed in these skins can also be used in third-party site scripting tools, such as Stylish. Additionally, a logged-in user has the option to go to their Preferences and activate "Hide additional tags". This turns the entire content of the "Additional Tags" field to a "Show Additional Tags" link. Currently, both of these options are primarily available to logged-in users and do not apply to email subscriptions or tag ATOM Feeds.
  • Wranglers and Coders alike have been considering ways to additionally mark these tags in the front-end code, so that via a site skin, a third-party plugin, or another method, a user can have more fine-grained control over tag viewing when browsing. (Any coding solution will, almost by definition, require more data pulled from the servers, so there's a lot of evaluation before we push any buttons.)
  • The wrangling interface does need some improvements. (Depending on who you ask, a lot of improvements.) We are working on them, but our coders' time is a limited resource. As well, we have wranglers on as many browser and OS combinations as our users in general, so it takes significant testing to make sure the interface doesn't degrade for anyone, which is time-consuming.

There will be a second post tomorrow stating the Tag Wrangling Staff's official point of view on the sustainability of the current Wrangling system. If there's something you have a particular question about, leave a comment and we'll try to get an answer for you!

Comment

Count ALL the tags!

Published: 2012-10-17 15:43:32 -0400

We've made it another year! As part of the OTW October Membership Drive, we thought we'd share some of the stats and growth over the last two years, and a breakdown of the tags on the Archive.

Warning right up front: this post is very graphics-heavy.

Overall tag growth and canonical growth

Overall, our growth has (inevitably) been nothing but up. There has been proportional growth between the number of works and number of unique tags over the last two years. In October 2010, there were 107,430 works on the Archive and 105,750 unique tags; in October 2012, there are 459,655 works and 395,099 unique tags. The number of canonical tags (the ones which come up in the autocomplete and filters), however, scaled up much more slowly: from 55,697 in October 2010 to 140,306 in October 2012. This reflects the fact that the AO3 tagging system is designed to give creators as much freedom as possible in how they tag their works, so while the arrival of a new fandom on the Archive might generate only a few new canonical tags so the fandom name and characters can come up in filters, there might be a whole host of non-canonical synonyms reflecting the different preferences of creators.

a multiple-line graph showing three lines for the number of works, unique tags, and canonical tags over the last two years by month

Ratings

For these next few categories, we don't have a historical comparison.

The greatest number of our works - just over 31% - are tagged "Teen and Up", with "General Audiences" close behind at just under 30%. "Explicit" works make up roughly 18% of the Archive, and "Mature", 16%. Roughly 5% of the works are "Not Rated".

a pie chart showing the percentages of works using each Ratings tag.

Warnings

Unlike Ratings, Warnings are non-exclusive: a work can have multiple warnings. The vast majority of works on the Archive - almost two in three - are tagged "No Archive Warnings Apply". Around a quarter of the works are tagged "Author chose not to use warnings." "Major Character Death" has roughly 18,000 works; "Graphic Depictions of Violence" has 17,000; "Underage" has around 10,000; and "Rape/Non-con" is tagged on just over 8,600 works.

a horizontal bar graph showing the number of works using each Warning tag.

Categories

Like Warnings, Categories are also non-exclusive. Roughly four in nine of the 460,000 on the Archive are tagged "M/M", making up the largest Category by far. "Gen" has roughly 125,000 works, and "F/M" has just under 100,000 works. The other three categories are much rarer with 23,000 works or fewer.

a horizontal bar graph showing the number of works using each Category tag.

Tags by type, 2011 - 2012

All User-generated Tags

The following two graphs show the month-by-month growth of total unique tags and canonical tags on the Archive, with the vertical bars broken up for each type of tag.

The unique tags have a linear growth from the 105,000 tags on 01 October 2010 until around December 2011, then they start showing a slight upward curve to their increase, to a current total of just under 400,000 unique tags on the Archive. Characters and Relationships are almost as large a percentage of the total tags as Freeforms (aka Additional Tags).

a stacked bar chart showing the increase in unique user-created tags, stacked by type, over the last 24 months.

The canonical tags, on the other hand, are maintaining a roughly linear increase, from 56,000 in October 2010 to today's 140,000. Proportionally, characters comprise the majority of canonicals, followed closely by Relationships. Freeform canonicals are roughly as common as Fandom canonicals. (Reasons for this can be seen in our Freeform Wrangling Guidelines.)

a stacked bar chart showing the increase in canonical user-created tags, stacked by type, over the last 24 months.

Fandoms

Fandoms have had a very consistent growth, with a notable bump in unique tags in May 2012, when many new users imported existing works from other sites. The number of canonical tags roughly follows this increase, but has been slowing down in recent months. 50% of the 14,000 Fandom tags were canonical in October 2011, decreasing slightly to 43% of the 23,000 unique tags in October 2012.

a vertical bar graph showing the increase in unique and canonical fandom tags over the last 12 months.

Characters

We can see a similar pattern with the Character tags - a linear increase in unique tags, and a slowing down of canonical tags. The increases aren't proportional, however: while almost 74% of the 72,000 character tags were canonical in October 2011, only 57% of the 114,000 tags are canonical in October 2012. This may reflect a greater diversity of fannish terminologies being contributed by newer users of the site.

a vertical bar graph showing the increase in unique and canonical character tags over the last 12 months.

Relationships

Relationship tags also show the same linear growth as the other two, with a slight decrease in the number of canonical tags. Due to the ever-climbing number of combinations, these increased more proportionally: in October 2011, canonicals were 47% of the 68,000 relationship tags; in 2012, they're only 44% of the 118,000 tags.

a vertical bar graph showing the increase in unique and canonical relationship tags over the last 12 months.

Additionals

In what should not be a surprise, the majority of growth in unique tags comes from the unique Additional tags (also called Freeform tags). The number of freeforms has increased along an increasing slope from 43,000 last October to 138,000 this October. However, as the vast majority of freeforms entered are not intended for searching and indexing, far fewer have been marked canonical: there were just under 9,000 canonical freeforms in October 2011, and there are only 11,500 canonical freeforms in October 2012, as most freeform wrangling consists of glancing at a list and picking out the ones that would be useful as canonical tags (for example, common terms such as 'Angst').

a vertical bar graph showing the increase in unique and canonical additional tags over the last 12 months.

Last Words

We always enjoy taking a look at stats, and tags are particularly interesting because they often give a snapshot of different fannish communities or traditions. We love the way different communities of users on the Archive take advantage of our unique tag system to tag in all kinds of different ways!

The growth in tags reflects the massive increase in the number of users on the site. If you're enjoying using the AO3 and you'd like to help with our running costs, please consider donating to our parent Organization for Transformative Works. Donations help fund the AO3 and all the OTW's other cool projects!

A note on tag filters

In any post about tags, we know people will want to ask about tag filters. We know that the Archive is much harder to browse without this feature, and we're sorry it's taking us a long time to restore it - the rewrite is a significant piece of work. The good news is that we're so close now we can almost taste it - the new filters are on our Test Archive and if testing goes well they should be rolled out to the main site in a few weeks time. Wish us luck!

Comment

Spotlight on Support: AO3 ticket stats

Published: 2012-07-25 05:07:03 -0400

This is going to be a very boring post. It's going to be full of numbers, and graphs, those things that I may or may not have spent many years at school colouring in with lovely coloured pencil without understanding them much (because I was apparently too much of an innocent mind to turn them into rude, crude approximations of things not related to mathematics except in the most abstract sense), and yet, these will be very easy to understand numbers. I am not a statistician, nor are the levels of data I have access to very deep. What I am is a member of the Support Committee with a curiosity about the numbers and types of tickets that pass through our hands, and who decided to add up the numbers one day and turn them into graphs. That was last year, and somehow the lure of the bar chart means that I have continued to collate information through to where we are right now, having just finished the second quarter of 2012.

In this post, I'm going to summarise the types of tickets received, what categories they fall under, and the general trends we witness. But first, some explanation of the process.

Collecting The Numbers

I'm sure that the method I have used is going to come under some degree of criticism for being inefficient; however, our Support software, provided by 16bugs (see Sam's spotlight post for more information) was not designed for data export. This means that the only way to extract numbers of tickets is to do it manually. And by manually, I mean I go through the email duplicates of each ticket one by one, assigning them a category, then add up the numbers for each month and enter them into an Excel spreadsheet.

What this method is, for all its faults, is quick, which means that I can rapidly pull up a given time period to see what sort of tickets were received between those dates. These graphs were originally created as an informal overview of ticket stats (which is a position they remain in – production of these stats is not an official Support Committee duty). They are simply counts of the original tickets, what they are about, and when they were received. They are not a count of how quickly they were responded to, who responded to what, or what follow ups were conducted with the users.

Categories

I'm going to leave direct explanations of the categories until the sections for the respective quarters, as these change on a quarter-by-quarter basis. This is due to the simple fact that new features are added, which generates new issues, and old issues are resolved. For example, squid caching was not implemented until June of this year, so prior to that, it was not shown in the graphs because issues relating to it did not occur. Here I'll instead explain the process by which tickets are categorised.

If you've ever submitted a comment or query to Support you will notice that on our form is a drop down menu.

screenshot of menu options: Bug Report, Feedback/Suggestions, General/Other, Help Using the Archive, Languages/Translation, Tags

These categories are not the ones I have used to sort tickets. Since the categories in the menu are so few and so broad, I felt it necessary to granulate them further, and count tickets as they related to specific archive functions and features.

If a new category is created in my sorting, it's because an issue got a large number of tickets and wasn't a transient bug. For example, if a ticket is related to subscription emails, it is categorised under "Subscriptions", not "Email" because it is related to a specific Archive function (in this case subscriptions) that has an existing category. If it were related to invitation emails, it would go under "Accounts/Invitations/Login". However, if it's related to kudos batching, it goes under general "Emails", because there is no category for kudos.

The Stats

2011, In Brief

I won't linger on 2011 too much (see Q1 2012 for an explanation of categories), since this information was a little more awkwardly hacked together than for 2012 – by which time I had sorted out my process for quickly organising tickets.

bar chart with different colors for every month in 2011, representing absolute ticket numbers for each in 15 different categories
(full size)

Prior to August, tickets were collated by the Support Chair, using slightly different categories than I did. I attempted to meld the two sets of information as best I could to produce the above year overview.

What is easily and clearly visible is the spike in tickets in November, resulting from a change to the front-end presentation of the AO3. The biggest spike is split between Interface/CSS tickets and Feedback. While many of the tickets sorted under Feedback were directly related to the changes to the AO3's interface, they did not contain bug reports or requests for information, and therefore fell under the heading of Feedback.

Q1 2012

Categories for Q1 2012:

  • Error 502 - the 'server busy' messages
  • 1000 Works - queries related to why we have a 1000 work limit on the fandom landing pages
  • Activation/Invitation/Login - problems activating accounts, getting invitations, or logging in
  • Admin/Abuse - issues that need to be examined by Admin or Abuse teams
  • Bug Report - Reports of transient bugs that aren't separately categorised
  • Collection/Challenges/Prompts - any problems/queries about these
  • Downloads - errors, bugs, queries related to downloading
  • Feature Request - any 'can I have/I would like/will you implement' queries
  • Feedback - any complaints, or any positive feedback (alone with no other feature-related issue)
  • Help/Information - any questions about AO3/OTW in general, or how to use specific features
  • Interface/CSS/Display - problems/queries relating to how the archive appears on screen, i.e. interface
  • Imports - issues with importing from LiveJournal/Fanfiction.net/other
  • Open Doors - questions related to fics imported through OD
  • Search/Browse/Filter - Problem or queries about sorting through archive contents
  • Tag Wrangling - any tag related questions


bar chart with different colors for Jan, Feb, and Mar 2012, representing absolute ticket numbers for each in 14 different categories
(full size)

Possibly due to the fact that the holidays are still going on at the beginning of January (and thus, people have more time to spend on fandom sites) we saw more tickets in general than during the following two months.

Q2 2012

Categories added for Q2 2012:

  • Embedding – queries/problems with embedding media (images/audio/video) into Works pages
  • Bookmarks – queries/problems involving bookmarking
  • Caching – bug reports that are actually caching issues (e.g., reporting 0 works in a fandom as a bug – this is a caching issue, or appearing as logged in as another user). The kind of caching which causes these particular bugs was only implemented in June.
  • Email – email issues unrelated to other categories (e.g., kudos email batching)
  • Subscriptions – issues/queries to do with the subscribe feature


bar chart with different colors for Jan, Feb, and Mar 2012, representing absolute ticket numbers for each in 14 different categories
(full size)

To break down the invitations emails, in June we received 140 tickets related to Invitations.

