Website Analysis: E-business key digital Optimization online section (next)

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

"One sentence per issue"

Without effective protection of innovation, there is no innovation. We have no jobs, no Zuckerberg, not a national quality, but a system.

Objective】

In the previous two articles in this series (above, Medium here), we talked about how to optimize the two aspects of key performance figures in E-commerce: identifying key drivers, initiating analysis, and attempting to upgrade key drivers. In this section, we will show you how, through some specific methods, our efforts to promote key drivers can ultimately be achieved.

Body】

Remember the optimization path we discussed earlier? In this path, we emphasize starting with the definition KBR, then breaking down the drivers that affect KBR performance, and then determining which of these drivers are the underlying drivers, which are not the underlying drivers, and then trying to analyze the underlying drivers and proceed with the improvement, Try to analyze the non-basic drivers and proceed with the improvements, and then test whether your improvements are effective and fix them effectively (optimized). Since optimization is not possible for all people and interests, finally you have to do dynamic processing (customization) on the basis of optimization.

　　

Figure 1

Today we look at testing and customization, and how they both make our efforts to optimize the final landing.

Analysis is not all

Website Analysis (web Analytics) or digital marketing analysis (Digital Marketing Analytics), are inadvertently for us to emphasize the importance of "analysis."

Analysis is of course important, especially in the country we are in and such a complex truth and falsehood of chaos, there is no serious independent respect for the facts of the analysis, we can not find all kinds of cheat to know many truths, will be some kind of we have to worship the power of the bluff. But for the business purpose of the website we want to optimize, the analysis is not entirely.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

I have said (http://www.chinawebanalytics.cn/10rules-for-using-wa-tools-1/, principle V) that, in most cases, web analytics tools can tell you what happened (what), But you can't directly answer the reason for what happened (why). In order to find out as much as possible, we then use one of the most basic methods-subdividing, trying to cobwebs the truth. But in many cases, it is hard to believe that this is the cause, and we are just saying that this may be the reason. The following example illustrates this situation:

I want to know why

For many years I have been trying to understand the causes of a lot of strange phenomena, but some of them have been solved, but some of them have not.

In the sweltering summer of June 2010, I was plagued by a problem about what site elements could be more effective in facilitating orders. This site is a travel class website that provides users with online hotel booking business. This site has several types of pages, my first step is to find which page has the most impact on the conversion rate.

　　

Figure 2

It's hard for me to do the Google Analytics e-commerce monitoring and set up a beautiful goals and tournaments site, a tool can solve the problem--ga $index, this metric (metric) to measure the contribution of each page to transformation. Because each transformation means earning a certain amount of money, so if a visit generates a 10-dollar conversion, and the visit browses through 10 pages to produce this transformation, then each page contributes 1 dollars for the conversion, that is, for this visit, the page's $ Index=1 dollars. Therefore, if a page's $index is very high, then this page is the key to the transformation of the page.

　　

Figure 3

In fact, without this tool common sense can also tell us that the booking page for the impact of the transformation was undoubtedly severe. So we focused our efforts on booking page.

Determined that a page has a greater impact on the transformation than other pages, we are a step closer to solving the problem, and the next thing we need to do is to find out what elements in the page are most likely to affect the transformation.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

Our approach is to make the booking page of the click of the hot map, click on the hot map is not the perfect tool (in my chinawebanalytics.cn the most recent "Challenge website analysis of the Public Wisdom" series of articles will be for you to click on the hot map of some of the lovely and defective places), But you can give us at least one answer: which areas of the page inspire users to continue to explore the interest.

　　

Figure 4

In this hot-click report (the left and right two pictures are actually the top and bottom parts of the page, because the original page is too long, so we cut it to two charts, we see some obvious points of interest--"expand quotes", "Pictures", "near hotels" and "hidden quotes" get more clicks. One thing that is surprising is that "comments" do not get the high attention they intend.

The question of "commenting" is clearly confusing, and the Web Analytics tool is a good way to tell us "what," but it's not clear to tell us about "why" here. We can guess the reason "comment" is less concerned because it doesn't get the ideal location, but now it's only a guess. (In the "mass wisdom of the Challenge Web Analytics" series, we'll show you what tools and methods are available to further validate this speculation)

This kind of confusion often occurs in the course of our website analysis. Many times we can not help but sigh, know more may not be good, we have to bear more unknown.

This is not the end of the case, because we need to solve the problem of what elements have the most significant impact on conversion rates. So we went on to look at the impact of these top clicks and the elements that we thought were important to the transformation. We made the following form based on the transformed data:

　　

Figure 5

Each column is the number of clicks of an element and the corresponding amount of order conversions that occur after those clicks. As you can see from the table above, if a user has a view of viewing comments or viewing pictures, the chances of conversion are greater than those who click on other elements. In particular, "view comments," although it does not have a good click, but has the highest conversion trend.

Data to do here, we can identify some things, but we also have a few more questions. Why does the promotion of conversion seem most pronounced in this case? What kind of commentary promotes transformation, or is there only part of it that promotes transformation? Why does the hidden quotation promote the conversion rate more than the launch price to promote the conversion rate? These questions, especially those related to the content of the comments, I believe you can not directly answer the Web analytics tool.

But no answer does not mean that we are not able to achieve optimization, the following content will help you to solve this problem in other ways.

In the above example, we can learn from web analytics tools that there are some site elements that are more important to transformation, but we still don't know how to modify these elements to further improve the transformation. The analysis tells us what we should be trying to do, but does not give us an answer as to what we should be doing.

Where does the answer come from?

Now, like me, you want to ask a question, how do we get the answer?

We need to rely on two things, one of which is the great wealth that the universe has bestowed upon us-our experience and intuition, and secondly, the fact that these experiences and intuitions have been validated.

In the above example, we know that "view comments" have a great deal of relevance to the conversion rate, and we know that "review comments" has not received much attention, so we have, with our experience and intuition, put forward solutions to this problem that we think might work:

1. We need to "view the Comments" button, and even the comments themselves on the page in a more prominent position, and give more explicit guidance, in order to obtain greater user attention;

2. We believe that "view comments" can be extremely relevant to conversion rates because of the role of positive comments, we need to guide users to send more positive comments to reassure other users.

Will these two solutions work? Experienced practitioners would say, "no problem, but it's true that none of us can make 100% effective promises before we make these changes, not only we can't, all the constellation analysts."

If you're an analyst, the best way to "protect" yourself and your company is to tell your boss-I believe these measures are positive for further improvements in our conversion rate, but for the sake of insurance, we need to do some testing to make sure that they really work, And these tests can tell us more about the user habits and ideas of the key information.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

If you say so, that's great. The biggest feature of the analyst is that it's not easy to say anything, they seem indecisive--is your friend judging you now? Congratulations, you are already a qualified analyst.

Test

There is no doubt that testing is the best weapon God has given us. Testing does not represent something formal, and if we fail, we can say that this is a test. So we have more freedom and a more relaxed mindset-it's great for finding something really valuable.

Another important part of the test is that no one can represent all the users, the boss's opinion, the experience of the operating department, and your own intuition are not users ' feelings. Haven't you ever been sulking?—— your boss's advice makes you think the two-month plan is going to go down in a minute, because your boss believes you can best represent the user's feelings, but you know it may not be so, but you have no way to persuade, or even exert a little positive impact of the possibility is not. This time, perhaps we need to do some testing, let the real user to participate, to vote, to prove who is the closest to the right.

No test, no final victory, or through a temporary setback to lead us to victory, no matter what, test can give us some real landing conclusions, otherwise our analysis may only float in the sky, our work is probably only floating in the air or just a few data to do several statements.

There are many types of tests, such as those shown in the following figure. In fact, many user usability experiments are a test in themselves.

　　

Figure 6

AB testing is the most important and most commonly used tool. Unfortunately, there is not a single article in my blog that specifically describes AB testing that I can't cite. This article didn't really want to tell the basics, but for the AB test, I'd like to say that this is one of the most important basic lessons of web analytics and even all analysis, so I'll introduce the AB test in the quote section below, if a friend you already know can skip directly.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

AB test

There is no comparison without the truth.

In our neighbourhood, where domination does not allow people to get any foreign information, or when the country sends some IT personnel to work here, the whole world screams--they are not afraid of their brains being "contaminated"? The country does not allow external information because it does not want people to make comparisons.

But the site analysis must be compared, we are all the times in comparison. Do subdivision is to compare, I have always been on the tip of the mouth of the "No Subdivision no truth", in fact, that there is no comparison without the truth.

The AB test helps us achieve comparisons. The AB test works well because it provides a fair comparison of the environment-we don't compare an elephant to a hippo compared to elephants and elephants.

The comparison of the most feared problems with the AB test helped us avoid it. First, comparisons are influenced by many external environments, such as time. You can compare the same thing at different times, such as a website on June 1 on the new home page, you can compare the entire May and the entire June home page performance, but the problem is, May and June There are many differences, such as June may be done promotions, or May traffic source is mainly the search engine traffic and June mainly direct flow. Second, the comparison itself may be affected by the sample, especially when it comes to user usability testing, we are less likely to randomly select 10,000 people for usability testing, so we can only assume that some of the samples we choose can represent the majority, but in fact no individual can really represent the majority.

The reason the AB test applied by web Analytics can solve these two major problems is that it enables comparisons to occur randomly at the same time (rather than the sample, unless you want to limit the sample). This sentence is a bit awkward, to put it simply, the AB test is able to compare two or more different comparison objects at the same time, based on all users.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

A common example of a website analysis-we now feel that the home page is not so good, so we let design department design a new home page. The new home page looks good, but we don't know if our current site users like it. So we do AB test-we put the original home (a page) and the new design of the homepage (b page) at the same time on the line, our users or enter the domain name of the previous site, all the flow channels are still pointing to the previous home (a page). But the wonderful thing is that although all traffic is entered a page, but a part of the user (typically half of the users) see not a page, but a B page, and they will always see the B page, while the other part of the user still see a page, until the test is over. Does this section see the user on page B make us select specifically? Absolutely not, they are completely randomly assigned to page B.

　　

Figure 7

So it's great that AB tests do different versions of the same page-it's not about time, because the tests happen at the same time, and it's not affected by the sample, because all users can participate in the test and the user's distribution is completely random.

By the way, two concepts are worth remembering--the user assigned to the original a page is called the control group, and the user assigned to page B is called the Exposure Group (exposure group).

AB test is not only used to test two objects, in fact, can test two or more than two objects, in the test more than two objects, the control group will increase, the proportion of each object assigned to the corresponding decrease.

The principle of AB test is so simple, but there are some other considerations and methods of use, in today's article will not talk more, we in the future of the article to do the corresponding special introduction.

The AB test can really help us solve some of the problems that are hanging around. In the previous example of the hotel Web site, we could decide what kind of picture or comment would be more transformed by AB test.

The product manager told me that his experience was that a picture with a clean bathroom would make people feel better about the whole hotel (in fact, this experience can be validated by the method of Web analytics, that is, to see if the photos of the bath room will be more viewed, but due to the impact of the location of the photo emission, this validation is slightly troublesome. So we put a picture of a hotel into three categories, a kind of no bath photos (A), another class with ordinary bath photos, did not do a special optimization (B), the last category is the selective optimization of photography and treatment of bath photos (C). On this basis, we have made a special three versions of the hotel's booking page, except for the different photos, all the same. #p # subtitle #e#

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

Before the test, we made a little bet, and for C, it was agreed that there would be a better conversion rate for A and B. I chose B might be a little better than a.

The final result is a slightly better conversion rate than B, but the difference is not particularly significant, or there is no special statistical significance, but C conversion rate is really better, higher than A and b about 10%. We think this is a terrific conclusion and we find that good bathroom photos may be valuable for conversion. I used the word "may" here because the results of a hotel AB test were still not representative of the overall situation, and we needed to do as many AB tests as possible to get a universal conclusion. In this case, the AB test is not so simple, the complexity is not the principle, but the need for meticulous and patient implementation.

The test always has the harvest, always

A friend once told me that he did a lot of ab tests and found out that the AB test didn't always solve his problem.

His reasoning was that he was often able to rely on the results of a direct judgment, and the results of the AB test were consistent with his intuition. If his intuition does not show the quality of each version, the results of the AB test are very close enough to be statistically significant. Does AB test really work?

I admire the friend's intuition, but I think it is narrow to understand the AB test, or it should be said, that we can't just do tests and not analyze.

AB tests often contain two layers of meaning? The first level of meaning is to compare the conversion rate (or CTR) of different test subjects, which is called "macro Contrast Analysis", and in the example shown in Figure 7, the version a page and version b page are randomly assigned to equal numbers of users, but version A There are only 50 registrations (singups), and version B has 75 registrations, so we think that B-page conversion is better than page A. But we will certainly not be limited to such a simple comparison, and will not immediately take a page, replace the page B. We will also continue to study the differences in user behavior on the two-version page to speculate on what caused the conversion of version B to be better, and it is possible to focus on the strengths of each of the two pages and design another new page version C and do a AB test again. This is called "Microscopic comparative analysis". This is like our experimental mice in medicine, the two groups of mice are the same, and were artificially caught in the same disease. We were feeding a new drug to Group B, and Group A was only giving a placebo (if the rat world understood the placebo), we found that group B mice were getting better and faster. The macro-contrast analysis concluded that drugs may be effective, and microscopic comparisons allow us to continue to explore what biochemical causes make the mice in Group B better and faster, and whether there are other changes in their bodies.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

So I think that implementing the appropriate AB test can always bring us a harvest, even if no version in one test can really win, we can find other user behavior differences, and based on these differences to analyze user habits, and thus still be able to create a better version of the possibility. This is very meaningful.

The following case illustrates this situation. The test of this case revolves around an ecommerce site AB two versions of the home page. The website hopes that the new version (which is considered to be more atmospheric) will not be worse than the original version in key metrics such as bounce rate, to ensure that it does not affect performance when replacing the old version. The purpose of the test itself is not to be particularly interesting to me, and to be honest, the purpose is a bit bland, but it is a good battlefield to study microscopic comparative analysis.

Let's take a look at the conclusion of the macroscopic comparative analysis:

　　

Figure 8

From the data results, even the difference between the larger shopping cart add rate and conversion rate, in fact, there is no significant difference. In some indicators, the old version even better, such as bounce rate and exit rate, and in the E-commerce conversion rate, the new version is better than the old edition. These data do not reflect the difference between the two, or macro-comparative analysis does not give us any strong conclusions.

But the microscopic contrast is different, under the same calm sea surface, is different scenery. Microscopic comparison of the methods we use to click on the hot map to see the user behavior differences between two pages.

　　

Figure 9

　　

Figure 10

Two pages Click on the comparison of the map to give us a lot of information, we can see the B version (the new version) of the first screen click ratio of a (original) higher, which shows that the new version in the first screen play more advantages. The new version of the navigation also obtained more clicks, seems to derive from the B version of the further rationality of classification adjustment (here we can only guess is this, if you want to know the real reason, we can again specifically for the navigation classification to do a AB test). Another notable feature of the new edition is that on the top of the large navigation bar, "new" access to 24% unusually high percentage of clicks, and in the old version, we also noted that the bottom of the "Global New" column "More" also obtained 7.3% of the click Ratio, showing the user's strong interest in new products.

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

However, the old version of up to 8.9% of the "second Kill" in the new version is "clearance" replaced by the same position, but only received 0.2% attention, it is surprising. Apparently seconds kill brings people joy, and clearance makes people feel cheap and low-quality.

In this case, the macroscopic analysis does not give us much useful information, but the microscopic analysis gives us too many meaningful stories. If let me design a better C version, perhaps I will be in the B version of the basis, adjust the "clearance", replaced by "second kill". And, emphasize "new product" in the focus diagram. A simple little example tells us that the AB test is far more interesting and more meaningful than the direct look.

Test optimization of basic driving factors and test optimization of non-basic driving factors

In the first two articles of this series, we mentioned the underlying drivers and the non-underlying drivers, as shown in the following illustration:

　　

Figure 11

Because the driving factor is divided into two kinds, the test work of the driving factor optimization is divided into two kinds, namely the test optimization of the basic driving factor and the test optimization of the non basic driving factor. Of course, these two types of methods are not different.

The two examples mentioned earlier, one is the example of hotel photos and comment impact conversions shown in Figure 2, the other is an example of the Home page optimization shown in Figure 9, Figure 10, which can be categorized as a test optimization for a typical, non-basic driver, since the former is the product page, which can be used on one or some products (this is the hotel product Optimization plays a role, but not entirely global optimization. The latter, home optimization, is the test and optimization of the underlying drivers that affect the global, such as the revisions to navigation and catalogs included, as well as the critical impact of the home page as a very important "bridge" in the transformation path.

For testing and optimization of non-basic drivers, activity/match is an extremely important content. For example, as shown in the following example, the same activity, exactly the same activity landing page, except that some of the items in the landing page are more clearly labeled as an attractive "premium price" for the product.

　　

Figure 12

This test shows a significant impact of the optimization of non-basic drivers on the effectiveness of short-term marketing campaigns. Some of the goods were marked "value" after the flow through the B-page conversion rate has been significantly increased, resulting in sales more than twice times the promotion.

My experience is that the AB test of the underlying drivers may not be easy (because we are not likely to always optimize the underlying drivers, we are progressive), the analysis and optimization of these factors depends on the vision of the company leadership and the real emphasis on "important but not urgent" things that we often do not touch, Therefore, the test and optimization of basic driving factors always appear to be met and not to be asked. But I would like to say that organizations that are really able to work on the underlying drivers are more likely to surpass their peers and achieve a real competitive advantage.

By contrast, the AB test of the non-basic drivers gives us a big stage--because the need for non-basic driver optimization is always recurring, and always get more attention (although this emphasis is always accompanied by a short-sighted pursuit of short-term interests), so we can continue to have the opportunity to try and verify, And it's easier to get results and a thumbs up. This case in Figure 12.

Whether you are lucky enough to find time and opportunity to optimize for the underlying drivers in addition to the optimization of the infrastructure drivers, the preceding three examples also show that, whether it's a basic driver or a non basic driver, after you've completed your analysis and solutions, It is only through testing that it is possible to finally give us a real proof of the optimization of the landing.

Customization

The AB test tells us a lot, and it also tells us that if there are multiple versions, we can always find that some versions are better for most people, but for a small number of people, they seem to like other versions.

I have said that the biggest advantage of the Internet is that customization can be achieved. We do not need to ignore a little bit of preference, or we can use a great way to attract a small number of people, even with the promotion of most people's conversion rate, we can drastically increase the conversion rate of a small number of people in order to eventually achieve "common prosperity."

This is not so difficult to achieve an ideal, at least compared to our real world, it is much easier.

Customizing this part of the content, I do not intend to continue to explain in this post. Instead, I'm going to open a special article. A friend said, Miss Song, your article is too long, we see the difficulty. I think I also need to take care of most of the readers ' preferences, not just the happiness I write. Please look forward to it.

Summary of this series

This series finally ended, this series of articles, from the last chapter to this article, separated unexpectedly half a year, today finally let me be able to catch my breath. I have always thought that the analysis of the Web site, digital marketing analysis, although very important, but the entire closed loop can not be less the goal of the establishment, and finally let the analysis to the landing test. And let us further plug in the wings, you can also target the different needs of users, and further provide them with customized Web site content display or functional services.

From the big point, no matter what analytical means, no matter what kind of business optimization proposition, if follow this analysis optimization path, can be more effective, can also really make optimization possible. And the pleasures that are experienced in this process are bound to weaken or even eliminate the pain of thought and distress in this process.

Unfortunately, to this day, I still find that many many enterprises, many of my clients, they may still focus on the analysis itself, but ignore other equally important parts, so that the original can improve performance may be halfway. On the other hand, in this process, the Executive force is also extremely important, I also see a lot of examples, although the intention, but no (execution) force, and finally wool, the end, empty leave regret.

Method is only method, no attempt, no development, no innovation, in fact, the method is nothing. In this sentence, it is also appropriate to replace the two subjects with the word "analysis".

This article copyright return "the website Analysis in China CWA" and its author sing, wants to reprint, please contact the author

Finally, I hope all like this blog and this article friends, can enthusiastically leave a message for me to write these words into some confidence and happiness. Thank you!

Original: http://www.chinawebanalytics.cn/kpi_optimization_part3/