Delivering a use case（交付一个用例）¶
To create value for your organization using the Palantir platform, you need to work across the platform to build tools that support operational decision-making processes. A use case is a time-bound effort by a dedicated team to support a specific decision-making process. Use cases are central to the delivery of new capabilities on the platform for a set of users.
Examples of use cases¶
Use cases in the Palantir platform can involve a wide range of different activities and workflows, such as:
- Generating, investigating, and resolving alerts related to important operational processes.
- Optimizing inventory across a network of your facilities to improve resilience and mitigate supply chain uncertainty.
- Helping you make decisions about how to optimize the allocation of salespeople to all the different regions you serve.
Use cases require a structured and thoughtful approach to building in the platform along with gaining an understanding of some new ideas and terminology. You can find all the details of our solution design methodology as well as case studies and reference architectures in the use case lifecycle section.
A data-driven mindset¶
Imagine a world where data scientists, quality analysts, assembly line workers, and executives use data as their daily communication language. Palantir turns data into a communication language by:
- Maintaining data lineage and attribution so people can trust what they discover and learn.
- Meeting users where they are, regardless of technical ability or experience, so everyone can bring data into their daily work.
This vision of collaboration is the driving force behind the Palantir platform.
The Palantir platform was built to create data-driven loops across an organization, in sync with your colleagues and collaborators: use data to make a decision, capture the decision that was made, and then use data to assess the decision's impact over time. Rather than relying on emailed spreadsheets and static analyses, you and your colleagues can collaborate directly on data in real time.
Project success on the Palantir platform involves creativity and thoughtfulness. There are often several solutions to any given analytical question, organizational workflow, or operational need. To choose the right path, you will need to balance three factors: the desired outcome, the available data, and platform tools.
Outcome¶
As you break down your project, it's more important to think about the outcome rather than the method you use to get there. For example, instead of starting with a need to build a sales dashboard, seek to understand what decisions and outcomes your work might enable. For example, is the outcome about making decisions on time and resource allocation to different sales regions, or something else?
This level of understanding might involve more work at the start of your project, especially if you are building a tool, report, or analysis for someone else to use. Consider an outcome-oriented framing to help you move towards a realistic goal.
Flexibility and adaptation can help ensure a successful Palantir project. Identifying a clear, outcome-oriented goal makes it easier to break your project into small, logical steps. This problem decomposition is an important skill, as projects will often require multiple sources of data and several platform tools working together.
Data¶
Figuring out the right data to back your project can be a daunting task. However, if you have an outcome-oriented framing and decompose your project into smaller steps, it's easier to work backwards from that outcome and identify the necessary data.
If your organization has been using the Palantir platform for a while, the data you need may already be in the platform. Try exploring datasets curated in the Data Catalog or objects and links in Object Explorer. From our outcome example, we might identify that we need data on our sales force, sales territories, products, and individual sales. Each of these objects should have a primary representation in the Ontology.
If you are unable to identify the key dataset for each type of data you need, contact your platform administrator. Sometimes, it is necessary to expand the Ontology to include new organizational objects or to add new properties to those that already exist. As we will discuss later, you can use the tools in the data integration layer to connect external data sources and bring new data into the Palantir platform.
Tools¶
Each application is designed to operate as part of the entire platform. It will take time to familiarize yourself with the different capabilities in the platform and which tools are best for a given job.
Once you understand the outcomes of your project and the necessary data, it's easier to map each step to a particular tool. For instance, suppose that in your project you recognize that one sub-project is to generate new sales metrics for each region. This sub-project creates several additional steps:
- Identify the key relevant metrics
- Source the data
- Develop logic to transform the data
- Develop logic to aggregate the data into the metrics
- Display the data for consumption by the sales team
Each of these steps will map to different tools in the platform, and the correct tool may vary as your project matures. For instance, you might start by prototyping transforms and metrics in Contour, an application for point-and-click analysis and data transformation. Contour makes it easy to understand the shape of your data and generate charts or metrics. You can add these metrics to a dashboard and create a quick prototype for the sales team to provide feedback. This could be a great end point for this project: a few well crafted dashboards that provide new insights to drive decisions for the sales resource allocation process.
For larger projects or those focused on production usage you can convert your logic to a pipeline in Code Repositories. There, you can collaborate with other technical users and use robust platform tools to schedule your data to update regularly. To create a more tailored user experience, you can set up Object Views or build a custom application in Workshop or Slate to enable the sales team to not only view data in dashboards, but also capture their decisions back into the system.
While many projects in the platform will not need this level of complexity, it is useful to understand how framing and decomposing your project will help you identify the data and tools you need to succeed.
中文翻译¶
交付一个用例¶
要利用 Palantir 平台为您的组织创造价值，您需要跨平台协作，构建支持运营决策流程的工具。用例(use case) 是由专门团队在限定时间内完成的工作，旨在支持特定的决策流程。用例是为一组用户交付平台新功能的核心。
用例示例¶
Palantir 平台中的用例可以涵盖各种不同的活动和流程，例如：
- 生成、调查和解决与重要运营流程相关的警报。
- 优化您设施网络中的库存，以提高韧性并缓解供应链不确定性。
- 帮助您做出如何优化销售人员分配到所服务不同区域的决策。
用例需要在平台中采用结构化且深思熟虑的构建方法，同时理解一些新概念和术语。您可以在用例生命周期部分找到我们解决方案设计方法论的所有细节，以及案例研究和参考架构。
数据驱动思维¶
想象一个世界，数据科学家、质量分析师、流水线工人和高管都将数据作为日常沟通语言。Palantir 通过以下方式将数据转化为沟通语言：
- 维护数据沿袭和归因，让人们能够信任他们所发现和学习的内容。
- 无论用户的技术能力或经验如何，都能满足他们的需求，让每个人都能将数据融入日常工作。
这种协作愿景是 Palantir 平台的驱动力。
Palantir 平台旨在跨组织创建数据驱动循环(data-driven loops)，与您的同事和协作者同步：使用数据做出决策，记录所做的决策，然后使用数据评估该决策随时间推移的影响。您和您的同事无需依赖通过电子邮件发送的电子表格和静态分析，而是可以直接在数据上进行实时协作。
在 Palantir 平台上取得项目成功需要创造力和深思熟虑。对于任何分析问题、组织工作流程或运营需求，通常都有多种解决方案。要选择正确的路径，您需要平衡三个因素：期望的结果、可用的数据和平台工具。
结果¶
在分解项目时，思考结果比思考实现方法更为重要。例如，不要从构建销售仪表盘的需求开始，而是尝试理解您的工作可能促成哪些决策和结果。例如，结果是否涉及对不同销售区域的时间和资源分配做出决策，还是其他方面？
这种理解程度可能在项目初期需要更多工作，特别是当您为他人构建工具、报告或分析时。考虑以结果为导向的框架，帮助您朝着现实目标前进。
灵活性和适应性有助于确保 Palantir 项目成功。明确以结果为导向的目标，可以更容易地将项目分解为小的、逻辑清晰的步骤。这种问题分解是一项重要技能，因为项目通常需要多个数据源和多个平台工具协同工作。
数据¶
确定支持项目的正确数据可能是一项艰巨的任务。然而，如果您有以结果为导向的框架，并将项目分解为更小的步骤，就更容易从结果反向推导，识别所需的数据。
如果您的组织已经使用 Palantir 平台一段时间，您所需的数据可能已经在平台中。尝试探索数据目录(Data Catalog)中策划的数据集，或对象浏览器(Object Explorer)中的对象和链接。从我们的结果示例来看，我们可能需要关于销售人员、销售区域、产品和单个销售的数据。每个对象都应在本体论(Ontology)中有一个主要表示。
如果您无法为所需的数据类型找到关键数据集，请联系您的平台管理员。有时，需要扩展本体论(Ontology)以包含新的组织对象，或为现有对象添加新属性。正如我们稍后将讨论的，您可以使用数据集成层中的工具连接外部数据源，并将新数据引入 Palantir 平台。
工具¶
每个应用程序都设计为作为整个平台的一部分运行。您需要花时间熟悉平台中的不同功能，以及哪些工具最适合特定任务。
一旦您理解了项目的结果和必要的数据，就更容易将每个步骤映射到特定工具。例如，假设在您的项目中，您认识到一个子项目是为每个区域生成新的销售指标。这个子项目会带来几个额外的步骤：
- 确定关键相关指标
- 获取数据源
- 开发转换数据的逻辑
- 开发将数据聚合为指标的逻辑
- 显示数据供销售团队使用
这些步骤中的每一步都会映射到平台中的不同工具，并且随着项目的成熟，正确的工具可能会发生变化。例如，您可以从在 Contour（一个用于点击式分析和数据转换的应用程序）中原型化转换和指标开始。Contour 使您能够轻松理解数据的形状并生成图表或指标。您可以将这些指标添加到仪表盘(dashboard)中，并为销售团队创建一个快速原型以获取反馈。这可能是该项目的一个很好的终点：几个精心设计的仪表盘，为销售资源分配流程的决策提供新的洞察。
对于更大的项目或专注于生产使用的项目，您可以将逻辑转换为代码仓库(Code Repositories)中的管道。在那里，您可以与其他技术用户协作，并使用强大的平台工具来安排数据定期更新。要创建更量身定制的用户体验，您可以设置对象视图(Object Views)，或在 Workshop 或 Slate 中构建自定义应用程序，使销售团队不仅可以在仪表盘中查看数据，还可以将他们的决策捕获回系统中。
虽然平台中的许多项目不需要这种复杂程度，但了解如何构建框架和分解项目将有助于您识别成功所需的数据和工具。