Daily data analysis of information flow, what should be analyzed in routine data analysis of information flow?

The cost of big search has been quite good recently, but the cost of information flow has increased, especially Toutiao. Guangdiantong is basically reasonable by optimizing costs. I have talked about a lot of search skills before. This article will talk about the experience of information flow optimization. The content is superficial and is for reference only.

The first is information flow data analysis. How to analyze information flow data? I think it can be analyzed from three dimensions. The first is display. The display of information flow is much more important than the display of search. In principle, the more information flow is displayed, the better. The more it is displayed, the more people are covered, and the more people there are, the more clues there will be. The display of information flow can be measured in tens of thousands. Leaving aside niche industries, for conventional industries, if the display is less than ten thousand, it can be basically determined that there is a problem with the targeting, bidding or materials. Usually the orientation problem is more serious, so you should check the orientation settings at this time. The general approach is that the orientation should be wide rather than narrow.

The second is click . After the impressions exceed 10,000, check the clicks. If there is no lead after more than 200 clicks, it can be basically determined that the landing page is not good . At this time, you can either observe further or change the landing page. Let’s talk about the first one first: observation. When the number of clicks reaches 300 or 400 and there are still no leads, the plan can be stopped. The second method is to change the landing page. To change the landing page, you can pause the time period for an hour, then change the landing page, and then wait for the time period to open and go online naturally. Stop the period rather than suspend the plan. In the flow of information, it is better to have a pause than a schedule.

Third, the results are unstable or poor, and the money spent is neither too much nor too little. What should we do in such an embarrassing situation? In fact, there are only two ways to make adjustments. One is to change it immediately, and the other is to remain unchanged and observe for a while. When should you change and when should you observe? I think it depends mainly on your acceptable cost value and prediction. For example, if your normal cost is 80, when the cost rises to 120, you need to pay attention. If the cost is acceptable to you, don't do it. If the cost is unacceptable, adjustments must be made. On the other hand, it is about prediction, judging whether to observe or adjust based on your own experience and the data of recent days. Therefore, every promotion must develop the ability to predict.

The above mentioned when to adjust and when to observe. This paragraph will talk about how to adjust. For example, your normal cost is 80, but over time the cost rises to 120. How should you adjust in this situation? I recommend:

1: Pause the time period first, do not pause the plan, stop the time period until the traffic peaks and then reopen it . This may help. For example, the effect is not good from 6 to 7 pm, and the cost is very high, so you can stop from 8 to 9, and open it again from 10 to 12, and maybe it will be effective;

2. Copy this unit (the plan in the headline) and run them in parallel. Then compare the results . Keep the ones with good results and stop the ones with poor results. If the results are all poor, then there is a problem with the landing page, materials or targeting.

3: Observe for another half an hour or an hour, or observe another 100 clicks. If the effect is better, keep it. If not, stop it and launch a new one.

The above is a routine data analysis, and we will analyze it in units of ten thousand to see if it can be displayed. Clicks are measured in hundreds. Generally, you should pay attention to lead conversion after 200 clicks. Similarly, you should pay attention to the conversion situation for every additional 100 clicks. Effect adjustment. Effect adjustment is closely related to clicks. If there are enough clicks but no effect, you should consider whether to observe or adjust. As mentioned above, whether to observe or adjust depends mainly on your own tolerance and prediction. It is usually recommended not to observe for too long. If there is no improvement in half an hour or an hour, you should go for adjustment.

In the next article, I will talk about information flow budget, bidding, time period, verification code and landing page data accumulation. Welcome to communicate.