  • How Do I Use This Invite: 22
  • Did Not Receive Invitation Email: 31
  • Fell Off Invite List (unaware of security changes): 41
  • (of those, who admitted to re-adding themselves: 6)
  • General Invite Queue Unhappiness: 10
  • Can I have An Invite?: 9
  • I Requested Invites, Where Are They?: 12
  • Paid Accounts: 3
  • My friend on FF.net needs an invite: 12
  • Need Invites for a Challenge: 5
  • Please Remove Me From Queue: 1

The remaining 37 tickets in that category were related to account activation or login issues.


bar chart with different colors for each week of June 2012, representing ticket numbers for each in 20 categories
(full size)

This graph shows how the tickets were distributed during the weeks that span the month of June. In week 23 (commencing 4th June) we received the greatest number of queries regarding invites, as this was the point at which the invitations queue started growing at the rate of nearly 1000 new additions per day (a rate since slowed to around 300-odd per day). This coincided with the point at which the AO3 servers started creaking under the strain of lots more visitors and a filtering system that was originally designed with a smaller user base in mind.

When squid caching was implemented to help ease the strain (around week 24) we saw an increased number of tickets related to this change. In week 25, when filtering was disabled, we began to see an increased number of tickets related to that. (Originally, the message was ill-worded, appearing to be an error message, rather than an admin message – this has since been altered, and tickets regarding the filtering being 'down' have disappeared.)

And This All Means...

I always have fun posting these stats to the support committee. Everyone already knows more-or-less how things have gone, but sometimes looking at the numbers surprises us. When I originally created them, one frequent question was "what's the most common ticket you get" to which we would generally reply "queries regarding the 1000 work limit". I was curious as to whether this was actually the case. As it turned out, Feature Requests came in more often. Questions about the 1000 Works came lower down the list.

If you are wondering how many tickets we answer altogether, I can tell you that at the time of writing there are no unanswered tickets in our support software (except for one bugged ticket, which we are attempting to resolve with 16 Bugs). Every single ticket we receive is read and personally answered by a member of our staff, usually within a day or two. So, the answer is: we answer all of them.


graph showing the number of tickets for each month from Jan 2011 (170) to May 2012 (590)
(full size)

This post by Support staffer Yshyn. If you find a bug, have a question about the site, or want to request a feature, you can submit a Support request.

Comment

My, how we've grown! A few AO3 stats

Published: 2012-07-16 12:09:58 -0400

We've been talking a lot recently about how much the AO3 has expanded over the last few months. One easy statistic for us to lay our hands on is the number of registered accounts, but this only represents a tiny portion of site activity. Our awesome sys-admin James_ has been doing some number crunching with our server logs to establish just how much we've grown, and provided us with the following stats (numbers for June not yet available). Thanks to hele for making them into pretty graphs!

Visitors to the AO3

Line graph showing the number of visitors to the AO3 per month, December 2010 to May 2012. The line progresses steadily upwards with a significant spike from 1,197,637 in April 2012 to 1,409,265 in May 2012.

The number of unique visitors to the site has increased almost every month since December 2010 (each unique IP address is counted as one visitor). There are a few points where the rate of increase gets more dramatic: there was a jump of 244,587 across December 2011 and January 2012, compared to one of 137,917 over the two months before that. This can probably be accounted for by the fact that during December and January, holiday challenges such as Yuletide bring more people to the site. This theory is borne out by the fact there was a slight dip in the number of visitors during February 2012, indicating that some of the extra traffic in the previous two months were 'drive by' visitors who didn't stick around.

May 2012 saw a steep increase in the number of visitors: there were 211,628 more visitors to the site than there had been the month before! The rapid increase in visitors was not without its price: this was the month of many 502 errors!

Traffic to the AO3

Line graph showing AO3 traffic in GB per month, December 2010 to May 2012. The line progresses steadily upwards with a significant spike from 2192 GB in April 2012 to  2758 GB in May 2012.

The increase in the number of visitors to the site has also been accompanied by an increase in overall site traffic (how much data we're serving up). Again, there's a significant spike during December/January. Interestingly, there's no dip in traffic for February 2012, showing that even though there were some 'one time' visitors over the holiday period, there were also plenty of people who stayed and continued to enjoy fanworks on the site.

The increase in traffic to the site clearly accelerated in 2012. Between January and May 2011 traffic increased by just 159.92 GB; the same period in 2012 saw an increase of 1,870.26 GB! In fact, with an increase of 566 GB during May 2012, that month alone saw almost as big a jump in traffic as the whole of the previous year (595.63GB)!

And the other stuff

With these kinds of numbers, it's not surprising that there've been a few bumps along the way. For information on how we're dealing with the growth in the site you can check out our posts on performance and growth and accounts and invitations.

Many thanks to our dedicated volunteers for their hard work dealing with the growth of the site, and to our fabulous users for their patience with our growing pains - and for creating the awesome fanworks so many people are flocking here to see!

Comment


Pages Navigation