Application Insights 中的 Impact 分析功能Impact analysis with Application Insights

Impact 可以分析加载时间和其他属性对应用的各个部件的转换率造成的影响。Impact analyzes how load times and other properties influence conversion rates for various parts of your app. 更准确地说,Impact 可以发现 页面视图任何维度自定义事件请求 对不同 页面视图自定义事件 造成的影响。To put it more precisely, it discovers how any dimension of a page view, custom event, or request affects the usage of a different page view or custom event.

Impact 工具

仍不确定 Impact 的作用?Still not sure what Impact does?

举例来说,假如团队中有人在争论站点某个部件的缓慢运行是否导致用户退出站点,则 Impact 就是解决争论的终极工具。One way to think of Impact is as the ultimate tool for settling arguments with someone on your team about how slowness in some aspect of your site is affecting whether users stick around. 尽管用户能够容忍一定程度的缓慢,但 Impact 可以提供见解,让你知道如何最好地对优化措施和性能做出平衡,以将用户转换率最大化。While users may tolerate a certain amount of slowness, Impact gives you insight into how best to balance optimization and performance to maximize user conversion.

不过,分析性能只是 Impact 的一部分功能。But analyzing performance is just a subset of Impact's capabilities. 由于 Impact 支持自定义事件和维度,只需单击几下鼠标,就能解答用户浏览器的选择与不同转换率之间的相关性等问题。Since Impact supports custom events and dimensions, answering questions like how does user browser choice correlate with different rates of conversion are just a few clicks away.



Application Insights 资源必须包含页面视图或自定义事件才能使用 Impact 工具。Your Application Insights resource must contain page views or custom events to use the Impact tool. 了解如何使用 Application Insights JavaScript SDK 将应用设置为自动收集页面访问次数Learn how to set up your app to collect page views automatically with the Application Insights JavaScript SDK. 另请注意,由于分析的是相关性,因此样本大小非常重要。Also keep in mind that since you are analyzing correlation, sample size matters.

页面加载时间是否影响了页面上的人员转换数量?Is page load time impacting how many people convert on my page?

若要开始使用 Impact 工具解答问题,请选择初始页面视图、自定义事件或请求。To begin answering questions with the Impact tool, choose an initial page view, custom event, or request.


  1. 从“针对页面视图”下拉列表中选择一个页面视图。Select a page view from the For the page view dropdown.
  2. 将“分析依据”下拉列表中的选择保留为“持续时间”(在此上下文中,“持续时间”是“页面加载时间”的别名)。Leave the analyze how its dropdown on the default selection of Duration (In this context Duration is an alias for Page Load Time.)
  3. 在“影响以下对象的使用”下拉列表中,选择一个自定义事件。For the impacts the usage of dropdown, select a custom event. 此事件应该对应于在步骤 1 中选择的页面视图的 UI 元素。This event should correspond to a UI element on the page view you selected in step 1.


在此情况下,随着“产品页面”加载时间的增加,“单击‘购买产品’”的转换率下降。In this instance as Product Page load time increases the conversion rate to Purchase Product clicked goes down. 根据上面的分布,可将最佳页面加载持续时间 3.5 秒指定为目标,以实现潜在的 55% 转换率。Based on the distribution above, an optimal page load duration of 3.5 seconds could be targeted to achieve a potential 55% conversion rate. 可将加载时间降到 3.5 秒以下的其他性能改进措施目前与带来的其他转换优势没有关联。Further performance improvements to reduce load time below 3.5 seconds do not currently correlate with additional conversion benefits.

是否能以自定义的方式跟踪页面视图或加载时间?What if I’m tracking page views or load times in custom ways?

Impact 支持标准和自定义的属性与度量值。Impact supports both standard and custom properties and measurements. 可以任意使用这两种方式。Use whatever you want. 如果不使用持续时间,可以使用基于主要和次要事件的筛选器来获取更具体的结果。Instead of duration, use filters on the primary and secondary events to get more specific.

来自不同国家或地区的用户是否以不同的比率转换?Do users from different countries or regions convert at different rates?

  1. 从“针对页面视图”下拉列表中选择一个页面视图。Select a page view from the For the page view dropdown.
  2. 在“分析依据”下拉列表中选择“国家或地区”Choose “Country or region” in analyze how its dropdown
  3. 在“影响以下对象的使用”下拉列表中,选择与在步骤 1 中选择的页面视图中的 UI 元素对应的自定义事件。For the impacts the usage of dropdown, select a custom event that corresponds to a UI element on the page view you chose in step 1.

在本例中,结果不再像第一个示例中一样拟合到持续 X 轴模型中,In this case, the results no longer fit into a continuous x-axis model as they did in the first example. 而是显示类似于分段漏斗图的可视化效果。Instead, a visualization similar to a segmented funnel is presented. 按“使用情况”排序,可以按国家/地区查看自定义事件转换的变体。Sort by Usage to view the variation of conversion to your custom event based on country/region.

Impact 工具如何计算这些转换率?How does the Impact tool calculate these conversion rates?

在底层,Impact 工具依赖于皮尔逊相关系数。Under the hood, the Impact tool relies on the Pearson correlation coefficient. 结果是在 -1 与 1 之间计算的,-1 表示负线性相关,1 表示正线性相关。Results are computed between -1 and 1 with -1 representing a negative linear correlation and 1 representing a positive linear correlation.

简单而言,Impact Analysis 工作原理的分解如下:The basic breakdown of how Impact Analysis works is as follows:

A = 在第一个下拉列表中选择的主要页面视图/自定义事件/请求。Let A = the main page view/custom event/request you select in the first dropdown. 针对页面视图)。(For the page view).

B = 选择的次要页面视图/自定义事件(影响以下对象的使用)。Let B = the secondary page view/custom event you select (impacts the usage of).

Impact 会查找选定时间范围内所有用户会话的样本。Impact looks at a sample of all the sessions from users in the selected time range. 对于每个会话,它会查找出现的每个 AFor each session, it looks for each occurrence of A.

然后,根据以下两个条件之一,将会话分解成两种不同的子会话:Sessions are then broken into two different kinds of subsessions based on one of two conditions:

  • 转换后的子会话包括以 B 事件结尾的会话,并包含 B 前面发生的所有 A 事件。A converted subsession consists of a session ending with a B event and encompasses all A events that occur prior to B.
  • 如果所有 A 事件是在未发生终结性 B 的情况下发生的,则发生未转换的子会话。An unconverted subsession occurs when all A's occur without a terminal B.

Impact 最终计算方式根据是按指标还是按维护分析而异。How Impact is ultimately calculated varies based on whether we are analyzing by metric or by dimension. 使用指标时,将求子会话中所有 A 的平均值。For metrics all A's in a subsession are averaged. 使用维度时,每个 A 的值为 B 的赋值贡献 1/N,其中 N 是子会话中 A 的数目。Whereas for dimensions the value of each A contributes 1/N to the value assigned to B where N is the number of A's in the subsession.

后续步骤Next steps