How I "twisted" behavioral factors

Greetings to you, habrazhiteli! Today I would like to tell you about a very intriguing thing, like behavioral factors. Or rather, about their artificial improvement. And more precisely, about the shortcomings of services-cheaters that do it.


Why am I writing this? The fact is that I am the owner of two commercial sites and, of course, I am interested in finding them as high as possible in search results. And at some point, after reading various articles, he decided to experiment. At once I will make a reservation that the result was a bit sad and I hope that this article will serve as a warning.

Start

Given:

• 2 commercial sites;
• Subject: construction;
• Region of promotion: Moscow;
• The age of sites: one is 5 years old, the second is 3 years old;
• Traffic before the experiment: site A - ~ 350 users / day, site B - ~ 200 users / day;
• Begin / end of experiment: January-April 2014;
• Links are not purchased from November 2013;

For more “cleanliness,” I did not make any changes on the site, except for posting articles once every one or two weeks. I note that I am not going to compare the positions before and after, but the traffic, since for me personally this is a more important metric.
')

Choice of service cheat

It did not take long to think about this moment - I decided to use the most actively advertised and new service at that moment. I will not give a link to it, but I will only say that it has an interesting name and logo. :)

Promotion strategy

After sitting in the forums and studying articles from several sources I was pleased with the following decision:

• On site A add 1500 transitions per month
• On site B add 900 transitions per month

That is, approximately I added 15% of the current monthly traffic. I think that it will be clear to the majority that it is somewhat doubtful to wind such a number of transitions on 1-3 requests, then I selected 80-90 LF and MF requests for each site.

The disadvantages of cheating: the depth and time of viewing

During the project setup in the service, I was immediately jarred by the lack of installation of the viewing depth and time on the site.

I believe that each site of a particular subject has its own depth and time of viewing the site. If for my sites the average depth is 4 pages, and the time is 4-5 minutes, then the shaft of visits with a depth of 2-3 pages and a time of 2-3 minutes somehow gets out of the general statistics. In the end, it happened.

Disadvantages of cheating: visitor characteristics

Yandex.Metrica has very remarkable and informative sections “Gender and age” and “By countries of the world”.

So, if in the subsection “By Countries of the World” Russia, the Center (and a little North-West) always prevailed, then after a month of progress I saw the Urals, the South, Siberia, the Volga region. Also, the CIS “actively joined”: Ukraine, Belarus, Kazakhstan and many other things.

In the subsection “Gender and age” the picture is typical for my sites: the main audience is men over 25 years old. After a month of promotion, a female audience began to grow, as well as an audience under 18 years of age.

Well, is it not a failure case ? Naturally, the support of the service actively persuaded me in this, after I sent them a ticket with comments. The essence of the belief was one thing - " We firmly believe that the Yandex Yandex does not take into account the data you wrote about ." Point.

Search Behavior

We should also highlight how tasks are performed:

1. The Employee opens the search engine and sets up a region (in my case, if it is not from Moscow).
2. Enters a search query.
3. Switches to a couple of random sites.
4. Goes to my site.

It is very strange when a user from Poltava changes the region to Moscow and searches for concrete blocks, when yesterday he was looking for GDZ for the 8th grade. Then the user clicks on two random sites. Yeah. Most often these are sites from the TOP-5 or TOP-10. After viewing the sites, the user abruptly goes to the site from the 29th place and then finishes the search. Yes-ah, very natural behavior.

My light, mirror, say or watch WebVizor

“What is this?” You ask. And this is the scale of user activity from the web browser . It displays typical user behavior from a cheat service on my site. According to my calculations, 7 out of 10 users behaved this way. For those who do not know what these colors mean on the activity scale:



That is, the user went to my site, “twitched” for 20 seconds, then switched to another window and did something for 2 minutes. Then he came back, clicked somewhere and closed the page. It is clear that most likely the “worker” performs several tasks at the same time, which causes him to switch to another window.

And in the appendix to everything: in WebVizor, you can filter out visits from a specific visitor, which helped me find another flaw - one user can complete the task several times a week.

Total

After a difficult 4 months, I decided to finish the epic with this experiment. I will not bore me with a story about other problems, such as incorrect removal of positions, an incomprehensible algorithm for performing tasks and writing off funds, numerous minor bugs in the interface (which they promise to fix, but this does not happen) and much more.

It turned out that in 4 months I spent about 40 thousand rubles, a lot of time and nerve cells . At the beginning of May, a week after stopping the promotion, I recorded the following traffic:

1. Site A - 220 user / day .
2. Website B - 180 user / day .

At the beginning of the post, I promised not to compare the positions, but, as a digestif, it is worth doing:



If you decide to use behavioral cheating, remember the drawbacks of “cheaters” :

1. There is no possibility of setting the depth and time of viewing the site
2. When you install the desired promotion region, you will receive users from all over Russia and the CIS.
3. Unnatural behavior of the “workers” of the service on your website and on the search results.
4. The same “employee” may visit your site several times a week.

For myself, I decided that in the future I will not use dubious gray methods. And I do not advise you.
Thanks for attention!