This season is providing a good case study in how Amazon Japan can, sometimes, be very tough to use in predicting sales!

Right now the top 10 titles by total DVD+BD accumulated points (not final predictions) are:

6,154 – Free! Eternal Summer
5,519 – Sword Art Online II
5,491 – Persona 4 Golden
3,607 – Gekkan Shoujo Nozaki-kun
3,468 – Sailor Moon Crystal
3,146 – Kuroshitsuji Book of Circus
2,554 – Fate/kaleid liner Prisma Illya 2wei
2,490 – Bakumatsu Rock
2,484 – Aldnoah Zero
2,011 – Tokyo Ghoul

For most of these, Amazon is problematic.

Free!, Bakumatsu Rock, and Kuroshitsuji are certain to be underestimated because their demographic is strongly skewed towards stores like Animate, particularly in Free!’s case.

SAO II, Aldnoah, P4GA, and Kuroshitsuji (again) are also likely underestimated due to Amazon only offering less-competitive “Amazon-exclusive” editions, aka significantly more expensive than normal releases and with truly pitiful extras nobody wants.

Prisma Illya should also be underestimated, probably more than season one (which was 25% under on BD) since its sales look like they’ll be down in general, leaving mostly core fans who I’d think are more likely to want store bonuses.

This leaves Nozaki-kun, Sailor Moon, and Tokyo Ghoul as the only current titles over 2k points for which we can realistically think Amazon may be reliable. Nozaki-kun and Ghoul have event tickets, though that’s not an Amazon-specific skewing factor. And Sailor Moon… uh I have no idea really. It’s been dropping and dropping since its strong start and I dunno how it’ll sell.

It’s just one of those seasons.

25 Responses to “Prediction Talk: Sometimes Amazon seems kinda useless”

  1. Anon says:

    It’s just one of those seasons where every Oricon report is going to be entertaining as hell.

  2. 90% underestimates ahoy!

    If stalker ever gets overhauled, I hope they take a look at whether series with bigger DVD#/BD# ratios have more storefront. That’s about the only thing I can think of that might help mitigate female demo underestimations without specifically singling out shows on a case-by-case basis.

    • something says:

      Now that Stalker is regularly tracking rankings at six additional stores, I hope it’s keeping all the data and not purging it at each update. The only way we can even think about factoring non-Amazon shops into the point totals is to let the system collect as many data points as possible over the next year or so, and then hope some stats genius is able to correlate that to sales and come up with point adjustments.

      Like was done for Amazon, but with much much less robust data sets… and there may be no trends to identify so it may not be possible. It’d be nice though! But on the other hand it would make the calculations a lot more opaque to the end user. I sort of like the simplicity of the current method, even if we know it will be wildly wrong sometimes.

      • Primadog says:

        I doubt alternative store rankings will ever become a serious augment to STALKER points. STALKER logarithm works for Amazon, because it uses general DVD ranking of some 150,000 on sale DVDs as background for anime sales. A consistency that you would not get if STALKER uses a smaller subset of the overall ranking like, say, only the “DVD > Anime” rank instead. Anime sales may have up and down seasons, but as long as the overall market have some level of consistency, you can assign points to general ranking with confidence.

        We knew from Sakura Tricks’s performance at Gamer’s that high ranking in a specialty store may be entirely meaningless in a down season. Even though Animate moves enough volumes to impact sales, it doesn’t move enough units in other genres. Without that background ranking to work with, rank 5 at Animate may mean 10 pre-orders; it may mean 100; we have no way of knowing.

        We can’t assign value to things there.

        Therefore I think it would be a mistake to change the STALKER algorithm to account for storefront. Human intuition works better than algorithm when it comes to when alternative stores have nontrivial contribution to sales. We can also account for demographics better. In those cases, STALKER can work just fine as a baseline for us hobbyists to make educated guesses against.

        Personally, I’m highly doubtful that Free S2 will see higher underestimation than S1, percentage wise.

        • something says:

          I know, I’ve expressed those same reservations before. And there is no human being on this planet more keenly and painfully aware of the Sakura Trick/Gamers situation than me! That’s why if it were done, it would need to be done in the aggregate, only after collecting thousands of data points, and only if trends could be found. Lots of “if” in there. And even in the best case scenario, it would only be applicable as either a small adjustment and/or only applied to the most successful titles in storefront rankings.

          Like I said, the extra complexity is a downside and may be reason enough that it might not be worth it. But categorically saying “NO NEVER DON’T EVEN TRY” isn’t helpful at all either, it just stops people from even attempting to look at the data and see if there is in fact something there that we’re totally missing because we’re too confident in our blind, untested assumptions.

          (As for what it would be like if anyone did try, it wouldn’t even have to change the current formula one bit – it could be an additional alternative, just like Stalker shows both with and without the high price penalty for expensive items. No better way to know if it’s an improvement or not than to be able to directly compare! But again this is all hypothetical, assuming someone could even work something out, with enough historical data. Data I hope Stalker is capturing even if it never ends up using it!)

        • If there’s a wide range of values gained, but things that do better in one store very consistently get higher sales overall, it would still be possible to tack on the minimum amount from that range as a basic correction. Giving a release +10 points for a day spent in the #1 slot at Animate, say, wouldn’t be that different from the way nico preorders are used now (-3 point penalty, iirc).

          Too, one could tether the strength of the season on amazon (say, avg daily point accumulation of the top 10 BD/DVD combos) to how much alternate stores count for in that season. That’s assuming they found a correlation there, of course, but the point is that there would be plenty of ways to try and solve the problem. I don’t see anything wrong with testing out ideas that could lead to a more accurate model, beyond the time it would take to parse the data.

  3. Fipse says:

    In the end, how much do the sale numbers mean anyway? We know we never get the full first week numbers and we certainly do not get numbers for 1-2 years of sales (besides some big names). And the very least we know if an anime made profit or broke even…

    • something says:

      I’m pretty open in the FAQ about how DVD/BD sales numbers are only part of the story, and furthermore that the data we get is only part of that part of the story. Sales numbers are best used as a relative guide, not an absolute number. If Oricon says one show sold 1,000 and another sold 5,000 and another sold 15,000 you can be pretty darn confident those are three very different levels of performance. Even if the 1,000 actually sells 2,000 and the 5,000 actually sells 8,500 and the 15,000 actually sells 21,000 (or 30,000!), or whatever! It’s about getting as level a playing field as we can, to tell what’s doing better or worse than its peers.

      • Fipse says:

        Yes, it is good to compare shows with each other but people tend to call shows “a flop” if it does not sell 4k+ numbers. I know you see things a bit more rational and we really do not have much other sources to say something about the success of a show.
        I just have the feeling that more and more people use those numbers as hard evidence to judge the success of shows while it becomes more and more clears how unreliable the sales numbers and especially stalker can be.

        • something says:

          You’ll never stop people from abusing statistics.

          But it’s better to have decent numbers than no numbers at all, because people are going to call shows hits or flops whether they have data or not! Might as well give a small subset of us the tools to be a bit more rational about it – and I have seen plenty of people get interested in this stuff they’ve never cared about before, and learn a lot about it. We’re still better off than before.

  4. lolivampires says:

    Since you know something is going to be underestimated is there any realistic way of knowing how much it’s going to be underestimated by? That way the Amazon can still give you something to base an estimate off of.

    • something says:

      Realistically, nope, not a prayer. The only non-Amazon shop that we can say has a consistent, large impact on sales is Animate. But even then, predicting a degree of underestimation is just utterly impossible because it depends on so many other factors.

      Amazon can be messy for predicting sales but every other store is exponentially worse!

      • primadog says:

        Do you have data readily available for underestimation for Fujoshi-oriented shows? Are the underestimates consistent for those?

        • something says:

          I’d have to get together a list of shows aimed at women, but I’m pretty sure the underestimates are not all that consistent (due to a variety of factors I’m sure), other than that underestimation is very common.

          • musouka says:

            I’d be interested in the numbers too, but my gut says if there’s an underestimation, it usually occurs when the series is incredibly popular with fangirls, not middling successes. I seem to recall there being discussion somewhere on some Japanese site (gee, how helpful is that?) that I read about how women buying online tended to buy from Amazon, but I know it’s been a few years since I’ve seen it and I don’t think that particularly applies any more, if it was ever true to begin with.

            • something says:

              Ask and ye shall receive! I didn’t intend to go this into it but the numbers were waaaaaay too good to bury in a comment. Needless to say, yeah wow huge difference between shows aimed at women and the rest.

              And sub-3k sellers are only marginally less storefront-heavy than over-3k sellers among women so the middling successes and below behave quite similarly to the hits.

  5. Siliziumleben says:

    These “amazon” editions are really worthless. Would amazon at least have something interesting like a storage box, that other shops offer for a whole volume purchase and ship it internationally, I wouldn’t have a problem with paying more.

    But even with the higher price point it seems, that amazon is still slightly cheaper than for example HMV, because amazon still calculates the import fee deposit wrong, at least for Germany. Would HMV ship their exclusive stuff to Germany, I would order from them…but they excluded that on their page, so I think I will stick with amazon.
    Has anybody outside of Japan ordered from HMV and received the exclusive stuff?

  6. Sandf5 says:

    SAO II, Aldnoah, P4GA, and Kuroshitsuji (again) are also likely underestimated due to Amazon only offering less-competitive “Amazon-exclusive” editions, aka significantly more expensive than normal releases and with truly pitiful extras nobody wants.

    But i read somewhere Animate and Gamers with exclusive bonus are more expensive than Amazon, also most of stores without bonus sell Aniplex titles only with 15% off, only stores like Rakuten and DMM sells it significantly cheaper but also without bonus. Is that Amazon-exlusive editions really significanly affected the sales when expensive or not doesn’t seem to be the case here because Aniplex titles actually generally expensive compare to average release no matter what the store is.

    • something says:

      “But i read somewhere Animate and Gamers with exclusive bonus are more expensive than Amazon”

      Absolutely, they charge full MSRP. But if you’re an obsessive fan or collector-type (which describes many of the people who would buy anime BDs!), the much nicer extras that stores like Gamers get can be very enticing.

      When the cost of those extras is losing a 27% discount (Amazon’s usual discount), a significant number of people are going to say “alright, that is a lot of money, I’ll go with Amazon”. When the cost of those extras is losing a 10% discount, suddenly you’re it’s not nearly as big a factor.

      Look at it this way… buying a one cour series on 6 discs at ¥7560 MSRP each from Amazon will save you ¥12,247 (~$120 USD) total if the discount is 27%. That’s a lot of money to pay for an artbox, or poster, or drama CD (and even then, many people go the non-Amazon route anyway).

      But if Amazon is only giving you a 10% discount you’re only saving ¥4,536 (~$45 USD) across the whole release by shopping at Amazon. Suddenly that tapestry or towel looks way more tempting, especially if you were already interested.

      As for the other retailers without exclusive extras, even if this change in discount only brings them “on par” with Amazon, it nonetheless opens them up as more realistic options. They may have better S&H policies, or may offer a points program, or whatever.

      Basically what the slashed discount for “Amazon-exclusive” versions does is make everyone else in the industry a lot more competitive vis a vis Amazon if that’s the only version Amazon has. And that means Amazon will almost definitely represent a reduced portion of sales.

      We saw the first major hints with Mahouka’s huge underestimation and if we see the same from Aldnoah and SAO II and Persona, I think that’ll be really strong evidence.

      Note: I’m including sales tax and ignoring S&H because I’m thinking from the point of view of a Japanese customer, not an importer. For importers who buy a lot, Amazon is sadly still the most realistic option, because they don’t ship by weight and all the shops that offer worthwhile bonuses require proxy services that dramatically increase your total costs. So it’s not worth it unless you desperately, desperately want the retailer bonus.

      I’ve been very lucky that Aniplex has distributed nothing at all I’m interested in for quite a few seasons in a row now, so I haven’t had to deal with their crap. But Winter ’15 is going to be rough if this nonsense between them and Amazon continues. 3 of my 4 extremely likely imports in Winter will be Aniplex… and for all I know the fourth could end up being one too. Ugh.

      • Sandf5 says:

        We saw the first major hints with Mahouka’s huge underestimation and if we see the same from Aldnoah and SAO II and Persona, I think that’ll be really strong evidence.

        I read Mahoukafans said the bonus from Aniplex store for Mahouka are really tempting can be a reason why it got huge underestimation. Aniplex store counted toward oricon right?

        • something says:

          Aniplex+ reports to Oricon, yes. It’s very hard to know how much they account for. We have small hints with Monogatari, which suggest it sells at least 1.5k of the ~40k sales reported via Oricon, but that’s a very rough guess at best. No idea how that scales down for much lower-selling shows.

          But yeah, if Aniplex+ has a good bonus, they benefit from Amazon offering a worse discount because their own offering looks more tempting than it would if Amazon were 27% off. Frankly, that’s probably Aniplex’s basic goal with all this.

          • Sandf5 says:

            But my point is i read SAO II and Aldnoah get less tempting bonus though Persona 4 can be similar case with Mahouka due to its good bonus.

            • something says:

              Even if their bonuses are less tempting at Aniplex+, that store is just one small player. They should still be heavily underestimated (especially SAO) based on strong performance at a whole range of non-Amazon retailers.

              Right now, SAO II v1 is only predicted by Stalker to sell 12k.

              SAO s1v1 sold 45,372 and s1’s lowest selling volume sold 31,556 and its average was 35,879.

              We can expect SAO II to drop, maybe even drop a lot, but the drop would have to be truly huge for 12k to be even in the right ballpark. Or the right country!

              SAO II v1, I’d guess, will be underestimated by a bare minimum of 50%, and I would not be surprised if it were as much as nearly 100%.

              It’s a bit hard to compare to season one in terms of Stalker accuracy because Stalker ranged from 20% overestimate for s1v1 BD to 63.5% underestimate for s1v9. But comparing v1 to v1 I expect a massive swing from that overestimation to this underestimation, largely because Amazon is so much less competitive now than other stores. Obviously we can never be 100% sure because there are so many factors, but this is the only large and obvious difference between this time and last time.

  7. Natapori says:

    How do you think with Free! Eternal Summer?I’m expected it still around 20K for average even it so many drop.

    • something says:

      For it to even sell 20k vol 1, let alone 20k average, would require over 100% underestimation in both formats, which is, while not technically impossible (at least on BD), still unlikely.

      It has been on an upward swing since the release specs were finalized and the cover was revealed. The Stalker estimate has increased from about 8000 to 9,337. And if it continues to get stronger it could get its estimate over 10k. And still no event ticket announcement, which could boost vol 2 or 2-3 if they recreate what they did for s1. But I always felt it needed a minimum of a 12k estimate to lock in a decent chance of a 20k wk1 for v1. And given v1 can usually be expected to be the best selling volume, it needs to do a lot better than it’s looking to do now.

      So I dunno it’s still not looking food, in terms of chances of being anywhere near s1. This baffles me because I feel like this season blows the already great season one out of the water but sequel sales are funny like that! We’ll just have to wait for the data, I too would love it to keep over a 20k average given the >29k average of s1, but I do expect a pretty severe drop, which makes me sad.

      (Though not as sad as Hanayamata and Locodol doing terribly or Yama no Susume s2 putting in a terribly anemic performance compared to how s1 sold so well for a short. This may be a great season for putting out shows I like but not for getting people to buy them. At least Free will still be a hit, even if it could be dropping far more than it should.)

Leave a Reply