How to educate the user and not drive him crazy. Yandex Browser Experience

    Yandex.Browser, like any other multifunctional program, has features that not everyone knows about. Moreover, not only the users themselves lose from this, but also the product as a whole - its potential as a result is not revealed to the entire audience. You can polish the design for as long as you like, fill out reference materials and write articles on Habr, but the bulk of users will still not be aware of all the possibilities. We needed a new, smarter and more accurate tool to accurately train the user. In this article, we will share our experience in creating this, and perhaps it will be useful to you when developing your products.

    As we have already noted above, it is possible and even necessary to talk about product features in detailed reference materials. The only situation now is that Help is being accessed less and less. This is about the same as the user manual. Tell me honestly, did you read the user manual for your new smartphone before using it?

    In addition to Help, there is such a thing as a start page (we call it “welcome page”) with a description of several of the most striking features. But there are many opportunities, and the user does not have much patience. Therefore, it was necessary to come up with something else. We will not tell you that as a result, we invented tooltips. After all, they have been used in various programs for a very long time. Think of at least the world-famous Clip from one very popular office suite, which had both strengths and some annoying moments. We took this idea as a basis in order to build our own system of smart tips on the basis of it. And then the fun began.

    Smart tips are what?

    What could be simpler than showing a person a tooltip about the availability of a particular function, you would think. In fact, everything is not so simple. To understand this, let's go back to the welcome page, which we already talked about above. Why can not this start page be used to teach a person all the wisdom of the Browser? There are at least three reasons for this:

    • A multi-page start-up briefing would rather cause a desire to remove the Browser or remember the kind words of the developer and his relatives than try all these features. You can, of course, give a button to close this training, but then the situation will be approximately the same as with licensing agreements. We all read them, right?
    • The theory, in isolation from the practice of application, will be forgotten quickly enough. It’s like with foreign languages: if you even read the textbook, then without communicating in the language you are studying, it will be very difficult for you to speak it.
    • It makes no sense to educate the user about something that he may never use. And this is not about modern higher education.

    These thoughts are applicable to any method of training, and, based on them, the developers proposed to show the hints contextually and only when the need for them is maximized. But this is only the surface of the iceberg, because each function has its own optimal moment, which still needs to be determined. But let's not get ahead of ourselves yet, we will look at specific examples below.

    Imagine that we have many functions, and the user quickly encounters them, which leads to the constant appearance of tips. You can imagine it, but the development team decided to try it for yourself. If you show ten dozens of unexpectedly pop-up messages within ten minutes, then you are unlikely to be satisfied, despite all their usefulness. Having experienced all this on ourselves, we realized the need to normalize the frequency of occurrence of prompts. After that, we conducted several more experiments on almost volunteers from the Yandex.Browser team, which allowed us to find the following sparing mode of working tips:

    • Show the same hint no more than once a day.
    • During the day, show no more than two tips of a different type.
    • If the user saw the same tooltip three times, but did not respond, then we do not show it for 1-3 months.

    A separate point is how the user can act after a notification appears. Our tips are contextual, which means that they always appear exactly when the user is studying the content or taking some action. To show a hint blocking the Browser at such a moment (for example, to demand clicking on the cross or the button “I read and understood everything”) - this is guaranteed to cause inconvenience to the person. In the course of all the same field tests, it also became clear that the tooltip should not take the focus away, otherwise the text entered at this time may go down in history forever. Man, to put it mildly, will be unhappy. Therefore, our messages can be ignored and continue to work - they simply disappear.

    And one more thing. The last wish for the tips was the ability to disable them. Therefore, now in the settings of the Browser they can be turned off immediately and that's it. Or click on the cross in each specific message, which will disable it for the next three months.

    We are done with the theory, and now you can look at the practice of applying these recommendations. We have selected a few examples that might be of interest to you.


    Readers of Habr are well aware of the ability to search immediately through the omnibox, but many other users often do the old-fashioned way: first they drive the address of the search engine, and only then they search in it. Different browsers solve this problem in different ways. At the time, Chromium offered to put the search engine on a new tab and immediately transfer requests to the omnibox. Also now you can find references to the Origin chip experiment , which consists in removing the page address from the omnibox in general.

    In Yandex.Browser, the omnibox was originally designed in the style of the Yandex search bar in order to prompt the user at the associations level. Now, along with this, tips are used that clearly talk about this possibility. Since the essence of the omnibox is widely known in narrow circles, we did not immediately tell every user of the Browser about it, but chose to use the hints only if a certain condition was met.

