Do you visit Habr more than once a day? Not to read something useful, but just to scroll through the main page in search of "what to add to the list to read later"? Have you ever noticed that posts published in the middle of the night get fewer views and ratings than daytime ones? And what about the publications that appeared in the middle of the weekend?
When I published a previous post-analysis on the dependence of publication indicators on its length, in comments that "there is some correlation between release time and publication rates (but also weak correlation)". Do you understand that I could not pass by?
So, is it important to publish on Habré from 09:00 to 18:00? Or maybe only on Tuesdays? What about the day after payday? Vacation period? Well, you get the idea. Today we will try to find out a temporary recipe for the best publication in the world.
Introduction and data set
Since we do not know exactly in what time frame there may be some interesting (or not so) dependences of publication indicators, we will analyze everything that is possible. Let's try to consider what happens during the year (are there any seasonal dependencies), during the month (are there any social / household dependencies - I wasn't joking about the pay day), during the week (is there any dependence on the degree of readers / authors' fatigue) and during the day (is there any dependence on the amount of coffee drunk).
To analyze reader reactions to a post, consider the number of views, pros/cons, comments, and bookmarking. Perhaps the cons are more placed early in the morning, and the pros late in the evening (or vice versa). And to identify author's dependencies - the size of the publication. After all, maybe the author writes less during the day, and more in the middle of the night. But it is not exactly.
The article analyzed 4 804 publications from hubs , , , и for 2019. These posts were discussed in the previous .
What's happening…
during the year?
Since the number of views of a publication is infinite, it is clear that publications at the end of the year collected slightly less than at the beginning of the year. If we take this fact into account, then no dependence on the date of publication can be distinguished. So, on the chartFig. 1) there are no special features for Christmas, February 14, or any other holiday. Not noticeable and the season of holidays, sessions or September 1st.
Rice. 1. How views of publications published in 2019 look like, depending on the date of publication
But the voting for the rating of the publication is currently valid for 30 days. Therefore, the only expected deviation is the publications of the second half of December, since 30 days have not yet passed for them. However, posts collect more ratings for the first day and first week, and only a little for the rest of the month. As seen (Fig. 2), users were not particularly diverse in exhibiting both pluses and minuses. It's worth noting that since we use a logarithmic scale to display the number of votes, the graphs don't include all posts that get 0 upvotes/downvotes.
Rice. 2. The number of pluses (left) and minuses (right) collected by publications in 2019
As surprising as it may be, although you can comment and bookmark posts as much as you like, posts are usually discussed and "saved for later" not so long. After - they are forgotten and that's it. Therefore, there are no interesting dependencies on the scale of the year here either (Fig. 3).
Rice. 3. Number of comments (left) and bookmarks (right) collected by publications in 2019
What can be said about the authors of all these publications? Not surprisingly, but now you can find a seasonal dependence - the number of short publications during the holiday season (late July - early September) has decreased (Fig. 4). And here are the medium and long posts in place. Therefore, it is worth noting that we found the holiday season for editors rather than for all users.
Rice. 4. Length of publications in 2019
So, the main result is that no really interesting (and not very interesting) dependencies were found during the year. Let's move on.
within a month?
Number of views (Fig. 5) publications does not depend on the day of the month in any way. To be honest, when building this schedule, I expected to see some kind of surge or fall on one of the days (something like payday - we don’t go to Habr, but celebrate), but I didn’t find anything like that.
Rice. 5. Views collected by publications, depending on the day of the month
But the votes given for the publication show a funny dependence. Habr users obviously do not mind putting a minus (as well as two, three, and so on) on any day of the month. But the pluses are basically put at least 10, although there are exceptions. Basically, the total number of pluses is from 10 to 35. However, there are no obvious dependencies on the day of the month here either.
Rice. 6. The number of pluses (left) and minuses (right) depending on the day of the month
Monthly statistics did not allow to identify the dependence of the number of comments or bookmarks (Fig. 7) from the day. We noticed that on the 24th of any month there are practically no publications with only 1 comment.
Rice. 7. Number of comments (left) and bookmarks (right) depending on the day of publication
What can be said about the authors? It seems that they absolutely do not care what day of the month they write their works (Fig. 8) and how long these works will be.
Rice. 8. Length of publications depending on the day of the month
Actually, I did not expect to see any dependencies on the day in the month, but it was worth checking?
during the week?
And here you can see the expected dependence. Posts posted over the weekend are less likely to garner a small number of views (Fig. 9). However, you should be careful, since there are simply fewer publications on Saturday and Sunday, you should take this into account.
