Tips for preventing malicious clicks in SEM promotion!

Those of us who do SEM promotion know that each keyword is converted by clicking on our ads to enter the website, and behind each keyword there is corresponding IP data as the source that is counted by us.

Because search engine promotion is mainly based on a pay-per-click model, there have been many cases of malicious clicks and unlimited keyword brushing , which are what we often refer to as bad points and keyword brushing. Here we will share with you several knowledge points separately, how to prevent malicious clicks and check and handle abnormal IPs from keyword and search term data analysis to IP data analysis.

First, start with the search term, determine its structure (such as the combination of verbs and nouns) and intention, whether it conforms to the search habits of netizens, enter the search in the search engine, and check whether the drop-down words are popular search terms . Generally, the drop-down words will also be sent to the hot search due to brushing words, and then further determine the needs of netizens; then analyze the SEM advertising position of the page that comes out of the search results:

① If it is determined that the search term does not meet the needs of our industry and netizens, it can be rejected to avoid being swiped all the time; if there is no advertising space for the search term in the searched page, it should also be considered rejected, because it can still be clicked in this case, which further proves that using software to swipe keywords is a malicious swiping behavior.

② If we judge that the search term meets our needs and the SEM ranking is relatively high, we will lower the bid at this time to reduce the possibility of being clicked.

1. What are the criteria for using spam words?

For example: if a keyword is clicked multiple times within half an hour (ranging from 8 to 10 times), then we need to determine whether there is a situation of word-spamming. If so, we will start to work on word rejection. If not, keep monitoring. With a deeper understanding of their respective industries, it is easy to tell at a glance whether many words are spammed or not. Therefore, it is very important to predict in advance.

2. Determine the source of spam , whether the search term is a planned keyword, and whether the word is not in the account and does not affect the normal delivery of the keyword;

3. When the search term is also a keyword in the plan , first determine the source of the click, then find the corresponding keyword to reduce the price or modify the matching method;

4. Determine whether the search term will match other keywords in the account . If it can match, then simply lowering the price cannot solve the problem. In order not to affect the normal delivery of other keywords, directly delete the keywords in the plan and add precise negations to other normal plans, and put the keywords that are being brushed into a separate plan or unit for separate optimization and monitoring.

5. Placing keywords in a separate word plan or unit does not mean that they are ignored. It is also necessary to observe whether the matching method of keywords in the unit is set correctly. If the keywords in the separately optimized units have similar parts of speech , this will still lead to the continued mutual brushing of keywords within the unit. It is necessary to adjust the same type of words to exact matches and lower the prices, only in this way can the prevention and control of word spamming be completed.

1. For brushing words, generally speaking, when the number of brushing words accumulated in a single plan or unit is large, these keywords need to be grouped and processed. Generally, it is a separate plan, divided into multiple units , and the same type of words are divided and optimized separately .

2. Regarding the keywords that have been brushed, you can consider calling them back after a period of time. When calling back, be careful to try to select a few keywords each time to conduct a small improvement test .

3. When there is no problem with the keywords after the callback, it should be done in two steps. The first step is to callback until the display ranking and clicks are normal, and then observe for a period of time before considering putting it back into the original plan unit. The keywords that are put back must be labeled with keyword color to facilitate later inspection and analysis. The second step is, if there is still a problem with the keyword you are trying to call back, set it back to a low price and try calling back with other keywords.

4. Put the word back into the original plan, and pay attention to whether there is a negative word for the word in the unit or plan. Either change the word to an exact match, or clean out the negative words and put them back into the original plan and unit.

1. In most cases, Baidu Statistics is used as a data monitoring tool. We usually analyze the IP data within a fixed time period (such as 7.15.30 days) and then sum it up.

The above screenshot is the header content of IP data optimization, which can be adjusted according to actual conditions. However, several major data sources that are absolutely indispensable are: time, region, keywords, search terms, access IP, and access identification code . At the same time, this table also serves as the source data table (historical data table).

2. Monitor and analyze IP data

After the table is created, you can sort the IP data through function formulas to find the collection of these IPs within a period of time, and then observe the situation of the IP data. If you have any questions, you can mark the color as the focus of the inspection .

It is generally believed that if the IP segment data is close or basically consistent, and the data volume or access volume is large, it is preliminarily judged as abnormal traffic and needs to be marked for further monitoring. Filter out the IPs that are initially judged to be abnormal, then put them into the source data for investigation and check the problems one by one.

① To judge abnormal traffic , you need to put the data back into the source data table (or regular period of data) and then make comprehensive judgments and adjustments. Different (but similar) IPs or IP segments, with the same or similar search terms, close visit times, and the same visit regions can be preliminarily judged as abnormal and then judged in conjunction with other data.

Note: IP represents human behavior and can be searched with multiple keywords. In other words, there may be multiple keywords behind one IP, and the IP segment represents more traffic. Therefore, before blocking an IP address, comprehensive, multi-dimensional data judgment must be performed before execution.

