How to judge whether the users of the promotion channel are real?

Channel analysis probably includes, for example, new users, active users, startup times, one-time ratio ..., and whether the time for users to activate the APP is normal; Define a core operation index to track the proportion of users' core operations from various channels and maintain the proportion of users' core operations. Generally, the whole wall, machine brush and meat brush are only started and maintained, and the cost of core operation is relatively high. To some extent, this ratio can be used as the standard of channel user quality. Some operators do APP promotion, and every channel is put in, with a particularly high number of clicks, but the activation amount is only single digits. It is also possible that the number of clicks is high, but the retention rate is low. All the expenses were spent, but the effect didn't come out. I can't draw a conclusion by analyzing the background data myself. What Romero wants to say is that the premise of analysis is to get reliable data. If the data is not accurate, the conclusion can be imagined ~ How to obtain accurate data can refer to my previous article on the selection of statistical platforms. Of course, no matter how reliable the platform is, there may be some unreliable situations. Why? As the saying goes, where there is a list, there is a brush list, where there is a platform for data statistics, there is a workshop for data cheating. In the mobile internet ecosystem, there are many unknown channels to brush the single studio, which contributes the same low-quality user data at a very low price. The SDK of the early statistical analysis platform is based on the jason data package of plain text. Studio can easily use programs to forge these data packets to simulate user data, such as new, active, reserved and duration. With the development of statistical analysis platform, many analysis platforms have launched SDK based on binary protocol, and developers can also call encryption switch by themselves. The improvement of these technologies improves the security and data accuracy of the statistical platform. If the App is upgraded to the SDK of the security protocol version, it is difficult for the brush studio to directly simulate the brush in the form of data packets.