Figure II
Looking at Figure 2, we got the sales and business email list refunds, and found that the sales in January and February were not high, but the refunds were very high. The sales began to rise in October. At this time, we have already refunded the refunds in Figure 1. The rate change creates a "skeptical" attitude.
In Figure 3, I changed the formula for calculating the refund rate to an algorithm, [refund rate = refund amount of the current month / sales of the previous month], and the overall refund rate has stabilized.
Students in the e-commerce industry probably know that the first half of the year is a low season and the second half is a peak season. Cross-border e-commerce has a long distribution cycle, and refunds for orders often occur across months. For example, for orders in December, refunds are concentrated. In January and February, the sales in January and February were not as high as in December, so the refund rate in January and February in Figure 1 was too high. The algorithm I described is not the most reasonable algorithm, but a tricky way. We need to optimize the specific algorithm based on the background of the industry.
Data analysis that is divorced from actual business is meaningless. We must attach context to any data when we look at it. After drawing plausible conclusions, we still have to maintain a "skeptical" attitude towards the data.
4. Maintain a "skeptical" attitude towards the leadership's arrangements
Maintaining a "skeptical" attitude towards the leader's arrangement allows us to have a "upward management" mindset. We must gain insight into the meaning behind the leader's instructions. If the work arranged by the leader is difficult to carry out, we must give timely feedback and reach an agreement with the leader. .
Here's an example of a scene: