技术控

    今日:35| 主题:49312
收藏本版 (1)
最新软件应用技术尽在掌握

[其他] 2 A/B Testing Failures and 1 Win: Why A/B Testing Failures Are as Important as t

[复制链接]
懂我与否 发表于 2016-10-20 01:21:19
130 1

立即注册CoLaBug.com会员,免费获得投稿人的专业资料,享用更多功能,玩转个人品牌!

您需要 登录 才可以下载或查看,没有帐号?立即注册

x
You’ve probably seen a lot of blog posts floating around the internet about A/B testing successes. This blog is no exception.
  But hardly anyone talks about their failed experiments. I don’t blame them–it’s hard to expose to the world that you weren’t right about something.
   Which leads us to believe…is anyone running failed, insignificant tests? I mean, if no one’s talking about it, it must not be happening, right?
   Let me tell you a secret: Everyone is failing. Everyone is running or has run an experiment that got them nowhere.
  However, at Kissmetrics, failure is part of our A/B testing process. If none of our tests fail, we know we’re not running enough tests or our ideas are too safe.
  In fact, the bigger the failures, the closer we are to an even bigger win.
  We’re never 100% correct about our hypotheses. No matter how many years of experience you have, no matter how much you think you understand your customers…there’s always room for learning.
  Otherwise, why would we test in the first place?
  Now let’s take a look at a couple of our own failures so you can see what I mean.
  Failure #1: Too much white space on our product page

   Test Hypothesis:There’s too much clutter at the top of the page. By removing the background image and reducing white space, we’ll make the page copy more visible, enticing people to scroll down and interact more with the page.
  You already know that this test failed, but just from looking at the hypothesis–do you know why?
  I’ll give you a hint: it has a lot to do with data.
  We technically had data that indicated a dip in conversion on this page.
      Control:
      
2 A/B Testing Failures and 1 Win: Why A/B Testing Failures Are as Important as t-1 (why,animal,testing,why,is,r,d,important,why,art,is,important,why,it,is,important,looks,are,important)
           Variant:

2 A/B Testing Failures and 1 Win: Why A/B Testing Failures Are as Important as t-2 (why,animal,testing,why,is,r,d,important,why,art,is,important,why,it,is,important,looks,are,important)
       However, we didn’t have evidence that people weren’t scrolling down, or that the space at the top was stopping them from converting. When a hypothesis has little or no evidence, we have a slim chance of winning the test.
  Results:

   Signups:
  Improvement over original: 4.41%
  Certainty: 55.27% for the variant
  What we learned:

  No significant data here.
  Having a hero image or not in this page won’t influence our conversions. In previous homepage tests, our hero image mattered, but perhaps not on this exact page.
  In previous tests on this particular page, we experimented with the copy and overall messaging. Therefore, our next test should be around the copy to see if we’ll get a lift.
  Failure #2: Copy and images on the product page

   Test Hypothesis:A more benefits oriented product landing page will lead to better quality leads.
      Control:
123下一页
友荐云推荐




上一篇:A new ecosystem for Haskell: the JVM
下一篇:A comparison of state-of-the-art graph processing systems
酷辣虫提示酷辣虫禁止发表任何与中华人民共和国法律有抵触的内容!所有内容由用户发布,并不代表酷辣虫的观点,酷辣虫无法对用户发布内容真实性提供任何的保证,请自行验证并承担风险与后果。如您有版权、违规等问题,请通过"联系我们"或"违规举报"告知我们处理。

忆柏 发表于 2016-10-27 05:43:30
支持楼上的!
回复 支持 反对

使用道具 举报

*滑动验证:
您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

我要投稿

推荐阅读

扫码访问 @iTTTTT瑞翔 的微博
回页顶回复上一篇下一篇回列表手机版
手机版/CoLaBug.com ( 粤ICP备05003221号 | 文网文[2010]257号 )|网站地图 酷辣虫

© 2001-2016 Comsenz Inc. Design: Dean. DiscuzFans.

返回顶部 返回列表