游戏邦在:
杂志专栏:
gamerboom.com订阅到鲜果订阅到抓虾google reader订阅到有道订阅到QQ邮箱订阅到帮看

解析应用商店推荐功能中的长尾理论

发布时间:2012-09-05 11:02:04 Tags:,,

作者:Rob Lockhart

Chris Anderson的《长尾理论》是一本关于市场营销和幂律分布(游戏邦注:幂律分布表现为一条斜率为幂指数的负数的直线,这一线性关系是判断给定的实例中随机变量是否满足幂律的依据)的著作。其基本理念是指当你不能同时持有多种类型的产品时(如满足各类人需求的视频商店),你便需要依靠热门内容(即大受欢迎的产品)来运营业务。而当你能够同时持有多种类型产品时(就像Netflix),你不仅能够提供利细分产品,同时还能够利用这些产品提高整体业务量——即使你并没有市场热门内容。

当我在阅读这本书时,我便想到了iOS应用商店,以及我们的工作室所合作的商店等。Anderson指出零售商总是会出售多种类型的产品,因为消费者总是会购买符合自己喜好的产品。他也表示创造者也能够从中获利,因为他们的作品有可能出现在热门产品旁边。

据我所知,创造者只有在自己的产品真正获得市场的推荐时才能受益。App Store的Genius推荐功能其实也只是一种可选项设置,所以从本质上看来并没有多大用处。这就意味着用户在选择自己喜欢的产品时只能经由游戏类型中的子类别,以及口头传播/媒体等方式。到目前为止电子市场中的口头传播方式仍只停留在现实空间中,也只有真正热衷于游戏的玩家才会去查阅游戏媒体(游戏几乎未曾出现在主流媒体上)。这就意味着比起前20名受欢迎应用排行的第一位,我们的产品更有可能排在之后的19位中。虽然这已经不错了,但却并不完美。

人们总是习惯将显示结果的难题称之为“检索问题”。从未在商店中购买过任何商品,只是根据前20名排行列表以及随机的应用商店研究结果(这也存在问题)的人们便会认为,这个列表以外的所有产品都是糟糕的,当然这也只是个见人见智的问题。而应用商店应该如何做才能为每位用户提供足够的选择呢?苹果/谷歌该如何在自己有限的销售空间中利用长尾理论?

使用推荐引擎服务,就像Netflix或亚马逊那样。这并不是一个选择过程,只能算是服务的一部分。即当用户登录时将会在头版页面上看到推荐应用,并且在消费后也仍能看到这种推荐。这种过滤方法能够同时让开发者和消费者受益。因为如此开发者便能够在相关的微类型产品间进行竞争,而不是所有人都只与《愤怒的小鸟》竞争,并且消费者也能够准确地挑选到自己喜欢的产品。

让我们着眼于Nook应用市场。这是一个再普通不过的应用市场,除了应用推荐是出现在消费者购买应用的页面下:

nook app marketplace(from gamasutra)

nook app marketplace(from gamasutra)

用户必须拉下标签选项才能看到“购买这一应用的用户同样也会购买”以及“更多该开发者所制作的应用”等内容。同时这里还设置了一个“分享”按钮,以帮助缓解口头传播所存在的问题。很明显之所以会出现如此设置是因为这些开发者也是Nook的eBook商店(基于bn.com网站)的创造者,所以他们不希望做出太大的改变。

还有谁的推荐能力强于亚马逊呢?下图便是我打开应用详细介绍时kindle fire上所呈现出的内容:

nook app store(from gamasutra)

Kindle-app store(from gamasutra)

这是来自amazon.com的一段文本内容:“用户看了这个**并最终购买了**”。亚马逊几乎在每个应用商店页面中都整合了推荐内容。当你打开商店时,你将会获得应用的详细介绍(如上图),当然了,在你购买了应用后,这些介绍也仍会继续出现。推荐始终贯穿于该应用商店中(右边最后一个标签),并扮演着非常重要的作用。

所以我怎么知道可发现性才是真正的问题所在?我怎么确认我的游戏糟不糟糕?我只能通过猜测,但是却不敢肯定。因为:在这个日趋成熟的市场中存在更多优秀的机制等待我们去挖掘,消费者至少会定期购买每一种产品。让我们着眼于这些数值:http://daveaddey.com/?p=893,并思考iOS应用商店的长尾为何会比标准的幂律分布来得均衡。

以下是我心中的一些理想条件:

1.任何应用商店的醒目页面都应腾出一定空间(20%至50%)去自动呈现应用推荐。确保推荐具有意义,否则只会遭到用户的无视。

2.为不同类型的应用划分不同区块。而每种类型的应用必须具有可扩展的(1)最畅销应用列表,(2)该类型应用的推荐,以及(3)子类型列表。

3.在用户购买应用后呈现出与他们购买的产品类似的应用(最好基于协同过滤方法)。

本文为游戏邦/gamerboom.com编译,拒绝任何不保留版权的转载,如需转载请联系:游戏邦

App Stores and the Long Tail

by Rob Lockhart

“The Long Tail” by Chris Anderson (of Wired fame) is a book about marketplaces and power-law distributions.  The basic concept is that when you can’t stock a lot of variety (like at a mom-and-pop video store), you depend on hits (products that are very popular).  When you are able to stock a lot of variety (like Netflix), not only are you able to offer niche products, but having them increases your overall business, even if you excluded the hits.

While reading, I couldn’t help but think of the iOS app store, and the place occupied by my studio, and many others, somewhere along the long tail.  Anderson points out that having variety works for the retailers, because consumers seem to be more willing to buy things which match their taste more precisely.  He says it also benefits creators because your work could potentially be listed alongside the hits.

As far as I can tell, that benefit only kicks in when the marketplace has highly visible recommendations.  Genius for Apps has been sidelined with an opt-in flow which makes it essentially irrelevant.  That means that the only ways users have of matching their taste is with the games section’s sub-categories, and word-of-mouth/press.  Word-of-mouth only goes so far with a digital marketplace scattered across meatspace, and only enthusiasts consult the games press (games almost never get covered in the mainstream press).  This means that instead of 1 list of the top-20 hits, we have 19 of them.  A little better, maybe, but nowhere near ideal.

People like to call the resulting difficulties “the discoverability problem.”  With no in-store experiences besides the top-20 lists and random app store search results (which have their own problems), people begin to get the impression that everything below the, for example, top 100 is crap.  Well, crap is in the eye of the beholder (hopefully not literally), so how can the app store hope to supply enough selection for each user?  How can Apple/Google take advantage of the long-tail economics, instead of acting as if they have limited shelf space?

Enter services with recommendation engines as first-class citizens, like Netflix or Amazon.  There’s no opt-in process, it’s just part of the service.  It’s there on the front page when you log in, and after you buy something. It’s this kind of filtering that benefits producers and consumers.  Producers can compete within the relevant microgenres, instead of everybody competing with angry birds, and consumers get their tastes catered to precisely.

Consider the Nook app marketplace.  A normal app market in every respect, except this is what you see after you buy a game:

You have to pull up the tabbed shelf, but thereafter you can easily see “Customers who bought this also bought” and “More by [Developer].”  There’s also a prominent “Share” button, which helps to alleviate the word-of-mouth problem.  It stands to reason that all this is here because these are the same developers who created the Nook’s eBook store, which is in turn based on bn.com, and they wanted to change as little as possible.  Laziness has served them well.

No one knows the power of recommendations better than Amazon.  Here’s what the kindle fire looks like after I’ve tapped to see more details about an app:

You may even recognize the text, copied and pasted from amazon.com: “Customers who viewed this ___ ultimately bought…”  Amazon has incorporated recommendations at every stage of the app-store experience.  They’re there when you open the store, when you get detail on an app (as above) and, of course, post-purchase.  Recommendations are consistently placed (the last tab on the right), and highly relevant.

So how do I know that discoverability is a real problem?  How do I know my games don’t just suck?  I guess it’s possible, but I doubt it.  Here’s why: in more mature markets with better mechanisms for discovery (the example in the book is Rhapsody), virtually every item is consumed at least periodically.  Take a look at these figures: http://daveaddey.com/?p=893 , and how the tail of the iOS app store is even flatter than a standard power law distribution (which would exhibit the classic 80/20 rule).

Here is the ideal situation, in my opinion:

1.A decent amount of space (say 20-50%) of the splash page of any app store should be devoted to automated suggestions.  The recommendations should make sense, otherwise it will be habitually overlooked.

2.There should be individual sections for different genres of app.  Each genre should have expandable lists of (a) top apps (b) suggestions within this genre (c) subgenres.

3.After making a purchase, users should be directed towards items similar to the one which was just purchased (ideally in terms of collaborative filtering).(source:GAMASUTRA)


上一篇:

下一篇: