This time, we will talk about how product managers encounter data abnormalities, how to do data analysis and find out the reasons. Once, as soon as I went to work on Monday, the leader asked, why did our App DAU drop by 30% this weekend? At that time, I had no experience, and I was in a hurry. I looked at the data, and looked for technical and operational discussions. Fortunately, I found the reason. At work, product data anomalies occur from time to time, and interviews are often asked. In particular, changes in core data (such as new additions, daily active DAUs, orders, income, etc.) are the focus of our attention. Once the data is abnormal, the product manager needs to analyze the data, investigate the cause, and solve it as soon as possible to avoid greater impact and loss. In this case, if you have no experience, do not know how to deal with it, and sometimes toss for a long time, you may not be able to find the reason.
Later, I was in charge of the data statistics platform, often helping product managers and operational analysis to answer various data changes. This gave me experience in data analysis. To analyze the abnormal situation of this kind of data, the conventional practice can be summed up in three steps. 1. Check common problems and confirm the accuracy of data The first step is to troubleshoot common problems to ensure that the data is accurate, and it makes sense to analyze it. These problems are mostly caused by version updates , changes in statistical methods , and service failures . Therefore, you can first see if there are any updates to the front and back ends of the product. If there are updates, you have to check with development and testing colleagues to see if these updates will affect the data.
Secondly, find out whether the data is accurate and whether there is a possibility that there is a problem with the data statistics. Most products now use the data statistics platform. If you use your own system, you have to confirm with the developer whether there has been any change in the recent data collection, reporting or statistical logic? Will there be mobile number list service updates or exceptions that cause problems with data statistics? If you use a third-party statistical platform, you should confirm with the other party whether there are any changes or abnormalities in the platform? If it is a manual statistical report, do you need to find a statistical colleague to verify the statistical method and source data? Caused by changes in statistical methods or service failures, usually multiple products are affected, the data changes are consistent, and start to drop at the time of change or failure.
For example, all product data fell at 8 points, or all were 0. Early products are prone to problems in this step; while mature products have a stable statistical platform, and the probability of problems is small. 2. Compare historical data to determine whether it is abnormal or fluctuating The data of some products will fluctuate periodically, which is a normal phenomenon. For example, for products in the workplace and office, the daily activity decreases on weekends; for student-oriented products, the daily activity will drop when the school starts, and the daily activity will rise again when there is a holiday.