② (Take investment and franchising as an example) The IPs corresponding to the above search terms are similar or identical, and it can be determined that one or more IPs are spamming the terms, which is abnormal traffic. However, before everything is confirmed, we need to do a comprehensive analysis of several data before we can draw a conclusion. Never say that it is better to kill by mistake than to let go. Often IP blocking and monitoring is a double-edged sword. If you block too much, more useful traffic will be lost.

So next, we need to do one more step.

From the source data in the above figure, we can see that the search terms are basically all from a few words. In this case, we can skip the IP analysis and directly negate the keywords. At the same time, we should pay attention to whether these keywords are keywords in the account. If so, consider optimizing and monitoring them separately to ensure that these traffic will not affect other units and plans.

If the keywords are different, observe the IP data, rank the top few with the highest clicks, and look at the corresponding search terms. If the amount of data after summing up the search terms is large and the access times are close, it is recommended to accurately negate the terms within the account. Because an IP address can represent multiple keywords, before you can completely identify the abnormal IP address, you should first deny the keyword if it is available.

③ Shield the identification code: The identification code can be understood as a device. Generally speaking, a computer can change multiple IPs. The identification code is relatively unique. If necessary, shield the identification code to prevent abnormal traffic.

④ The important role of bidders in account optimization operation record sheets also applies here. After the negative word, it is recommended to record the relevant IPs in a table. After a few days, check whether these IPs continue to enter, as well as the time, region and number of entries, and finally determine whether to block the IP operation.

① After analyzing all regional data within the time period (7.15.30 days), we can get the source data of the region. The regional data may have a certain error with the data in Baidu backend. After all, the accuracy of Baidu Statistics is inherently unstable and is also related to problems with the installation code and the website, so it cannot be used as the only reference standard.

② As shown in the figure above, if you find abnormal regional data, check the IP and search terms, and refer to the operation steps of the above two knowledge points for inspection. Many times, for some mature bidders, they can tell whether the traffic is normal by looking at the regional data.

③ If it is normal, calculate the proportion of each region, and then download the regional consumption report from Baidu backend to analyze its rationality. There may be some errors, as long as they are within a reasonable range.

④ If you find that the regional IP is abnormal , you should filter out the IP with the most data after analysis , perform positioning tests on these IPs, find the location of the IP, and see if it is the location of the same industry, and block it. There are many IP locators on the Internet, and you can go and look for them.

⑤ Finally, compare the regional proportions of sales consulting transaction data, adjust and optimize regional delivery, increase the regional coefficient or strengthen regional creativity.

Blocking an IP address does not mean a permanent solution; instead, it is the beginning of monitoring and ensuring the ultimate responsibility for accurate traffic . When you have completed IP blocking and made your own statistical table, collecting all the problematic IPs, you must check the blocked IPs regularly after a period of time.

Objectively, the first problem we face is that the number of manually blocked IP addresses is only 200. Subjectively, we cannot ignore the blocked IPs. Many times, bidders do not pay attention to the blocked IPs and block them without any agreement, which leads to a decrease in account traffic and quality and they cannot find the reason. Therefore, it is necessary to check the blocked IP regularly. After all, when you block, it does not mean that there is really a problem. We operate through data analysis and subjective judgment. The system needs to "test" your operation in the entire environment.

For example, looking at the picture above, we can clearly see a lot of IP data that was blocked a long time ago. Whether it is the IP segment or the precise IP, the last column explains a lot of problems. After all, the system can clearly tell you whether your shielding work was correct at the time through the data accumulated in the entire promotion environment.

As shown in the figure above, the last column shows how many customers have blocked this IP data. This means that it is an abnormal traffic and has been discovered and blocked by more people. The other type is that we blocked it ourselves, but other customers did not find the problem, or we blocked it before, but later unblocked it, and then blocked it again, and the system also recorded it. The more times it happens, the more it proves that there is a problem with the IP. These data can provide strong guidance for our judgment.

Here we emphasize again that the detailed data of abnormal IPs should be recorded in a table. When it is needed later, reverse reasoning can be performed through the table to verify the judgment.

Now the IP address has been upgraded from 4 to 6. There are many tools on the Internet that can convert IP4 to IP6 with one click, so I won’t elaborate on them here. If the network has not been upgraded to IP6, you can judge and analyze it, convert it to IP6 through tools, and then manually add it to the background.

1. Can you prioritize keywords and then deny keywords after analyzing the search terms ? Be cautious when blocking IPs, as the impact on traffic is far greater than that of denying keywords.

2. The judgment of blocking IP should be multi-dimensional and comprehensive before manual blocking can be performed.

3. After blocking the IP address, you need to check it again after a period of time and establish the standard for unblocking through data.

4. Develop the habit of monitoring and analyzing spreadsheet data , and ensure that Shangdun data and spreadsheet data are executed synchronously.