Fun With Numbers: Long-Tail Figures and Uncharted Territory

Recently, via the creation of a second OpusData trial account, I was able to acquire some new data that sheds light on how several newer US anime release performed in the long term. This includes 5 releases (Attack on Titan: Part 2, One Piece Film Z, The Wind Rises, Evangelion: Ha, The Tale of Princess Kaguya) for which we have at least one week of sales data already, plus 4 releases (Momo, Hal, Bayonetta, The Cat Returns) that came out in the 12 months and failed to chart once. Excluding Evangelion: Ha, and The Cat Returns, all of these titles came out between September and November 2014, so they’re roughly comparable in terms of the amount of time since release they’ve had to accrue new sales.

[Note: OpusData and TheNumbers, where I usually get what hard US video sales data I post, are owned and operated by the same company, Nash Info Services.]

The info that can be gleaned from them is interesting, but let me just say right now that I’m really happy about the clerical error that logged Attack on Titan’s second set as a movie, and hence trackable long-term in the Opus database.

Preface: This is a really messy batch of data culled from a database not without its problems. For my own sanity, I’m going to try to keep remarks brief here. If you want a better explanation, check this earlier post with a similar subject and more exposition. Or ask in the comments.

First off, here’s a comparison of on-chart versus total sales for the 5 in this new dataset which have both:


(Note that Evangelion has been available on video about 3 years longer than any of the others on the list.)

Compare these with what long tails were like for the previous generation of big hits (old image, apologies for the messier columns, one on the right is the important value):


It’s a little bit eye-popping. There are 3 releases beating the previous long-tail record of 0.21 set by Laputa’s BD rerelease, and Attack on Titan is doing so by about 80 percent! These are tails which an average of 10 months old which are (percentage-wise) beating tails which have been around for half a decade in some cases. It’s interesting because thresholds on the US charts have gone down a lot in recent years, so one could’ve made a case that tails should be shorter now. While 5 releases is a fairly small sample, the fact that all of them show tails much shorter than was typical of the older ones indicate heavily that modern anime releases accrue sales more gradually.

The sales of the other 4 movies, the ones that never charted but still got their names in the database, provides some info about the database itself.

A Letter to Momo (DVD) – 7374 copies

A Letter to Momo (BD) – 102 copies [this is laughably small and probably an error]

Bayonetta: Bloody Fate (BD/DVD combo) – 42,786 copies [logged as DVD]

Hal (BD/DVD combo) – 8838 copies

The Cat Returns (BD re-release) – 5648 copies

Let’s start with Bayonetta, the best-selling title of the four. This one is interesting because, like a lot of Funimation releases, it came out as a BD/DVD combo pack. However, unlike Hal and Attack on Titan, it got classified as a DVD release. If its tail was anything like that of the ones we just measured, it should have charted on week 1 if it was tracked as a BD; the BD threshold for that week was 5682, while the DVD threshold was an intimidating 17,165. So DVD/BD data can get mixed up or inconsistently classified in the Nash Info Services database. I expect that might have happened to A Letter to Momo, with a lot of DVD sales being reported as BD sales, as the BD number is a laughably small percentage of the total.

The Cat Returns, re-released in late June, sold less than I expected for a series that cracked the top 100 for multiple days. Given how Spirited Away also failed to parlay a release week in the top 20 to better than 30k sales, I suspect the competition at that time was weak enough to inflate both their rankings quite a bit. If these figures are accurate, they provide an interesting look at how mid-tier hits can perform beneath the thresholds.

This was going to be all I was going to say for now, but I found a little something extra very recently. All data above was collected on August 20th. The Funimation search was useful enough to set to a default, so, on a whim, I went back and collected it 2 days later – and the numbers went up a little! Specifically, the numbers for just the DVD releases went up a little, while the BD numbers remained constant. August 21st happens to be the day TheNumbers updated with a new week’s worth of DVD data. So there might be something there, a way to see how series long since released and well below thresholds do on a weekly basis. If the system works the way I think it does, with OpusData and TheNumbers updating simultaneously, it should be possible to get reliable figures within roughly a month’s time.

And then the sky’s the limit. The toughest part of the amazon rank-sales figures correlation has been that middle #5000-#20000 area, too high for low, slow-moving stock totals to be useful, and far too low to make a weekly top 20. Pinning that down would help a lot towards building a non-useless prediction model.

Source screencaps (sales):

opus-query-Arietty opus-query-evangelion opus-query-Kaguyahime_Kaze opus-query-momo opus-query-neko funi-opus-query_aug20

August 20th Funimation Query:


August 22nd Funimation Query:


2 thoughts on “Fun With Numbers: Long-Tail Figures and Uncharted Territory

  1. Following you on twitter, I thought for the last few months you’re ready to give up on this “STALKER USA ver” project. Has this splurge of new data shine some light at the end of the tunnel for you?

    • Well, the last few months I’ve been uncertain about a bunch of things. I felt burned out in general and had a hard time progressing any of the ideas I had into workable stuff for the blog. Switching tacks to interview translations helped – I was feeling gassed on numerical methods and worried about a lot of other incidental things. I’ll probably get back into them eventually, but what I’m doing now keeps me going.

      The US Amazon tracking stuff has always been my baby. Nobody else I know of has made the same sort of general effort to make the US market more transparent, and it only takes about 20-30 minutes out of a day. Those are minutes I would typically be surfing the web aimlessly, so it’s a really small daily effort required. I also feel like I’ve been having close calls for over a year in regard to cracking it, and every bit of new information, even the confusing ones, brings the project closer to success. Too, it’s worth keeping at because the standard for success is actually really low on account of it being a very opaque topic.

      The diversion with the continuously-updating numbers is going to be key – if they do indeed correspond to amazon ranks, given the amount of anime I can monitor (not counting BD/DVD splits), I’ll very quickly have hundreds of datapoints corresponding directly to a copies-sold figure. I’d be very surprised if I couldn’t build a better model out of that.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s