使用 Optimizely 进行多变量测试
我刚刚开始使用 Optimizely，并且非常喜欢它。但是，我需要进行多变量测试。有没有人想出一种方法，无论是 hack 还是其他方式，在 Optimizely 中进行 MVT？我不喜欢手动创建 25-100 个变体。或者，也许有某种方法可以自动创建变体？
I've just started using Optimizely, and really love it. However, I need to do a multivariate test. Has anyone come up with a way, hack or otherwise, to do MVT in Optimizely? Manually creating 25-100 variations isn't my idea of fun. Alternatively, maybe there's some way to automate variation creation?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
我是 Optimizely 的新用户，但我的印象是它现在具有多变量测试功能。 （至少，他们有一个标记为“多元”的选项。）
I am a new Optimizely user, but it is my impression that it now has a multivariate testing feature. (At very least, they have an option labelled "multivariate".)
编辑：Optimizely 早已支持多变量测试（至少从 2011 年底开始）。不再需要下面建议的解决方法。
原始答案为后代
保留，以防它对可能尚不支持 MVT 的其他测试平台有帮助。
所以，我一直在处理同样的问题，据我所知有两种解决方案。
只需同时对每个变量运行单独的实验即可。换句话说，如果你的两个变量是标题的大小和按钮上的语言，只需将它们视为单独的实验即可。只要它们在 DOM 方面不互相冲突（确保您的选择器非常具体，并尝试使用 ID 而不是类似
$j('p' ).eq(44);因为一个实验可能会改变另一个实验的引用。您可以生成所有的 。这不能很好地扩展，但是对于 2 或 3 个变体，我编写了一个非常简单的多变量生成器，我没有对其进行广泛的测试，但是，它允许您选择部分和不同的数字变体。对于每个部分，它将生成这些值的所有组合。在这里尝试一下： http://htmlto.com/optimizely/
EDIT: Optimizely has long since supported multivariate testing (since at least late 2011). Below suggested workarounds are no longer necessary.
Original answer kept for posterity and in case its helpful for other testing platforms that might not yet support MVT.
So, I've been dealing with this same problem, and AFAIK there are 2 solutions.
Just run separate experiments for each variable, concurrently. In other words, if your two variables are, say, the size of headlines and the language on buttons, just treat those as separate experiments. As long as they don't conflict with each other in terms of the DOM (be sure your selectors are very specific, and try to use IDs instead of something like
$j('p').eq(44);since one experiment could change which that refers to for the other.You could generate all of the permutations manually. This does not scale well, but for 2 or 3 variations, it could be ok. I scripted a very simple multivariate generator. I haven't tested it extensively, but, it will allow you to select sections and different number variations for each section, and it will generate all of the combinations of those values. Try it here: http://htmlto.com/optimizely/