Weekdays, on the other hand, are about the same in terms of views, although the minimum number of views on Friday is higher than on Monday.
Rice. 9. Views collected by publications, depending on the day of the week (starting from 00:00 Monday, UTC)
It seems that publications of the weekend rarely collect only a couple of pluses, and more often - a couple of dozen (Fig. 10), unlike weekdays, when only 4-5 votes for the publication is quite normal. The number of minuses on weekends is also reduced.
Rice. 10. Number of pluses (left) and minuses (right) depending on the day of the week (starting from 00:00 Monday, UTC)
At the same time, Saturday and Sunday publications are commented and bookmarked about as often as any other (Fig. 11).
Rice. 11. Number of comments (left) and bookmarks (right) depending on the day of the week (starting at 00:00 Monday, UTC)
What to say about the authors of publications and the length of posts? They are no different from day to day. What is Monday, what is Wednesday, what is Saturday are the same in terms of analysis of the length of the publication.
Rice. 12. Length of publications depending on the day of the week (starting from 00:00 Monday, UTC)
An analysis of the dependence of publication indicators on the day of the week led to one of the most interesting conclusions. The chance of getting not 5, but 15 pluses, as well as 1 minus instead of 5 on weekends is higher than on weekdays. At the same time, it is advisable to publish no earlier than Sunday morning, then you still have a chance to get into the TOP of the day on Monday morning. The latter will help you get more views and more votes.
during the day?
You did realize that no one posts in the middle of the night, right? An interesting fact is that the night for Habr is quite a standard night in UTC - from about 22:00 to 6:00. But according to MSK, this corresponds to 01:00 - 09:00.
It is impossible to unambiguously identify the dependence of the number of views of a publication on the time of its appearance on Habré (Fig. 13). However, such a graph clearly shows a series of publications at 2:00, 7:00, 9:00 and 9:30 UTC, about which . Basically, these series are publications of editors and corporate authors who have the feature "«.
Rice. 13. Views collected by publications, depending on the time of day (UTC)
Now consider these 4 series of publications. All of them are clearly visible in the dependences of the number of pluses on the time of publication, but not the minuses (Fig. 14). In general, it is worth noting that such series increase the density of publications at a certain time, they do not stand out from the entire set of post metrics data.
However, for all publications in the period 0:00 - 4:00 UTC, there is no large number of minuses.
Rice. 14. The number of pluses (left) and minuses (right) depending on the time of day (UTC)
But by the number of bookmarks and additions to favorites (Fig. 15) there is no significant difference between night and day fasts. As with the views and upvotes charts, there are “editor streaks” visible here.
Rice. 15. The number of comments (left) and bookmarks (right) depending on the time of day (UTC)
What about text lengths? As it turned out (Fig. 16), authors do not have a preferred writing time for very long or very short posts. In general, posts that are longer and shorter are evenly distributed throughout the day.
Rice. 16. Length of publications depending on the time of day (UTC)
Instead of a conclusion
So, when is it worth publishing on Habré in order to get the maximum number of views / ratings / comments, and so on?
If we consider the time of day, then there is practically no difference. Of course, if you publish in the middle of the night, then your post will be at the top of the list of all publications for a longer time. On the other hand, in the morning and afternoon there are many other publications that will move yours lower. On the other hand, while you are in the first position, the chance to get more pluses will be yours and then you can claim a good position in the TOP of the day, which will bring additional views.
In terms of days of the week, there is less competition on weekends and especially on Saturday. But if you still want to take the opportunity to hit the top of the day and get some extra views, then you should aim for Sunday afternoon. Then you can also get readers who view the TOP of the day on Monday before noon (when Monday's publications have not yet had time to collect a significant rating).
If we consider a whole month or year, then there are no special dependences of indicators on time or date.
In general, you know, publish your posts at any time. If they are interesting and / or useful to the Habra community, they will be read, upvoted, bookmarked and commented on.
And that's all for today, thank you for your attention!
PS If you find any typos or errors in the text, please let me know. This can be done by selecting a part of the text and clicking "Ctrl / ⌘ + Enter"if you have Ctrl / ⌘, either via . If both options are not available, write about the errors in the comments. Thank you!
P.P.S Perhaps you will also be interested in my other studies of Habr, or you want to propose your topic for the next publication, or maybe even a new cycle of publications.
Where to find the list and how to make an offer
All information can be found in a special repository . There you can also find out which proposals have already been voiced, and what is already in the works.
Also, you can mention me (by writing ) in the comments to a publication that you think is interesting for research or analysis.
Source: habr.com