How to optimize keywords and rank on Baidu’s homepage?

In the past few years, there are some things that I have not dared to share. Not only because of privacy considerations, but also because if these things were made public on a large scale and everyone knew about them, it would be very difficult to predict the changes in the industry.

And at this point, writing about it again may not be a big deal.

What I want to write in this article is the theory and practical operation I used to rank the super hot word "SEO" . How competitive is this term? In the industry, there were at least hundreds of forums, blogs, training websites, third-party official websites, etc. that were ranking for this word, and the competitors were front-line practitioners and even teams.

But on the other hand, to put it bluntly, these competitors are just a group of people who don’t even understand the basic concepts.

Yes, sadly, even the backbone of this industry mostly lacks basic concepts.

Therefore, I have to start with the basic concepts in this article.

1. Summarize the search engine ranking rules in one sentence:

In the shortest possible time, let the highest proportion of users find content that meets their needs

Starting from this ultimate principle, many possible operating points can be directly inferred through deduction.

Here, for easier understanding, let me assume a situation . When a user searches for the word "Naruto" , there are only three related web pages on the Internet. At the same time, assuming that Baidu only displays 3 search results in total, then these three web pages are displayed at the same time, and their respective click-through rates are as follows:

Naruto Animationa CTR: 50%

Naruto Animation b CTR: 30%

Naruto manga CTR: 20%

What if only 1 search result is displayed in total? Among the three web pages above, web page No. 1 will undoubtedly be displayed.

But what if there are 2 search results in total? An interesting thing happened. It is obvious that web pages 1 and 3 should be displayed. This will meet the different needs of different users and maximize the overall click-through rate.

2. Behind many words, there are very different segmented needs.

For example, "Naruto" can have segmented demands such as "Naruto animation", "Naruto comics", "Naruto games", etc.

It should be easy to understand that among the 10 search results on the first page of search results, the more comprehensive the segmented needs displayed, the better. But how does Baidu determine the specific segmented needs for each word?

This involves a non-critical knowledge point, but knowing it will help you figure out the problem easily. In fact, it is also common sense in the search engine field. Most users, when they look at a page of search results and don’t find what they want, will not choose to turn to the next page, but instead search for another (usually more specific) word.

So it is very simple for Baidu. Just look at what proportion of users who search for "Naruto" will switch to searching for "Naruto comics" to know the extent of this segmented demand.

3. How did I thoroughly determine the above theory?

When I was in charge of SEO at Ganji.com, I got the rights to use the data tool Baidu Sinan because Ganji was a major customer of Baidu's bidding. At the beginning, Baidu Sinan provided a data called "subsequent search terms", which clearly listed the proportion of users who switched to searching for other terms after each term.

For example, when you first see the word "recruitment", subsequent search terms include "recruitment website", "recruitment job search", "recruitment 58.com", etc. By comparing these subsequent search terms with the specific rankings, we found that, as expected, the ranking of the corresponding web pages is basically the same as the proportion of the subsequent search terms.

Occasionally there will be a slight deviation of one or two places, and it is not surprising if it is larger, because there is a special last-place sorting adjustment rule called "satisfactory click rate", which will not be expanded here.

4. I wonder what readers think when they see this. The problem is that Ganji.com is a major customer of Baidu, so I have the data, so can I figure this out?

No.

I understood this long before I saw the data. What is very noteworthy is that, up to the second paragraph, all of them are just guesses based on Baidu's ranking rules and logical deductions.

Once you have a "guess" about how subsequent search terms will affect rankings, just looking at the rankings of a few more words will be enough to determine the sorting rules.

In this process, no project experience, profound technology or complicated data were used.

5. Understanding the concept of "subsequent search terms" alone is not enough to improve the ranking of super hot words such as "SEO". Next, a concept that needs to be introduced is coarse sorting and fine sorting. I still figure it out through "logic".

However, even if you have not thought deeply about it, as long as you read a few books on search engine principles carefully, or do a simple full-text search yourself, you will definitely know the issues of coarse ranking and fine ranking. However, even though I have always believed that books on search engine principles should be read within the first week of learning, few people actually take the time to learn the basics of the basics.

Let's look at this in detail:

Assuming Baidu wants to display the corresponding web pages of 10 subsequent search terms on the search results homepage, there are several technical strategies:

Directly search for the results of the 10 subsequent words separately, and then search for the hypothetical 10 million search results corresponding to the main word on the main word homepage, then search 10 times in all 10 million web pages to get the hypothetical 10 million search results corresponding to the main word, and then search 10 times in, for example, the top 1000 with the highest scores. Obviously, although it seems that 1>2>3 in terms of search quality, the computing time cost of 1 and 2 is much greater than 3.

Up to this point, any of the solutions Baidu actually adopted is possible. If someone studies Google's rules and finds that it uses 1 or 2, it will not be surprising.

Continue using logic:

If, for example, there are 10 million search results for the main word, there must be subsequent words;

For example, "Naruto Animation" contains the subject "Naruto". But incidentally they do not absolutely need to be included.

Then, in this case, the final display results of the above strategies 1 and 2 are the same;

There are always some special cases, such as caching strategies, etc., so you need to search for more words to confirm.

Conversely, if a high proportion of the first-ranked results for a direct search of the subsequent word do not appear on the homepage of the main word, it can be seen that Baidu adopted strategy 3.

In fact, Baidu uses strategy 3.

One of the best things about SEO is that the ranking results are there. To verify a hypothesis, just search for a few more words. The cost is minimal and spending a few hours is a waste of time.

As the old saying goes, practice makes perfect. You have to try to be sure of the result. Testing can take weeks or even months. Those who always insist on "practice" regardless of the circumstances really have a bad brain.

6. Although it is logically clear, it would be easier to remember such a concept if there is a corresponding word, which is generally called rough ranking and fine ranking in the search engine field.

Rough ranking is when a search engine finds the top n pages corresponding to the main word from hundreds of millions of Internet pages.

Precision ranking is when the search engine further calculates more complex rules from the top n pages, which includes the issue of subsequent search needs (but it is far more than that).

I have shared the concept of rough and refined defecation in a small circle. Later, a friend asked a Baidu R&D staff member he knew directly, and the other party replied that the number of pages taken by Baidu's coarse ranking in this round is the maximum number of search results displayed, 760, and the sorting rule of coarse ranking is only BM25.

If readers don’t know what BM25 is, the basic principles of search engines will be a mess. I personally recommend starting with “Introduction to Search Engines”.

It is definitely nonsense to say that the rough ranking rule only uses BM25. If you only use BM25, you can pass the rough ranking by simply repeating the keywords a few times on the page. This is extremely unreasonable and does not match the large number of cases I have collected.

But the rough number of pages taken is 760. Based on my rough analysis of observations, this may be the case.

7. I have written a lot, but how do I get the word "SEO" ranked higher?

Let me introduce the initial situation of the website. At that time, I operated a small SEO industry forum with more than ten to twenty active users posting every day. The entire website has hundreds of pages, with very few friendly links (after the ranking went up, I could easily exchange links with others, so I added some to be on the safe side), and no other external links.

What happens after you start ranking? Because the process of ranking up is so simple, if I don’t explain the theory comprehensively, it will definitely be regarded as nonsense:

Modify the title of the website homepage and add "SEO School"

Bought two friendship links for a total of 50 yuan

That’s all.

It's really gone.

8. Some readers will definitely think, is there something wrong with this simplicity? If you intuitively have this idea, it is right. After all, misattribution is a mistake most people make.

I have seen many, many times that someone did some SEO operations, but forgot about them after a few months. When they came back and found that the traffic had increased, they said that they did the right thing. This is blind correspondence.

Attribution, at least in the context of SEO, requires at least two things to be correct:

1) Reasonable theoretical explanation

2) Reasonable timing

When my website's subsequent word "SEO Academy" was already ranked high, and "SEO" was not even in the top 760, I knew that the last step was only to get into the rough ranking of the word "SEO".

Moreover, according to observations, Baidu's rough ranking rules are mainly similar to the relatively basic weights of PageRank. This is also a problem that can be discovered by searching a few more words. So I bought two links, and about a week later, Baidu's ranking was updated and ranked on the homepage.

9. There is a strange question that has not been explained yet and readers must have discovered it. Why is the subsequent word "SEO Academy" such a strange word?

The answer is, I don’t know who posted it, it’s always there. I no longer had access to Baidu Sinan’s data at that time, but I could directly determine that “SEO Academy” was a useful subsequent word just by seeing that there was a page with “SEO Academy” in its title for a long time and that basically no other subsequent words were matched on the homepage. It was a lucky coincidence.

Later, the ranking was stable for a while, and I also operated the forum on my website quite well. I thought the website weight should be quite high. At the same time, I felt that the weird word "SEO Academy" was not suitable for the title, so I removed it and changed it to "SEO Tutorial". The ranking remained stable for a long time.

If it wasn't a small forum like mine with no content and no links, any slightly better website could have directly ranked for "SEO tutorial" or other similar follow-up words.

etc!

Why didn’t my website’s ranking drop after “SEO School” was changed to “SEO Tutorial”? Is it really like the theory I mentioned earlier? Or should it have been ranked first, and it's all just a coincidence?

As mentioned earlier, it is normal for readers to have doubts during the reading process. Not having doubts is the wrong attitude towards acquiring knowledge.

However, returning to the theory of subsequent words, even if you may not have done it or observed it, you can still figure it out simply through logic. Combining the subsequent search terms with the concepts of rough sorting and fine sorting, we can come to the following conjecture:

For main words with subsequent word effects, there are only two states for the ranking of the corresponding search result page: on the first page and at any subsequent position. There is no situation where the ranking on the first page drops and then moves to the second page.

The actual situation is a little more complicated, such as open platform space occupation, click-through adjustment, etc., but the overall situation is still as described above.

At the beginning, when my website was ranked up through "SEO Academy", the rankings of other websites corresponding to the original "SEO Academy" fell to the bottom at the same time;

Similarly, when "SEO Tutorial" went up, the ranking of the original "SEO Tutorial" corresponding website also dropped to the back.

The theory and timing are perfectly consistent, so it is the correct attribution.

Later, on the one hand, Jiao Da also figured out a general idea on his own and competed with me for the ranking; on the other hand, I was starting a business doing outsourcing at that time and the volume of business was too heavy to handle. Eventually, the keyword "SEO" dropped from the Baidu homepage after remaining on it for more than a year.

In addition, today, there is no need for readers to analyze the ranking of the word "SEO". Baidu's homepage is currently fully covered by various click-brushing websites, which is simple and crude.

Summarize

Readers can review the entire article. Everything is common sense that even a layman can easily and quickly grasp. Through a lot of logical deduction and a little word searching to verify the conjecture, a complete theory can be finally obtained.

Many of my colleagues know that I have extensive experience in the industry, so they may wonder, although I have explained the previous problem through logic, in fact, is it that I achieved the effect through "experience" and just explained it through logic, thereby giving the credit to "logic"?

I want to emphasize for the nth time that it is right to be skeptical.

No.

Why did I get "SEO" ranking on Baidu's homepage? Because I was too idle when I was at Meituan. I was 23 years old that year and had only been in the industry for four years. Moreover, it was a much earlier time for me to understand these theories, so there was no experience to rely on.

Finally, let’s put it into practice, change the title and buy two links. Ranking for a keyword like "SEO" seems almost impossible to achieve for most people, but in reality, it is just a piece of cake to do.

Changing links all day long, pseudo-originality...

Looks taller.

Writing code and crawling data all day long...

What are you doing?