    If the ratio of search queries from the omnibox and from the Yandex main page drops to at least 1 to 4 in favor of the site, then the user sees a hint. This ratio was deduced exclusively by trial and error. At the same time, it is imperative that the total achievement of search queries exceeds 100. At lower numbers, there is a high risk of reaching a user who knows everything well, but is currently actively searching through the site for some reason.

    If you look at the results now, the preliminary data are as follows: among those who saw a hint, the number of searches via omnibox increased by 7%. This figure is rather modest, and we had the assumption that these users more than consciously go to the Yandex main page and search from there.

    Page Translator

    Yandex.Browser can offer to translate the page into Russian. Previously, we used a pop-up panel to do this, which was already annoying because it moved content on the page. Therefore, we recently transferred the translator directly to the omnibox, which many of our users took very positively. By the way, the Chromium project went the same way and in the 34 version did the same.
    After translating the translator, we noticed that his CTR decreased by about 20%. This is quite logical, because now users can safely ignore it. If the old translator was himself intrusive enough and did not need hints, then for a new one training would not hurt.

    In this case, we show a hint if a translator appears in the omnibox, but the user has never used it. As a result, we, on the one hand, do not impose a function on those who definitely do not need it. On the other hand, we do not cause irritation for users already familiar with the translator. And here are the results:

    Judging by the fact that the share of those who agree to translate the page more than doubled, the tip reached its goal, and a fairly large number of users got acquainted with the function.

    By the way, while studying the tips, we accidentally noticed that moving the translator to the omnibox somehow positively affected the use of the latter. The number of searches in omnibox as a result increased by 3.4%. Either this is an error, or an example of an interesting interconnection of seemingly unrelated functions. And what are your versions?

    Translator of words and phrases

    While many are already familiar with the page translator using the Chrome browser, there are no translations of individual words and phrases. It may not be obvious to a beginner that the browser is capable of this, which means that a lot already depends on the prompts. But how to determine if a clue is needed? Almost every page on the network contains words in a foreign language, so the hint cannot focus on this factor. Fix the use of online translators? But who said that the translated words were found on the page, and not in the subtitles of the beloved honestly bought series.

    The decision came completely unexpectedly. Already known to readers of Habr kukutzHe hypothesized that the selection and copying of a word in a foreign language in some cases can serve as a sign of a desire to translate it in a third-party service. It is clear that there will always be those who copy words for other tasks, but the developers clung to this idea, added a filter that cuts off those who already used the translator, and created a tooltip.

    And the result was very pleasant. The share of users who at least once used the translator of words and phrases jumped.

    Scam Protection

    The browser can warn the user about visiting a site that is suspected of fraud. Previously, a pop-up panel was used for this. Unfortunately, a large number of incorrect advertising on the Internet has done its job, and many users have developed immunity to such warnings and simply do not pay attention to them. To fix this, we tried to add a tooltip to the panel as well. It turned out like this:

    And the attention was really attracted, as users began to read the details more often.

    But the most interesting thing was that the percentage of ignored warnings increased, and few could have expected this. There are various hypotheses on this score. For example, there is even an opinion that users were afraid of the warning, because they did not understand its essence, and after reading the details they began to act more boldly.

    Such a high percentage of potential victims could not be ignored. Therefore, here we took advantage of the experience of blocking sites that distribute malicious content, and immediately began to block the pages of scammers. In this case, it is completely acceptable to interrupt the consumption of content by the user, since we are not talking about training in new opportunities, but about curbing the crime. Do you agree with us?

    Up button

    The “Up” button is our conditional name for the function of returning to the top of the page by clicking on the tab. The simplest thing here would be to show a hint when located at the bottom of the page, but in most cases we do not return from there upwards. So, in order to make our message as useful as possible, we need to more accurately identify the desire. Therefore, the Browser monitors scrolling down (at least two screens) and then up (at least one screen) and only then offers to return to the top of the page in a simpler way.

    If the user used the prompt, they will be immediately informed of the possibility of returning to the previous place through a second click.

    Instead of a conclusion

    There are much more tips in Yandex.Browser than we described in this article. But even these examples show that even such a simple thing as hints requires a special approach and can sometimes lead to very interesting results. We realized that for some functions, tips are useful and necessary, while for others they do not bring great results. We learned how to select moments for showing tips so as not to cause irritation to the user. And they also realized that in security matters it is necessary to act more decisively and prevent problems due to changes in the product, and not tips.

    But how did people perceive these clues? According to our statistics, most people see no more than one clue per week. At the same time, only 12% of people prefer to close them through the cross, which is not so much. In support of the Browser, complaints about hints were almost never received. Based on this, we concluded that the user training model we selected does not create additional inconvenience and fits perfectly into the product.

    Also popular now: