What happened to the likes in VK

Explains the director of growth and research "VKontakte".

On August 6, users drew attention to the disappearance of the counter of likes and reposts for some users of VKontakte. This change is part of the testing of a new algorithm, and experimenting in this direction not only VKontakte, but also Instagram.

In the case of Instagram, the effect of the absence of meters of the main metrics of the “success” of posts is already known. As part of the experiment on Canada’s users, it became noticeable that they began to publish more content, getting rid of the desire to gain more likes, and ceased to evaluate the quality of other people's posts based on data from meters. It is likely that the Instagram team will continue to test the refusal of likes in other countries or roll out a new paradigm at all - and this will change the consumption of content in one of the most popular mobile services of the world.

Behind the solution of VKontakte are also worth the data of experiments and a certain philosophy. TJ asked to tell Andrei Lawonov about this, the director of growth and research on VKontakte.

Every day, in parallel, there are thousands of versions of VKontakte. They differ not only in algorithms, but also in interfaces and functionality - within the framework of tests.

Most of the experiments are quite imperceptible, but one of the recent has become very resonant. For the test, we removed the counter of likes and reposts in the news feed of a small part of users. On that day, many thought that VKontakte removes likes completely and everywhere - but no, we have a different idea. Likes remain, they are still available as part of this experiment on the screen of a separate post, but their tasks and how they are perceived can change.

We are aware that changes are difficult. But the only way we will win

Changing a product that almost 100 million people regularly use is very, very difficult. The news feed is one of the main sections, users spend tens of minutes in it daily. Any interface update can break the usual patterns of behavior. Often, after a couple of days or a week, the update becomes convenient, the new UX is familiar. But at the time of the changes, everything new causes negativity.

Even more difficult for professional authors. Changes sometimes affect not only subjective comfort, but also the audience coverage, monetization or business metrics. Given all this, we are as serious as possible to any significant update.

So maybe you don’t have to touch anything? But it is impossible not to change. Because everything around is changing: technology, rhythm of life, ways of using social networks, local and global trends, user habits and tastes. This is not only the theory, we check experimentally: periodically we return the version of the algorithm to a small test group, which was several months or six months ago. And each time we see that the old versions are really worse solving users' tasks in the tape. So it is with interfaces.

Do you need likes in the news feed

It is important to figure out what likes is. We perceive the tape not just as a content consumption service, as reading a newspaper or viewing television programs, but as a platform for communication and communication between the author and readers. Comments make it possible to supplement the publication, express an opinion, ask the author a question or argue with him. Repost - share content with subscribers.

And the like, which is actually called “I like” VK is the most concise way to tell the author “excellent recording”, agree with the opinion, praise friends for a class photo, playlist or video.

In this regard with likes, everything is super. In our experiment, the Like button from the tape does not disappear anywhere - it even becomes larger and more noticeable.

Why do not need likes

There is a very popular incorrect opinion that the number of likes is an assessment of content quality. Why is this not?

The number of likes in itself does not mean anything at all. You see one post with a photograph that has 1000 likes and the other, which has 50. Which users like it more? It is impossible to draw a conclusion, based only on these two numbers. The author of the first record may have a million subscribers, and the publication gained 1000 likes for three days. And the second received 50 likes in the first 15 minutes - despite the fact that its author has several hundred subscribers.

There is a more complicated and also incorrect example of analytics - in the mind to calculate the coefficient of involvement (Engagement Rate), dividing the number of likes by the number of views. But so we can only find out how views are converted into like. After all, the mark “I like” is only one of many ways to engage. Some posts collect few likes, but a lot of comments. Yes, there is a commentary counter, but this is only the peak of the iceberg, if we talk about the analytics of the post. The articles are important in the video - inspections, in podcasts - listening time, in posts with links - clicks. For example, when the tape algorithms evaluate the recording, they use hundreds of different metrics: reading time, the number of discussions, interaction with media content and documents, and others.

As a result, to be objective, the number of likes does not give any opportunity to adequately evaluate the post, even if you attract all available information and calculate conversions in the mind. And if the information is not needed in the interface, then it is logical to try to unload it, to make it more minimal and more convenient. That is why we launched an experiment with a new tape interface.

One of our recent experiments was connected with the same idea when we hid part of the attachments in long records. It is inconvenient to read the tape when one post can occupy two or three screens. In this direction, we will also continue to develop.

What information is useful in the tape

First of all, the content itself. The author and content are important in it. We focus it, unloading the interface. Record is communication between the author and the reader. To understand whether you like the post or not, you need to see the content itself, and not the reaction of other readers.

The analytics in the number of likes can be familiar, but it looks illogical - so perhaps this is the habit that it is useful to change.

In this experiment, we leave a block in the tape, in which we show which of your friends I liked the post. Because this is about communication: you know that friends also responded to this record, and you may be interested to discuss this content with them. The social component of media consumption in the experiment is not affected and even visually emphasized.

We left the counter of the number of comments - because it, unlike the husky counter, helps to decide whether to go to the post screen. If there are no comments or there are few of them, it may not make sense to make an extra click. If there are a lot of them and you were interested in the recording, then this is a hint that you can switch to the screen screen, read opinions and, possibly, take part in the discussion.

What are we expecting from the experiment

Did it happen that the post seems to be cool, but somehow it like the first to like it? We want to try to get rid of such psychological traps. It is important for the authors that good content receives fair feedback.

We expect the feedback will become less susceptible to the effect of the crowd. This will make the content system more balanced. Perhaps the experiment will not lead to an increase in grocery metrics, but the tape will become more convenient for users. We are waiting for feedback to analyze and understand whether it became better.

When we roll out on all

We do not initially know about any update that it will definitely be successful. At the first stage, the hypothesis is put forward - for example, "let's remove excess information from the news feed so that more attention is paid to content."

The next stage is the inner roar of the idea. Given the experience of the team, we are discussing what this can affect. Then we make different layouts to understand how it might look. Then we realize the idea on one or all platforms and begin to use the new version of the application inside the command. Often at this stage we ourselves understand how and what can be redone, and we finish it, we try it again until it is convenient for us. Obviously, this stage is very subjective.

Therefore, the next step is the A/B tests on small random user samples to get objective and statistically significant results. At this stage, we not only look at hundreds of grocery metrics, but also analyze feedback: from users in comments and in support, as well as from professional community. Very often, following the feedback results, we make changes and launch the experiment again.

So we will roll out the new tape without a flasking flask on all only if the experiment is successful in all respects. If the result is controversial or negative, we will modify the idea according to the new introductory.

What to expect next

Many things. We are now actively working out a lot of hypotheses, how to make it more convenient to watch the tape. We know that the like is used in different ways: someone says thanks to the author, and someone retains a post to read later. There are also those who do not like some of their favorite posts, because they do not want this information to be public. Therefore, we are looking for ways to make all these patterns as convenient as possible.

We also see that communication in the comments on the platform becomes much more active, so in this direction we also have a lot of grocery plans. Traditionally, Stay Tuned!

#VKontakte #social networks #opinions