Challenging the public intelligence in Web Analytics (1)--bounce Rate

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

Objective】

This series of articles should be specially written by Netconcepts Allen. It is also the first article to be completed on the commute to the subway.

Must the wisdom of the masses be the correct wisdom? Perhaps you have heard the word "the mob of the majority". In no matter where, because of the influence of "popular wisdom", a higher IQ does not always ensure that we make the right decisions, as the 36KR article says, smart people can sometimes be silly. So, this series of articles is to look at the field of web analytics which the public wisdom may let you "go astray". In the first article, we will focus on bounce rate.

Body】

Bounce Rate, web Analytics, one of the basic metrics, I believe that everyone is not unfamiliar. Bounce rate seems to represent a "magical challenge" given to us by some kind of web analytics, and if we have uniforms or the power to dominate it, we seem to have achieved something remarkable.

So people believe that the lower the bounce rate the better, but as the old saying "too much," lower bounce rate may not be as good as we expected.

Super low bounce rate implies monitoring technical problems

In any case, if the site (or some of the main landing page) bounce rate less than 5% or even below 10%, there is a potential for monitoring technical problems. There are at least two conditions that can cause bounce rate.

First, the frames page (that is, the page is nested through an IFRAME or frame) of the parent pages and subpages, if all with the monitoring code, then in most cases, when the page is loaded in the browser, the monitoring code will be run in a short time two times. For web analytics tools, such as Google Analytics, it is recorded that two different pages are opened (because the URI of the parent page and subpage is definitely different), so this situation is clearly considered not a bounce, Even if the user does not click on any of the links on this page to browse other pages. In the example I see, in this case, the bounce rate is less than 5%.

Second, if you accidentally add two identical monitoring codes to the same page, or for some reason, have the same monitoring code run two times. So when the page is loaded, for Google Analytics, as this page is refreshed once, the GA does not calculate this as bounce. In this case the bounce rate is even lower than the bounce rate in the first case. For Adobe Sitecatalyst, however, the problem does not exist because of the bounce rate definition.

Of course, there may be a third situation, that is, traffic cheating. In the case of botched cheating, the robot (or broiler) unifies the website visit according to some kind of preset procedure, will very serious reduce bounce rate. There is no doubt that this is not worth happy.

Bounce rate low may mean other negative information

If bounce rate unreasonably low, we can judge the technical problems that arise in this case. But if it is low, but it does not reach a very low range, can we breathe a sigh of relief?

I do not dare to jump to conclusions, in a sense, bounce rate low may also convey some other negative information, the biggest problem is that too low bounce rate may mean the lack of fresh traffic.

In the Web site analysis has a conventional law, that is, for commercial sites, bounce rate values often and% new visitors has a linear relationship, some even have a similar 1:1 relationship, quite interesting. It can be argued that, in order to obtain more new traffic, our price is often sacrificed to better bounce rate, this sacrifice is fully understandable. The following figure is the corresponding relationship between the bounce rate and the new access ratio of nearly 50 size sites I have monitored, and has a more obvious linear correlation. This means that we do not expect the actual existence of the "paradox"-if I expect bounce rate lower, then I should try to save my old users, but the proportion of old users, the site traffic situation I am afraid not so optimistic.

Of course, things are not absolute, bounce rate is not just determined by the new rate of access. For example, in the above illustration we also see that there are some points that are significantly off the trend line in the lower left corner, showing a "less beautiful" side-a very low new access ratio (10%), but still nearly 50% of bounce rate. Moreover, the overall trend of bounce rate and% new visits shows that the bounce rate performance is very discrete when the new rate of access is low. We seem to be able to conclude that the high proportion of new visits almost certainly affects bounce rate, which tends to be too high, but when the rate of new visits is low, the level of bounce rate is influenced by more other factors. We'll discuss these effects later.

Bounce rate low also means another possibility, especially when Bounce rate suddenly become low, we believe that there is an abnormal guidance or temptation. If the guidance or temptation is intentional or justifiable, it may be misleading to the website operator if an unexpected situation occurs. One example I've ever encountered is that the bounce rate of a website suddenly dropped by about 20% a day, which is encouraging. But looking at the data carefully, we found that the sales of the site did not occur any increase, which is not in line with commonsense. We found that bounce rate dropped the first one or two weeks, our one channel (art sales channel) on-line, we further found that in the bounce rate down the day, the site has a lot of traffic from the search keyword "human art" or "Human photography", And these traffic landing page is our art sale channel. Interestingly, there are some paintings or photographs with little or no clothes on the channel, and they do get better clicks. But anyway, that's not what we expected.

From the point of view of the actual operation. Bounce rate is a lot more enjoyable than high quality, but be careful, we do also need to rule out some very important reasons to ensure that our Bounce rate reflect the real effect, not the false prosperity.

Exclusion of monitoring factors caused by bounce rate abnormal height

Conversely, if the bounce rate is really high, it is not necessarily really a big problem, or, we should at least first ensure that the high bounce rate is not the result of monitoring.

The main monitoring is not in place easily on pages with more non-HTML interactive elements, such as a page with many links made up of Adobe Flash or JavaScript.

Whether it's Google Analytics or Adobe Sitecatalyst, additional code processing is required for non-HTML interactive elements, such as using GA event tracking or virtual page monitoring to mark these elements, This records the behavior of the user clicking on these elements, otherwise, all the interactive web analytics tools for these elements are not perceived. For example, in the following page, all the circled buttons are the buttons that the user clicks on in the middle of the page, but they're all made by Flash, and the new pages that you click on are not your own (but jump to someone else's page, which means you can't monitor the pages you click on)--that is, If there are no additional monitoring settings, this page, as landing page, even if every visitor comes to this page and clicks the buttons directly, the bounce rate of the pages will be very high. This does not reflect the real situation.

I think, bounce rate if really bad, do not panic, first of all to exclude possible monitoring problems. When you face up to 95% bounce rate, you should do this: the first step is to ensure that all user clicks are monitored, especially for non-HTML elements; The second step is to see whether new visits to the page (or Web site) are up to 90%. If you've checked and eliminated both steps, you need to be really worried about the bounce rate of up to 95%.

Oh, in fact, not necessarily so worried, we continue to look, if bounce rate should not be so high, we should do.

The two situations that cause bounce rate to be identified and handled separately

Without the other objective reasons mentioned above, if the performance of bounce rate is still poor, we need to improve from the business (and even the business) level.

Usually, we think bounce rate high is the site out of the question, specifically, is the main landing page poor performance. But in fact, for bounce rate analysis and optimization We do not simply think that the problem is the site.

Removal of technical problems, in essence, the decision bounce rate high and low only one factor-matching, that is, traffic and web site matching decision bounce rate level.

To cite an inappropriate example, a website is the sale of maternal and infant supplies, but the source of traffic is all Diablo III players, even if this site to do a very reasonable appearance, goods extremely cheap, I am afraid bounce rate will not be low. You can not simply think that the flow of the fault, or the site's fault, the fundamental problem is that the two are completely "donkey head not horse mouth."

However, as the only correct answer to the "match" is not the best answer, the experience of learning to tell us that the theory is often unreliable, not it is not correct, precisely it is too correct, so empty. We certainly hope that bounce rate lower, so we certainly hope that traffic and site matching, but the question is, how can we do match?

Luckily our problem does not enter the compulsory education textbook, otherwise it will always be a theory. Again, luckily we are analysts, and we know how to cut the so-called "match" problem further. Because the two ends of the match are traffic and the site, then there are three kinds of mismatches--either the site is not suitable for traffic, but the flow is not appropriate for the site, or the site is not suitable for traffic, or both are cock silk, are dwarfish, are not. The third situation I will not say, because in fact, if the first two cases can be resolved, the third case is completely similar to the solution.

Whether the flow is reliable

Flow by unreliable identification, to use the contrast method. In general, for a site, or for an important landing page, if part of the flow channel bounce rate performance and some poor performance, we can assume that some channels and the site (or landing page) have a better match, Other channels are poorly matched. For example, in the following example, the weibo.com crowd doesn't seem to like your site that much.

Because each source of traffic represents a certain part of the population, so we can through different flow channels bounce rate to read out whether the corresponding crowd of your site or a certain landing page satisfied.

In this example, we can also look at the third largest source of traffic google/organic, it is not very bad, but after all, compared to its friend Baidu/organic performance is much worse. To this end, we have the curiosity to see exactly which keywords bring the traffic and the site does not match. Decomposition of the search engine traffic is particularly meaningful, because the search engine's keyword contains the user's interest information (obviously, only if you are interested in something, you will search it), so different keywords bounce rate directly reflect the site's satisfaction to these interests.

Well, it seems that a lot of people are interested in Rafi and ray glasses, and can also find out through the search engine that your site has this information, unfortunately, or these people are not really meaningful to you, or your site content is not very good with relevant content to attract them.

Web site unreliable

Generally speaking, if all flow channel's bounce rate are very high, that is not the flow itself problem.

The example above is not particularly typical, but it can be seen that this important landing Page (important because of its large flow) the main source of traffic bounce rate performance is unsatisfactory. If here 360daohang_navs bounce rate also up to 70% or more, then we can basically judge this page bounce rate high level is not the flow of the past, but itself or other factors caused.

Here is a little explanation, you can see the name of the landing page is "official Website Product Details page", in fact, it is not a page, but the site of all the Product Details page collection. about how to carry on the analysis of the Web page cluster, is the website Analysis very important basic lesson, in after the blog article, I will specially use an article to carry on the introduction.

In this real-world example, we ultimately determine the quality of the page, rather than the flow problem, resulting in poor performance bounce rate. We open any Product details page, such as the left page, we can find that, in these pages, the description of the appearance of the goods, the description of the integrity of the protection and the description of the specific items of attention are extremely deficient. Not only this kind of merchandise, we found that the site all the details of the product page is extremely lack of detail.

You may wonder whether the screenshots are intentionally not intercepting more content, please believe me, this screenshot reflects the content of this product in this page all the content. The part that I did not intercept is just the copyright information of the website and so on footer content.

In addition to the details, there is a lack of convincing people to click on the orange button "put in the cart" on this page. All the merchandise is so, so we're not surprised. The Product Details page, whether as a whole, cannot get good bounce rate performance.

This example shows us that if most of your major traffic channels are bounce rate on a landing page, the biggest problem is not that the traffic itself is good enough, but that landing page has a problem.

What kind of bounce rate is good, what is not good?

Finally do a bounce rate benchmark summary. Because constantly have friends to ask my own bounce rate whether healthy, but I always feel not give them a satisfactory answer.

The reason is that I am "anti-benchmark faction".

The benchmark of website analysis, because each site individual has many differences, so it is difficult to compare. In an article, I specifically mentioned that too much pursuit of benchmark actually does not have much meaning: the site analysis of the top ten misunderstandings and alternates (iii).

But we always wonder, what kind of bounce rate is healthy?

I will give some empirical data I think, but only for reference, remember, only for reference.

To give these benchmark, I first want to make sure that an important factor does not interfere, that is, the site's new user ratio is certain, and about 30%. In this context, let's take a look at what the various bounce rate of the site might say:

Less than 10%: bounce rate, less than 10%, may indicate that the problem is being monitored, except for some super sites, such as Google, Baidu, or Sina.

10%-20%: The chances of monitoring problems remain, but another possibility is that your site has a strong incentive for visitors to click on the content.

20%-40%: Brother, your website is doing well!

40%-60%: Not the best, but not bad!

60%-70%: Well, look at the main landing page which did poorly. Because the percentage of new users is 30%, more than 60% of bounce rate may indicate a problem with your site's content.

70%-90%: There are serious problems that need to be corrected immediately landing page.

More than 90%: Well, need to update the site immediately, if not, look at the site is not a server problem. 70% of the old users in the case there are more than 90% of the bounce rate, either the site has been unmanned management, or is really a problem with the server, can not load the page smoothly.

Next Notice

I'll write a few of my readers ' friends in this series. If you have any expectations for this series, please leave a message for me. For this article have any suggestions or want to discuss with me, please leave me a message. Thank you very much! The next issue will focus on one of our common tools-a hotspot map. Let's look at what kind of traps the public wisdom may give you in this field and how you can avoid them.

"This article copyright belongs to CWA and original author sing all, if want to reprint, please contact author"