There are often complaints from classmates: I feel that I usually do descriptive statistics, year-on-year, month-on-month, how to do in-depth data analysis? Explain systematically today. Not much to say, go straight to the dry goods.
As a simple example, let’s analyze: Why did performance drop. The practice of many students is to compare this month with the previous month, and then make crossovers by product, region, branch and other dimensions. In the end, it was found that the performance of product A dropped by 10%, and the performance of product B dropped by 6%… One more step, it may be counted as a 5% drop as a whole, and then each product dropped by more than 5% is marked as red. This is the end of the analysis.
Common descriptive statistics
The focus of further analysis is to find enough data evidence. For example, there is a hypothesis here: the reduction of marketing coverage products leads to a decline in performance. Then it depends on the data, whether the decline in inactive products is very strong, and the decline in active products is very small. Similarly, if there is evidence for each hypothesis, a conclusion can be drawn: the lack of marketing efforts has led to the problem of performance decline.
Third, the second step of improvement: focus on key issues
Note that more than one direction is assumed. For example, after we come to the conclusion of “insufficient marketing efforts”, everyone will naturally ask: Is there only a marketing problem? Is there a problem with branch management? Is there no problem with the product?
In this step, you need to help you clean up other assumptions and focus on the core issue. Here’s a simple way to judge: which problem has more impact.
For example, I want to prove that the impact of marketing efforts is greater than the impact of products. So the assumptions I’m going to list are:
The analysis of this step requires a large number of positive and negative examples and evidence, which is very energy-consuming. Analysts are required to have strict logical sorting and a large number of detailed data demonstrations. And in this process, it is very likely to find a large number of special cases, making it difficult to draw conclusions. For example, it seems that the decline in marketing efforts has the greatest impact, but some products are very strong, and some branches are always rotten.
If there are many special cases, it is actually a good thing, indicating that performance is not affected by a single factor. At this time, it is necessary to use the MECE method to sort out the logical relationships of the special cases one by one. (The construction method of MECE will not be repeated here, and interested students can see the previous sharing). The final effect is as follows: