Those empty net things: The strange actuality of the artificial flow

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

Those nihility net thing, is my helpless. The reality of the Chinese network Marketing Gray, created a complex, perfect, profitable and deep inside the artificial flow of the industrial chain, and this industry chain for the clients I serve, I am engaged in the industry have a very negative impact. This is a world where the law of "bad money banishes The good currency" unspoken rules will really persuasive things out of sight, "disorderly flower gradually want attractive eye" result is "false to really true and false, do not have there is no", to you do not do something can not survive the point, indeed sad. What I want to share with you today is the fact that we are trying to find out the truth through the weird facts.

Below I will enumerate the strange phenomena I have seen and the reasons I think are possible. However, my vision is limited, the understanding of this gray world is only stint, if my friend you read some strange phenomenon after you tell me the real reason you think, will be my most want to get the knowledge, thank you!

Flow explosion

Traffic includes website traffic and advertising traffic. We use Uv,visit and PV to measure the flow of the site, using impression and click to measure the flow of advertising. No matter which measure, in short, the more the more, "a lot of pressure body", can always cheat the big silly Alexa, always fooled advertisers. So, when I suddenly found that one of my ads in one hours of sudden traffic explosion, I was not surprised.

Figure 1: The above figure is one hours before the traffic, the following figure is one hours after the performance, I blurred some sensitive information (after the same)

As you can see from the diagram, just one hours (please note that the top two table header time), this media total advertising flow, especially impression, has been magnified n times. Because one hours ago, the impression number bizarre (or more accurately click the number bizarre, because too big), at the beginning of the CTR high is unbelievable. With our warning to the relevant site, impression's problem was solved within one hours. It's amazing.

It is simply impossible to look at the data. I never pay attention to how huge the impression will become, because there are too many factors affecting this measure, to know that QQ clients can contribute 1 or 200 million impressions! to your advertisement in one day. Any popular game of BBS brings you millions of day impression also see more not strange. However, you can never ignore the size and change of the click Value, such as the following figure, this is a continuous on a Web site advertising time of the click:

Figure 2: Traffic explosion refers to a short period of traffic surge, as a supernova explosion (note 14 and 18 points of Click)

Such traffic trends are unlikely to occur if it is not a problem with statistical tools or if the server becomes self-conscious wall E. The booming of 14:00 and 18:00 is a fact that cannot be explained by conventional thinking. I just want to say, if you really want to use artificial means to do something, please, give some professional spirit good! I have a question here, AdSense (AdWords) will be blocked off such abnormal flow?

Unique is a good thing.

One day, when Tenly is studying a conversion rate (conversion Rate, which is the number of visit that the target site brings, conversion Rate equals 20%. means that 100 click to bring the site 23 visit, the other click in the page is really open before the browser is closed, the conversion rate was found to be less than 5%. In general, conversion rate in the 30%~80% range should be reasonable, after all, different forms of advertising caused by the number of false clicks is different. But 5% is unacceptable, which means either click or visit data is out of the question.

Tenly starts with the click and immediately discovers the problem. His method is simple, see the unique Click directly. The Unqie click and unique visitor are similar metrics that, if a period of time (this time is slightly different depending on the monitoring software, but the monitoring principle is exactly the same) you use the same browser on the same computer to repeatedly click on the same AD, The click Value increases, but the unique click does not. The larger the value of the Click/unique click, the more times an ad is repeatedly clicked.

Most people don't do it in a single ad, and only a few are too bored to do it, but they don't make a whole lot of difference. But when we see the site's Click/unique Click, we were surprised that the ratio was as high as 11 times times more. This means that the average person who clicks on the relevant ads on this site will click more than 10 times in a row! Other sites are normal. If it were you, what would you conclude?

Figure 3: Cannot believe up to 11 times times the Click/unique click Value

If you use Click to calculate CTR (click Rate), ctr=2.11%, very high. Using the unique click to compute the CTR, it is 0.19%, returning to a normal value. So I think the unique click is basically "real" clicks.

Unique is really a good thing. But now it is not good, the reason is that the current "website optimization" technology, has already crossed this step, such a low-level error is difficult to see. So when I see this case these days, I feel so interesting.

Don't be credulous Ctr can be greater than 1%!

A lot of friends asked me Ctr probably more normal. I have no answer. It can be said that most of the CTR I have encountered are surprisingly good, often more than 0.5%, or even higher than 1%. However, if I defying say what I think is true, you may not believe it: The true Ctr of the vast majority of China's ordinary online advertising (except for SEM) is good, and the CTR of those web portals that are so complex may be lower. But we see that the final statistical software is always much higher than this number. So when you ask me what is normal, I really don't have an answer, the accuracy of the answer depends on the sincerity you want to know the true image.

In any case, too much CTR is still a cautionary note, especially today, when impression generally explodes, meaning that the click is worth getting bigger. If ctr>1%, it means that every 100 people who see your ad will have at least one click on your ad, which is too high, and don't believe you can immediately launch a field survey (survey) study in your office.

Some forms of exaggerated advertising, such as suddenly full-screen pop-up and add a fake "X" (off) tag ads, or nude (naked beauty) and other deceptive ads will win very high clicks, even CTR can exceed 50%, but such "靐 advertisement" I believe you will not use, which is equivalent to click Investigative (click fraud). If your ad is normal leaderboard,pip, skyscraper or Textlink, and the content is normal (text and pictures are not too "provocative" or "selective"), please believe the approximate true Ctr range:

First screen leaderboard:0%~0.1%

Pip and skyscraper:0%~0.05%

textlinik:0%~0.02%

If it is SEM, may be higher than 0.5%, after all, SEM is a relatively segmented form of advertising, but more than 1% I am a little worried

Is it really smaller than you think?

Of course, the above value may be very arbitrary, after all, the click of the ads and the site and advertising matching degree. However, believe me, no matter how the content matches, more than 1% of the hits must be worth conducting in-depth investigation.

Helpless Action Rate

You may be a little strange to see this little headline. Bounce rate we common, and action rate is what? Action rate is not a standard metric, but in my work I have standardized it to measure the really meaningful click behavior on the page. My definition of Action Rate is: Action Rate = subpage visit/landing page visit, which can be simply understood as: the number of accesses to the next page from the landing page is the amount of access to all landing pages. In this definition, the higher the action rate, the more access to the landing page will click on the link to enter more other pages, which means landing page more effective.

Why do you need action rate such a metric? To measure the validity of landing page with bounce rate not on the line? And obviously, according to this definition of Sidney, if the action rate high, then bounce Rate should certainly be less.

Yes, that's what I thought at first. Until one day I found a "strange thing" that I was constantly discovering. The discovery of this "strange thing" began with a bounce rate. In general, commercial Web sites (such as promotional products, promotions, etc.) of the bounce rate will be relatively high, can be less than 70% in fact very uncommon, so when I found a probably only 40% of the bounce rate similar site, I was very surprised. Then I saw that the site pv/v over 2, while time on site was an incredible 205 seconds-definitely a pretty dazzling achievement.

At first I thought the site really had a good design and quality of visitors, but when I looked at the navigation summary report in top content in Google Analytics, I started to see that my initial judgment was completely wrong.

Figure 4: Can you tell? Home Page There is a large number of repeated refreshes, while the real content is no

Can you see anything in this picture? The two pages in the Red thick line box are exactly the same name, which is actually the home page (/sales/.../?ad=1509) itself. All we have to do is look at the right part of the picture, which tells us that 53.69% of all traffic to the home page leaves the site, and another 46.31% goes to the back page (Next pages). Now, the problem appears, please look carefully: 46.31% of the next page of traffic, 45.11% is to visit the home page itself, only 1.2% of the traffic in the access to other content! What does that mean? I remember a friend once asked me this question, I told him at that time, this shows that there is a large number of home page refresh!

In addition, this data also tells us, this page action rate roughly only 1.2%! 100 people come, about 1 people will really click on the link in the page, the other 99 people will not click on the page on a pile of interesting links. The 1.2% action rate and extremely low bounce rate, very high pv/v, and extremely long time on site simply can't match, and it's just out of the way! However, the true image often comes to the surface when contradictions occur!

Careful analysis of this data report is not difficult to reproduce the flow of the first page: many visits (visit) into the home page, a part of the site to leave (nearly half), the other part (the other half) although stay, but do not click on other links, just refresh the home page. This cannot be interpreted as an ordinary visitor's behavior.

My first feeling is the Trojan traffic, countless "broiler", "zombie" a period of time (usually a few minutes) to visit the target site, this is "hackers" pediatrics. Because GA in the definition of bounce is a single PV for the definition, so these visits, although only constantly visit the home page, but will not be recorded as bounce, so bounce rate will be very low, pv/v will be very large, visit the site's long also become very long! But it's totally rubbish, it doesn't mean anything to the website!

You will ask, that let these Trojan traffic continue to access other pages do not visit the home page not on the line? On the one hand, these Trojan traffic can be controlled to access a website page, but to let them division of labor continue to ask some links in the site can not be so easy. On the other hand, let the Trojan traffic access to other pages will be more obvious exposure. So, the end result becomes the strange website that the graph shows.

In addition to Trojans, robot traffic is similar to this performance. In short, not real human behavior, in the GA subdivision report is very easy to show off.

Well, write here, you may have a lot of questions, of course, there are many want to discuss and argue with me. I still say that the manual flow is a gray area, I am not an expert, I sincerely hope that you can discuss the debate with me. I don't want to affect the industry chain, I'm just curious, I want to know the true image.