This article first explain what is today's headlines smart personalized recommendations, and then explain headlines on the recommendation mechanism, the law.
1. What is today's headlines smart personalized recommendations?
No headlines with the biggest difference between micro-channel public number is: No. headlines own fans, even if the headline number creators No fans also possible to obtain the amount of reading millions and billions. No amount of reading the article headlines decision level is the recommended amount. No headlines from media quoted intelligence personalized recommendation engine, readers can find exactly what, no demand attention, seeking subscription can also have a mass readership.
Today's headlines belong to machine intelligence personalized recommendation mechanism, namely by machine algorithm users to post content (videos, articles) distributed to the audience content. Smart personalized recommendation engine based on the article characteristic (content quality, content features, starting, engagement), user characteristics (headlines on the history table, headlines on subscriptions) and environmental characteristics (time, place, weather) and so on, as interested readers find articles and recommended to them.
2, headlines on the recommendation mechanism, law
Headlines on the recommendation process is this: as long as approved, the machine will first try to recommend a certain amount, if the click-through rate and read completion rate, then a new round of broader recommend to users more similar, if reduced to a certain extent, the recommended process is over.
The first key point recommendation process: the initial recommended amount
After review by the machine will first try to recommend a certain amount. That in the end this is how a certain amount of work out of it? That is also recommended certain amount of influence the most critical factor in today's headlines try after checking what is it?
The answer is No. headline index. Headlines on the index can be interpreted as "how your content deserves to be recommended," the index is read by a machine reader of the creator and the content over a period of time, record and analyze the behavior of interest rates derived from the value of the account, including health degree, attention, spread, perpendicularity, five of the original dimensions. The higher the index number headline article recommend more.
The second key point recommendation process: Secondary recommended amount
Today's headlines after trying to recommend a certain number, will decide whether to recommend a new round of broader, or reducing the recommended amount or based on click-through rate and reading readers' completion rates.
The user clicks on, but not necessarily watching, reading is to read the number of clicks divided by the number of completion rates. Click on the title and the more people read the article, the higher recommendation. Of course, there are other behaviors also affect the secondary recommended amount, such as thumbs up, forward, favorites, comments, interactive, such behavior secondary recommendation is also a great help. There is also a big factor affecting the second recommendation, issued a document that is time. Article published the longer, the lower the recommended amount, and theoretically recommended time of 72 hours, in fact, most will not be more than 48 hours, and more than 80 percent of the recommended amount is completed within 24 hours.
Headline index number determines the initial recommended amount, click-through rate, completion rate reading, thumbs up, forward, favorites, comments, behavior and interactions of these secondary effects of the recommended amount. And only after the initial recommendation had more hits, read completion rates, as well as more thumbs up, forward, favorites, comments interaction of these acts in order to have a better secondary recommended amount. Only with more secondary recommended amount, the amount of reading the entire article to it, driven headlines on the rise throughout the index. The headline index increased number can bring more initial recommended amount, so the cycle repeated.
As mentioned earlier, the headline number comes with a fan, as long as approved, the machine will first try to recommend a certain amount. As long as click-through rate and a higher completion rate reading, the reader thumbs up, forward, favorites, comments and more of these behaviors, you can increase the amount of secondary recommendation, it can drive the headline number rose that can make the above figure this cycle into a virtuous track .
That in order to read the article click-through rate and completion rate, get more thumbs up and forward, depending on what they do? This is very simple, right? High-quality content is fundamental. Only good content, readers like to make these acts possible, otherwise all nonsense.
3, posting the best time
After a previous analysis, we know that today's headlines biggest feature is the intelligent personalized recommendations, even if your fans, rarely take the initiative into your home to see your article. The main recommendation is to give the reader through the headlines today, readers will click to read.
So today's headlines most critical is the recommended amount, and if you can find today recommended large headlines in what period of time, it means posting the best time to find out today's headlines. Because the longer the time from dispatch, the less the recommended amount, basically do not have the recommended amount, after more than 24 hours.
Drawn three highest recommended amount of time (in chronological order):
07:00 AM --- 10:00 a total of 3 hours
11:00 noon --- a total of 2 hours 13:00
At 22: 00--01: 00 a total of three hours
That point in time at which we should send it?
Combined with the previous analysis of the headlines on the recommendation mechanism (the law) is better draw the following conclusions:
Morning and evening is better than at noon, why? Because only if issued at noon to enjoy high-recommended amount of two hours, then you have to wait until the evening, and the longer the time from dispatch, will lower the recommended amount. So is issued immediately after the best recommendation is the high amount of time, but this time as long as possible.
That same morning and evening is three hours, which is the time it issued more effective? Good evening than in the morning light of the foregoing analysis and proved. This is because after posting the evening, from 07:00 01:00 --- separated by only six hours to enjoy high recommended amount of time three hours again. And if it is issued, then in the morning, from 10:00 --- 22:00, need to wait 12 hours in order to enjoy a high amount recommended time period of three hours again, but after posting 12 hours, has been greatly affected by the recommended amount influence.