Data team usually has to work on three types of projects: data engineering, data analytics, and data science. Data engineering projects focus on how to effectively process and manage data, data analytics projects try to find insights from what happened so as to make better decisions, and data science projects attempt to build models to predict the future. What kind of strategies to use for a project? A strategy is an optimization since we want to optimize our return on investment. However, we have to face many constraints (i.e., money, time, resources). Therefore, a strategy is a trade-off that implies we cannot achieve all the objectives. The process of finding the right strategy involves in asking a lot of questions and figuring out what really matters for a project. The outcome of this process is some guidelines that effectively guide us when working on projects. We usually follow three guidelines: flexible, convenient, and scalable. We built a unified s3-based data platform to achieve one source/format goal. An S3 object stores data in JSON format and some user-defined metadata (i.e., MD5, content size, last modified, UUID) for verification. The key idea is to use a relational database to keep S3 object key, S3 user-defined & system-defined metadata, a logical name (consists of a few fields), documentation, etc. A unified data platform can easily support data analytics and data science tasks well. JSON format is convenient for a human to understand data. For efficient processing, data can be converted into Parquet format via Spark.