如何让Uber标准化移动分析以获得跨平台洞察
At Uber, we prioritize the end user experience. To identify poor or broken experiences, we must understand how people interact with our apps throughout their journey. To accomplish this, product teams instrument UI components so the apps emit analytics events whenever a rider, driver, Uber Eats user, or courier views or interacts with them.
在Uber，我们优先考虑最终用户体验。为了识别糟糕或破损的体验，我们必须了解人们如何在整个旅程中与我们的应用程序互动。为此，产品团队对UI组件进行仪器化，以便在乘客、司机、Uber Eats用户或快递员查看或与它们互动时，应用程序发出分析事件。
At a high level, here’s the workflow for how mobile engineers add mobile analytics:
从高层次来看，以下是移动工程师添加移动分析的工作流程：
- Definition. Engineers design an event and any custom metadata, define it in a generic Apache Thrift™ schema, and assign it a UUID that’s later mapped to a human-readable name in the back end.
- 定义。工程师设计一个事件和任何自定义元数据，在通用的 Apache Thrift™ 架构中定义它，并为其分配一个 UUID，后者在后端映射到人类可读的名称。
- Generation. The Thrift schemas are code-generated into native Swift® and Kotlin® models and committed to the mobile monorepos.
- 生成。Thrift模式被代码生成到本地的Swift®和Kotlin®模型中，并提交到移动单体库。
- Instrumentation. Engineers attach the event and its metadata to the UI component that emits it.
- 仪器仪表。工程师将事件及其元数据附加到发出事件的UI组件上。
- Emission. When someone interacts with the UI, the event is generated, batched, and logged together with common app metadata from a central reporting library. Events that aren’t actively used can be disabled client-side, and re-enabled later remotely by UUID.
- 事件发射。当有人与用户界面交互时，事件会被生成、批处理，并与来自中央报告库的常见应用元数据一起记录。未被主动使用的事件可以在客户端禁用，并可以通过UUID在远程重新启用。
- Processing. A central back-end service receives the event, enriches the data, and routes it to the appropriate pipelines.
- 处理。一个中央后端服务接收事件，丰富数据，并将其路由到适当的管道。
- Consumption. Engineers, data scientists, product managers, and marketing managers access the data via online and offline sources.
- 消费。工程师、数据科学家、产品经理和市场经理通过在线和离线来源访问数据。
The data produced by this pipeline is critical for deriving insights, monitoring feature health, optimizing the user journey, and personalizing the app with ML-driven recommendations.
该管道生成的数据对于获取洞察、监控功能健康、优化用户旅程以及通过...