Guest Post: Anime and Sales Analysis, by Yuyucow

I am pleased to share a post by Yuyucow on an aspect of the anime industry that I am admittedly not well-versed in. You can follow Yuyucow on twitter.

We all know that anime fans more than anyone else get emotionally invested in works they enjoy, which in addition to a sense of insecurity makes them consistently seek proof that they like the right kind of show. While random forum comments and blog posts on the internet tend to be enough, some require further confirmation that their beloved cartoon is well received and thus deserving of their praise. That’s where sales become relevant. How could several thousand people willing to pay outrageous prices for a certain show be wrong about its quality? Isn’t that clear, solid evidence that they’re right?

So now we have two factors in play: rabid fans willing to (mis)interpret anything in their favor and the surprisingly complex data about anime sales. The situation is made worse by the fact that western fans are getting more acquainted with what used to be obscure and exclusive Japanese sites, meaning that they get their hands on tools they didn’t even know that existed. ANN’s weekly reports are no longer their only source of information, now they can stare at beautiful preorder graphs or look up data on wikis entirely dedicated to this. Data analysis and interpretation is based on knowing all the variables, but that’s not a problem on the internet since we’re all experts about everything and anything. Our anime connoisseurs are now armed with new tools they have vague knowledge about. And what’s worse: they’re willing to talk on the internet about it. But worry not, here I am to explain them how to properly use them so they become better, more informed people. And to spare myself from reading dumb stuff.

The Amazon Stalker is most likely the best place to start, since it’s complex to the point of making no sense to any newbie. It’s not just the overwhelming amount of data, using the site requires a degree of understanding of how anime preorder works and how do they relate to otaku buying habits. If it sounds like a hassle already – good, that’s the message I’m trying to get across. First of all, what is it? The site tracks the hourly updates of Amazon’s DVD/BD sales ranking (thus the name Stalker) and assigns a certain number of preorder points every 24 hours based on how it fared throughout the day. This fake currency is the result of the fancy algorithm the entire site is built around; based on the performance on Amazon it extrapolates how it’ll do in every retailer that reports its sales to Oricon (the source of the sales data) so that one preorder point is always equivalent to one real sale. This formula is obviously tweaked and updated on a regular basis, but the fact that the predictions are really close on most cases says a lot about the regularity of otaku habits. Amazon is a big site, no doubt, but it doesn’t account to most of anime sales by any stretch of the imagination, so it’s quite impressive we get spot-on predictions based on its preorder ranks.

With that out of the way we can move onto the site itself. This is what greets you:

From left to right – current rank preceded by a small sign indicating whether it’s gone up or down since the last update, accumulated preorder points, the rather irrelevant niconico preorders and the release day. So far so good, time to check the individual pages:

Looking simple as well. Current, best, worst and average rank followed by the niconico preorders and preorder points. Now it’s when it starts getting tricky, due to the last number you see down there. Those preorder expectations are calculated by taking the average number of points it’s earned in the last week and assuming the performance will be the same until the release date. Needless to say, this oversimplifies the situation and has a tendency of breaking the estimate in hilarious ways, more often than not overshooting the sales to impossible degrees. I won’t recommend to entirely ignore it since it might save you some mental arithmetic, but take into
account that it’s prone to breaking. Recently offered releases, fan reaction to certain key episodes or extras bundled with the BD/DVD are some of the situations it struggles with, but even more basic stuff like the weekly pattern for TV series with its spikes after a new episode has aired and the last week rush effect aren’t taken into account. Long story short: use, don’t abuse. Always take this value with a grain of salt.

The graphs are self-explanatory and don’t really add much other than a fancier way to display the data. Although I must say that the blue one is quite nice, since you can compare the current performance to the previous weeks’.

That’s about it as far as the Stalker’s mechanisms go, but most of the issues come from people failing to understand what this data means, not what it is. Since there are way too many variables to explain, especially if I intend to keep this remotely interesting, I’ll simply use sales data from the Summer 2012 season to exemplify many common situations.

Starting from the top we get Horizon S2, the sequel of a quite successful show that already had a large established fanbase to begin with. As it tends to happen with big franchises, all the volumes were massively preordered once they were available on Amazon, resulting on the blurays ranking really high for a great deal of time. And that’s where the singularities of the Stalker start becoming apparent; as a general rule, anything that stays up in the rankings for a while will be overestimated by the site. Which makes quite a bit of sense when you think that there’s no way to track preorder cancellations, so always shooting a bit too high and thinking of the miscalculation as cancelled purchases makes more sense than regularly underestimating releases. Jintai suffered from that on a much smaller scale. Funnily enough, what made it spike and thus caused the overestimation was Tanaka Romeo, the novel’s author, tweeting that he wanted 200 more sales to get more royalties.

Moving onto Tari Tari for a more amusing rule of sorts, which I can only sum up with “sometimes, shit breaks.” It looks off right now but it was even more extreme after the first week of sales, which only amounted to half of what the Stalker expected. While it did rank high for weeks, it wasn’t high enough that it would get overestimated like that, and the show did nothing to upset its fans so it can’t be a case of cancellations en masse – if anything, the sales seemed to get better as time passed. It’s not a case of special preorder bonuses being offered exclusively in one store. So why weren’t Amazon’s sales for this particular title consistent with its overall sales? I don’t know.

Sometimes, shit breaks.

Dog Days and YuruYuri serve as good examples of the so called storefront appeal, that makes them consistently sell more than expected due to people running onto them and buying on a whim. Love for an existing series like YuruYuri or even a certain studio like in Dog Days/Seven Arcs’ case is the main cause of this phenomenon. It most likely happened to Total Eclipse as well, since it’s backed up by a huge franchise like Muvluv, which combined with the great deal of cancellations due to fans pissed off by the adaptation ended up being one of the most spot-on predictions by the Stalker. It certainly gets tricky when multiple factors clash like that. Might as well mention Hakuouki and Arcana Famiglia in that regard, but their sales are quite low so the Stalker loses precision.

The Stalker is unreliable when it comes to shows with terrible sales. If you look at the lower part of the table you’ll notice that there are plenty of shows being overestimated in one format while selling less than expected in the other. The general rule of overshooting still applies, but everything is more chaotic and there’s no way to deduce actual sales numbers from them. It’s easy to tell when a show will sell really badly, but there’s no way to compare multiple shows with abysmal expectations and tell which one will fare the worst.

If you’ve made it this far then you should be aware that this is all overly complicated and not meant to be used by normal human beings. If you’re genuinely interested in this mess and are willing to keep using the Stalker, I hope this at least helped you understand its intricacies a bit better. I might write a follow-up explaining how Oricon works, how it doesn’t work, and why directly comparing anime sales is pants on head stupid. I’m afraid it might get too meta if I start talking about why certain shows sell so don’t expect it anytime soon, but it’s something that could happen.


5 thoughts on “Guest Post: Anime and Sales Analysis, by Yuyucow”

  1. Thanks for this post. It’s hard to describe why and what can be pretty accurate for one case and not accurate at all for another but I was able to follow you here. I suppose the more detailed algorithm behind it is not available to review (in English) somewhere?

    1. They’re very secretive about this kind of thing so I doubt it’s easy to find the actual algorithm. We’ll have to assume it’s a magic formula maintained by a Math Wizard.

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 )

Google+ photo

You are commenting using your Google+ 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