-
Yesterday, I had a little discussion with a friend on the issue of SEM management of very large accounts. I thought it might be useful to everyone, so I compiled and added it and posted it.
The account structure involved in the Q&A is actually the most important part of large and medium-sized account management. It is not discussed this time. If readers are interested, they can discuss it in a separate article.
Question: How can we group super-large accounts with more than 100,000 keywords into groups for easier management? Now this account is mainly divided by type, such as industry/general/question/competitive products, etc., but the management feels very cumbersome.
Answer: For large accounts, account structure is the most important part of SEM management. In essence, account size and the number of keywords have no essential difference in SEM management. The main problem with large accounts is that there is too much content and it is easy to get lost in details, making it impossible to effectively identify priority issues and solutions. The purpose of a clear and reasonable account structure is to improve the readability and operability of the account, and if it is prioritized, management will not be too cumbersome. In principle, I think that even if there are millions of keywords, you don’t need dozens of people to operate them. Three or five are enough.
Question: Yes. Now there is a problem. My client here has a few little girls who are doing it by themselves, and they often take out the low-quality ones and rebuild the plans again. As a result, I almost went crazy when I went in. It was a mess. If there was a lot of movement, the data would probably fluctuate greatly.
Answer: When taking over a new account, you should avoid making big changes. Too many moves at once can easily cause violent fluctuations. Even if you create a new account structure, it should be done step by step. For customers, it makes more sense to optimize on a stable basis. Severe fluctuations themselves can also create additional obstacles to optimization.
Question: Is it OK to just adjust the top 100 consumers? I plan to filter out the top 100 consumer keyWords and related search terms, and then regroup them.
Answer: You can do this, for example, choose the largest project industry as a sample, select the most meaningful keywords, and reorganize a test series. Then optimize this series. After completing the initial optimization, replace other projects/industries with the same test group concept, little by little, don't change it all at once.
Question: I’ve seen the PPT you used to write for Baidu’s intermediate training to test a set of keyword combinations. Isn’t this difficult to achieve without a system?
Answer: It’s not necessarily difficult, Excel is enough. The main thing is that you have to set up the structure yourself first. When you do the test, you will have a rough layout, and then gradually adjust it.
Question: Yeah. Do you have any samples of EXCEL data analysis here? Let me see how you do it and learn from it. Just DEMO data.
Answer: No... This is not complicated at all. It is basically a simple sorting and effect comparison. But it requires experience, right? As for the structure, in addition to organizing it yourself, you can also talk to their webmaster or business leader to echo the website structure or business process, with the goal of report readability and operation ease of execution.
Question: I would like to know what aspects of data fluctuations you pay attention to when doing these optimization tasks. If you look at a table, it is very intuitive.
Answer: Of course it’s a comparison of the effects of the old and new series. This is the purpose of making a small test group, ensuring that the data before and after the test are readable and comparable, using consumption and conversion as the main indicators, and displaying clicks and CPC as a means.
Q: Ad groups also need to ensure relevance, right? The correlation between creativity and keywords, if the purpose is different, the structure will be different.
Answer: Your test series must be designed with optimal goals, otherwise it will be meaningless. After the test is successful, the old words and old series will be gradually copied and eliminated when replacing, so as to ensure that the new structure is clear and controllable.
Question: Yeah. Understood. Sometimes it is good to start planning, but later optimization is likely to mess up the account.
Answer: Well, this is a matter of experience. I still don’t have clear ideas, and when there are too many things, I lose my priority. To do SEM management, you need to be able to get in and out, always understand the core of the problem, and be able to grasp the main line at any time. If the outline is too broad, you will not be overwhelmed by the large number of details in the later stage.
Q: There is another situation where I keep adding high-quality search terms during testing, which makes comparison difficult. If you don’t add it, you can only optimize it in CPC, creativity, and LANDING.
Answer: The test is only for the old ones, don’t complicate it. Word expansion is a matter of replacing the old keywords and the old series. Optimization must be implemented step by step, targeting a limited number of targets each time, otherwise you won’t know how to evaluate the effect. You are a little anxious about optimizing your account and always want to get it right in one step.
Question: Yeah. What are the general steps here? Yes, I am not as mature and sophisticated as you. The main reason is that customers are also eager for quick results and have no choice.
Answer: Some tasks cannot be rushed. The key is to confirm that optimization and optimization methods can be replicated. This requires data accumulation to verify. There is no point in just being anxious. You cannot win a lottery ticket twice. Reproducibility is the key to system optimization.
Question: Could you please provide some guidance on the specific implementation steps. I will record it carefully and execute it accordingly to see the effect.
A: First, address your priority goals. Sort from the two perspectives of consumption conversion, start with the most influential series or keywords, and establish an ideal test group or test series. Google European and American accounts now support the test function in the series settings, which can be divided into tests in the running state. Chinese users who use Baidu or Google China use traditional methods: they stop the old series of keywords, start the test series, and compare the effects of the same time period before and after the test.
(Click to view larger image)
Question: Google's test is to divide the old and new series/groups into two different CPC bids and see the final comparison results?
A: It’s not just CPC. In fact, in the first step of optimization, CPC is not important, what is important is the structure. You have to achieve the optimal structure in your test series and then adjust CPC, ads, LP, relevance, etc. After preliminary testing is completed, this model will be copied to other target series. In the new series that has completed the replacement of the old with the new, you can further optimize the words, adjust the budget, improve the advertising, etc. In fact, it is not complicated at all. The main thing is to have a clear idea and know what you are doing at every step, whether the goal is achieved, and whether it matches the means of implementation.
Question: Yes, I’ll record it. There is another problem. Some industries themselves are highly volatile. For example, it is possible for this week to fluctuate by 30% compared to last week.
Answer: First, extend the observation period. The second is to lock the series, grouping or keywords, and eliminate fluctuations to see the overall effect. Fluctuations are normal, but all grouped keywords fluctuate together, which is abnormal if there is no external influence.
Q: Overall volatility is really normal in some industries.
Answer: You don't understand what I mean. If you think a certain fluctuation is normal, just put this fluctuation range into effect observation. This has the same meaning as extending the observation period. That is to say, if the 30% fluctuation is normal, then our optimization can target 50%. On the other hand, you can weight all data using a 7-day average and a 15-day average to weaken the impact of short-term fluctuations. In the final analysis, it is still a question of thinking: you decide on a structure and optimization plan, how to implement it and evaluate the effect.
Question: Thank you so much. After listening to what you said, I feel that my mind has opened up. I am always confused about how to eliminate the influence of normal fluctuations from the weighting of data analysis.
Answer: You're welcome, be patient, the main thing is to identify the problem, determine the method to solve the problem, and confirm that the problem is solved. If you start small and use the right method, small improvements can be replicated and enlarged. But if you go through a lot of operations at once, and you don’t know which ones are useful and which ones are not, you won’t be able to see clearly at all, and it will be useless for customers to worry about it.
Question: Should we not put too many keywords and groupings into the test series?
Answer: Based on your ability to manage effectively, if you are confident that you can manage a large number of keywords at the same time, don’t be afraid to add more. But generally speaking, the more keywords there are, the higher the requirements for SEM management experience.
Article source: semwtach.org