When I first served as the product manager responsible for the SDK direction, I was actually full of misgivings. The biggest misgiving was "SDK still needs a product manager"? In practice, I discovered that the SDK needs a product manager very much. This is a deep participation. The first cognitive change after the work of the SDK.
Unseen demand
SDK is the core of this link of data collection. Asking it to be so clear, it's a source of fresh water. SDK is the living water of this source. From a technical point of view, the design of Shence Analysis itself is relatively sophisticated. In order to ensure strong performance even in the case of privatized deployment, it will try to avoid putting ETL-like logic into the system. The data quality reported by the SDK puts forward new requirements, and the report can be used.
From a business perspective, tangible requirements such as visualization of full buried points, heat maps, and Deeplink require a product manager to design the entire requirement. There are also preset all-embedded events and preset attributes that seem to be very technical, but in fact, the module that affects the entire data foundation needs to be responsible. Even if it is a name that seems to be a small problem, it will be met if it is not done well. To the big problem.
To give a simple example, the activation event Appinstall was not directly designed as a preset event when it was originally designed, so the system analyzed by Shence has no reliable basis to determine what event is the activation event. Is it possible that the two-end reporting is inconsistent due to the customer's mistakes, leading to the need to make an interface for users to clarify the name of the event when the channel is iterated? This is actually very painful, all of which stems from the earliest difference in thought.
As the product manager of SDK, the first task is to start a variable name that can be used for 100 years, and establish a reliable audit mechanism. This kind of demand may seem to be unsatisfactory at first glance, but it is objective. More data access requirements of various SDKs and the mechanism of writing preset attributes in preset events are important requirements that affect the statistical caliber. These requirements may actually be an extra attribute of the preset events, which requires technical consultants, long-term technical research on R&D, and business logic designed under many restrictions.
Visible demand
The most obvious of the visible needs is that the visualization is fully buried. Visualization of full burying points has always been a pain point. To make a good visualization of full burying points, you need to rely on the power of SDK development. However, it is not enough to rely on SDK development alone. It also needs the support of front-end and back-end and QA (Quality Assurance Department) students. The sense of responsibility of "starting from SDK, not limited to SDK", the secretary embodies incisively and vividly, from copywriting to interaction, project progress to quality management, the secretary participates in the whole process, not only from the perspective of SDK, but also from the perspective of the entire project. From a perspective, the project management is well organized. As a product manager, I can concentrate on solving business logic and business delivery issues, which is actually very "cool".
In fact, there are many students in SDK development who are like secretaries who can do "start from SDK, not limited to SDK". For example, Wei Zhangxiang, who is now in charge of the project management of "visualization full buried point", is like a secretary. Yang Zhankun who organizes platform product project management affairs. It is this sense of responsibility that allows the project to be implemented, so that everyone can grow and go further.
Better data foundation
Wen Feng mentioned data foundation in many internal and external speeches. In fact, our data foundation is composed of many things, including SDK, SDG, data stream and query and storage engine. In order to ensure performance and interpretability, we often try our best to avoid complex ETL behaviors and data modification behaviors. A lot of data can be said to be open without turning back.
How to build a better data foundation, the SDK is the first team in the forefront. I hope that in the coming time, we can take "allowing customers to obtain the most and most structured data with the least configuration and the lowest access cost" as the first value, and build the best open source SDK in China with the team. All users who purchase Shence or those who do not use Shence SDK.