How to avoid the Matthew effect by charts and random numbers

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

　　

The following are some of the things you actually encountered and tried to solve in your work a few days ago. The company has an excellent team, everyone is thinking about how to solve some of the existing unreasonable problems, whether it is product design, content operation or background architecture. The following problem is based on the needs of a data, design and summary of thinking according to requirements, but also need to constantly try to adjust and optimize.

In fact, whether E-commerce site commodity trading volume, Content Site page browsing volume or resource site resources download volume, most of them will exist in the list of this thing. The rankings provide an entrance to a popular message, and are a typical manifestation of Word-of-mouth marketing and user herd psychology. But for a good site, more than 80% will not use the natural ranking, how many will move some "small hands and feet", that is not deceiving users? Yes, we do do that sometimes. And compared to the money-driven approach to the irresponsible change rankings of users (which everyone should understand), the following tips for changing rankings are definitely "beautiful lies" to users.

The drawbacks of natural rankings

Let's take a look at what happens if the site is tailored to the natural rules of the list:

　　

Of course it may not be so exaggerated, but it is possible that the contents of the line remain unchanged for 1 months or even 3 months. The herd mentality of netizens is not to be neglected, for a user who is wandering around without a clear purpose in the website, all kinds of rankings become their best entrance, or we can explain this problem from the level of data, where is the general list hanging on the site? Home page, category index pages, sidebar? Anyway, those who are easy to see, can easily click where the common features of these pages or locations are high exposure, to the home page for example, to see my blog exposure ranked top 5 pages:

　　

From the data can be seen on the home page is almost 10 times times the number of other pages, so for a resource, whether the list will lead to at least 5 times times the difference in exposure, that is, if you want to complete the same conversion target, not put in the list of goods need to put the list of goods higher than 4 times times the conversion rate, For the same product, if you put it in the rankings as long as 4% of the conversion can be completed 100 transactions, so if you pull it out of the list also to complete 100 transactions will require up to 20% of the conversion, is simply a difference, this is the magic of the list.

This is the so-called Matthew Effect, is a very interesting phenomenon, but whether for the site or the user, the effect of the Matthew effect is unfavorable, the Web site can not be the top of the list of potential products sold out, and users will also not see those outside the rankings more valuable products. So we need to try to circumvent the Matthew effect.

How to avoid Matthew effect

Matthew effect (Matthew multiplying), simply said, "The stronger the strong, the weaker the weaker". It must be admitted that the Matthew effect is an extremely powerful natural law, and the world is trying to circumvent it, but how many can really solve the problem. The problem in economics and society seems to be that there are many solutions in the virtual realm of the Internet, and one method that is absolutely not applicable in other fields is random numbers.

In fact, there are many ways to circumvent the Matthew effect in the rankings, if the site's data and computing platform is strong enough to use some algorithm to effectively solve the problem, before I also introduced an effective content recommendation method, or if you can according to the results of user behavior analysis to provide customized personalized rankings based on user interest , you may not need to look at the following. But for a small web site or a new product that is not yet fully available, random numbers will be one of the easiest and most effective ways to circumvent the Matthew effect.

There are many ways to produce random numbers, and there are many different types of random numbers, such as a common range of two decimal places (0,1], or a randomly generated 1 to 100 natural number. There are a number of ways to change the list, and here are 1-100 of natural random numbers to illustrate some of the ways I think of adjusting the TOP10 list:

Random Ranking Method 1

Adjustment strategy: In the first 30 items randomly take 10 into the list.

Implementation: Assigns a random number to each content of the top 30, then sorts the top 10 according to the random number.

Application: The same applies to the first 50 or 100 scrambling order after the random 10, but whether the top 30 or 50 of the content rankings can not be too significant differences, such as hot books, perhaps the first 30 books of the popularity difference is not so obvious, then you can use this random sorting method.

Random Ranking Method 2

Adjust the strategy: the list of 8, 9, 10 to replace the 11-20, 21-30, 31-40 of each take a random content.

Implementation: In fact, is to generate 11-20, 21-30, 31-40 range of 3 random numbers, put to the list of 8, 9, 103 positions, first 1-100 of the random number to do a simple processing, divided by 10 to take the remainder, so it is the equivalent of 0-9 random number, in addition 11, 21, 31 generated 11-20, 21-30, 31-40 range of 3 random number, take the corresponding content to the list.

Application: There are obvious star products, for example, download the software download site before several will have been occupied by some software, these software is indeed the most commonly used software users, and other software downloads there is a significant gap, this time do not go to the top of the list of products, and as long as the random changes in the number of products on the list.

Random Ranking Method 3

Adjustment strategy: Randomly take a position in the list to place the recommended content.

Implementation: With the above example, this implementation is relatively simple, in fact, is to generate a 1-10 of the random number, the first 1-100 of the random number divided by 10 to take the remainder, plus 1 can be, and then the list of the random number of content to replace the contents of the recommendation can be.

Application: There are already good quality content to recommend to the user, and can expect these content is likely to be in the future ranking.

Okay, so here's what I've listed. Some of the practical applications of using random numbers to change the charts to avoid Matthew effect effectively, do you have a better idea and plan to share with me in the comments?

» This article uses the»in agreement, reprint please indicate the source: the Website Data analysis» "The list and the random number"

Original: http://webdataanalysis.net/personal-view/ranking-and-random/