Shulou(Shulou.com)06/02 Report--

About the author: Xie Rongsheng, founder of Digital Geek & CEO, former Director of Financial products of Gome and Senior Product Manager of Taobao.

The author has worked in the Internet industry in China for 16 years, has gone through many stages of Internet development, and has been in charge of products, operation, marketing and so on. I find that in recent years, due to the surge in traffic costs and competitive pressure, Internet companies rely more and more on data analysis, but they are troubled about how to do data analysis well.

How to make enterprises make better use of data analysis? As a result, several geeks have served more than 300 enterprise customers and accumulated more than 5000 trial users in the past two years. I would like to share with you some experience in the application of user behavior analysis.

In recent years, we have frequently seen such words as enterprise fine operation, data-driven growth and growth in all kinds of media, and the core behind this is data analysis. but many people do not understand the causal relationship between user behavior and business growth, just because other enterprises have been successful through user behavior analysis, so they learn from the trend, but only scratching the surface. This leads to the following extremes at the application level:

1)。 Purchased a user behavior analysis system, which is idle due to the lack of data analysis methods.

2)。 Underestimated the difficulty of user behavior analysis, because of insisting on internal self-building, the business department has been waiting for the available system, wasting a lot of resources and development opportunities.

3)。 Do not understand the value of user behavior analysis, only pay attention to the conventional PV, UV indicators.

How to solve the above dilemma, let's start with understanding the analysis of user behavior.

1. What is user behavior analysis?

User behavior can be summarized with 5W2H:

Who (who), What (what behavior), When (when), Where (where), Why (what purpose), How (in what way), How much (how long it took and how much money it cost).

User behavior analysis is to find the rules of users' use of products through statistics and analysis of these data, and combine these laws with the marketing strategy, product function and operation strategy of the website. find that there may be problems in marketing, products and operations, and solve these problems can optimize user experience, achieve more precise and accurate operation and marketing, and make products get better growth.

Second, why do you need user behavior analysis?

In the PC Internet era, the annual growth rate of netizens reached 50%, and you can get a lot of traffic by building a website. In the early days of the mobile Internet, APP also experienced a wave of traffic dividends, which cost less than 1 yuan to get a customer. In recent years, as the dividend of traffic growth recedes, the competition becomes more and more fierce. there are hundreds of peer competition in each field, the cost of getting customers has soared to an unbearable level, and business growth is getting slower and slower or even retrogressive.

Photo: competition in the Internet industry is becoming more and more fierce

In such a high-cost and highly competitive environment, if the enterprise can not make use of data analysis to do a good job in fine operation, it will lead to a huge waste of resources, which will inevitably lead to high operating costs and lack of competitiveness. For the Internet platform, the traditional data analysis mainly focuses on the data of the results, but lacks the analysis of the user behavior process that produces the results, so the value of data analysis is relatively limited. This is also the reason why many enterprises feel that they have done sufficient data analysis in recent years, but they do not have much effect. By analyzing the 5W2H of user behavior, we can know where users come from, what operations they have done, why they are lost, where they are lost and so on. In order to enhance the user experience, the conversion rate of the platform, with fine operation to enable enterprises to achieve business growth.

Third, how to collect user behavior data?

User behavior analysis is so important, why are there so few Internet companies that can do a good job of user behavior analysis? The main reason is that the data collection is not comprehensive and the analysis model is not perfect.

1. How to efficiently collect user behavior data

Because the traditional data analysis is not precise enough and the analysis model is not perfect, the analysis is too extensive, and the application value of the analysis results is low. And if we want to do a good job in the analysis, we must first have rich data, so from the data collection, the traditional user behavior data collection method is relatively inefficient, for example: when we obtain a user's behavior data, we need to add the monitoring code to the corresponding button, link, or page to know how many people have clicked this button and clicked this page. This method is called "burying point", which requires a lot of manpower, energy and tedious process, which leads to the high cost of manpower and material resources.

In the mobile Internet era, burying sites has become a more painful job, because it needs to be posted to the App Store after each burying point, and the review cycle of the Apple App Store is hard, which makes the timeliness of data acquisition even less timely. As data analysis is an extremely important part of business development, even if the cost of manpower and material resources is too high, this work still can not be omitted.

Therefore, we can also see that there are some excellent user behavior analysis tools at home and abroad to achieve the function of non-buried point collection, for example: there is Mixpanel abroad, and domestic geeks can collect data without buried point in WEB, H5, Android and iOS. Through the collection of non-buried points, the perfection and timeliness of the data can be greatly enhanced.

two。 How to accurately collect user behavior data

For some core business data, we want to ensure 100% accuracy, so it can also be supplemented by back-end embedding points, so that we can not only experience the efficiency and convenience brought by no burying points, but also ensure the accuracy of core business data. In data collection, geek supports data integration in four ways: no burial point, front-end burial point, back-end burial point and several geek BI import data.

Fourth, how to do a good job in user behavior analysis?

First of all, it is necessary to clearly define the business objectives, deeply understand the business process, identify the key data nodes that need to be monitored according to the objectives, do a good job in the collection and collation of basic data, and have a scientific model when there is enough data. in order to more effectively support the analysis results.

The previous generation of user behavior analysis (more accurately, website statistics or APP statistics) tools, the main function is limited to the analysis of browsing behavior, but there is no analysis of users'in-depth interaction behavior, so the value of analysis is relatively limited. At present, the impression of most Internet employees on user behavior analysis is still at this stage.

In my opinion, in order to do a good job in user behavior analysis, we should master the following analysis models:

1. Full tracking of user behavior, supporting AARRR model

Dave McClure, a 500 Startups investor, has proposed a set of analytical models to analyze the "piracy indicators" obtained by users at different stages, which has been widely used in Silicon Valley.

AARRR is the abbreviation of Acquisition, Activation, Retention, Revenue and Refer, which corresponds to five important links in the user life cycle. First of all, the user behavior analysis should be done based on the user's complete life cycle.

1)。 Get users

In marketing promotion, which channel brings the highest flow and what is the ROI of the channel? The conversion rate of different advertising content is the data analyzed in this step.

The source channel is the first step to get customers. Through the combination of system automatic identification and custom channel, the retention and transformation effect of each source channel is analyzed. The access source of the website, the download channel of App, and the search keywords of each search engine can be counted and analyzed conveniently through the data analysis platform. Using the multi-dimensional analysis of UTM promotion parameters, cross-analysis through promotion channels, event names, display media, advertising content, keywords and landing pages, we can distinguish between high-quality channels and low-quality channels, fine tracking, and improve channel ROI.

Through the channel quality model, formulate the corresponding customer promotion strategy:

Figure: Channel quality model

The channel shown in the above figure is an example, and the channel quality will change dynamically. In the first quadrant, the channel quality is high and the flow is large, so we should continue to maintain the channel delivery strategy and intensity; the second quadrant channel has higher quality but less flow. The channel quality should be increased and continuous attention should be paid to the change of channel quality; in the third quadrant, the channel quality is poor and the flow is small, so the channel should be carefully adjusted and optimized step by step; the channel quality in the fourth quadrant is relatively poor, but the flow is large, so we should analyze the channel data to put in more accurately and improve the channel quality.

2)。 Activate the user

Activating users is the most critical first step in achieving business goals. If there are a large number of users using your product every day, but without a strong relationship with you, you will not be able to carry out follow-up operations.

3)。 User retention

Nowadays, the key to the success of a product is not the viral mechanism or a large amount of marketing money, but the user retention rate. It is very important to develop products that attract users back. There is a "40-20-10" retention rule for Facebook platform. The numbers represent the daily retention rate, weekly retention rate and monthly retention rate. If you want the DAU of the product to exceed 1 million, then the daily retention rate should be greater than 40%, and the weekly retention rate and monthly retention rate should be greater than 20% and 10%, respectively.

Retention is one of the important links in the AARRR model. Only when retention is done, can we ensure that new users will not be lost in vain after registration. This is like a constantly leaking basket, if you do not repair the cracks below, but just pour water into it, it is very difficult to achieve sustained growth.

4)。 Get income

Achieving revenue is fundamental to the survival of every platform, so it is important to find a business model that suits you. According to different business models, the ways to get revenue are also different: media platforms rely on advertising, games rely on users to pay, e-commerce by collecting commissions or sellers pay, etc., while in the field of enterprise services, LTV: CAC is greater than 3 in order to grow effectively.

5)。 Virus transmission

Through the optimization analysis of the first four stages of the model, unstable users, active users and then the final loyal users will be retained and transformed to cultivate loyal users of the enterprise. through social word-of-mouth communication can bring efficient benefits to the enterprise.

Today, with the high cost of getting customers, social communication can bring enterprises a higher quality user base and lower customer acquisition costs.

two。 Transformation analysis model

Conversion rate is the core of continuous operation, so I also use a lot of space to interpret it in detail. The commonly used tool for transformation analysis is the conversion funnel, or funnel for short. New users continue to lose in the registration process, eventually forming a funnel-like shape. In the process of analyzing user behavior data, we not only look at the final conversion rate, but also care about the conversion rate of each step of the transformation.

1)。 How to construct funnel scientifically

In the past, we will build the funnel through the experience of products and operations, but we have no confidence in whether the funnel is representative and how effective it is to improve the overall conversion rate. at this time, we can understand the mainstream path of users through user flow analysis.

Figure: user flow Analysis

User flow analysis is very intuitive, but analysts need to have some experience and judgment ability. In order to solve this problem, digital geek has developed an intelligent path analysis function, which only needs to select the transformation target and analyze the mainstream path of user transformation with one click. The efficiency of creating a funnel is reduced to a few seconds.

Figure: intelligent Transformation Analysis

2)。 Funnel contrast analysis method

It is not enough to use ordinary funnel in the transformation analysis, it is necessary to analyze the detailed factors that affect the transformation, and it is very important to carry out subdivision and comparative analysis. For example, by comparing the source channels of users, you can grasp the conversion differences of different channels to optimize channels, while by comparing user equipment, you can understand the conversion differences of users of different devices (for example, for a higher-priced product, the conversion rate from placing an order to payment is significantly higher for iphone users than android users).

Figure: funnel comparative analysis

3)。 Combined analysis method of funnel and user flow

In general, the conversion funnel only has the trunk process, and there is no detailed information about the inflow and outflow of each step. When we analyze the user registration conversion, if we can know where the user who has not been converted to the next step has gone, we can more effectively plan the transformation path of users. For example, in the conversion path in the figure below, 88% of the users who did not enter the second step left directly, while 10% of the registered users chose to log in directly, and only 2% of the users bypassed the landing page to go to the home page of the site. 100% of the users who did not switch from step 2 to step 3 left. This is a typical closed landing page, so you only need to optimize the conversion rate of the third step to improve the overall conversion rate.

4)。 Micro transformation behavior analysis method

Many behavior analysis products can only analyze the transformation at the functional level and event level, but there are serious deficiencies in the detailed analysis of user interaction. For example, in the funnel above, we analyze that the last step is the key to the transformation, but the last step is to register the form, so it is very important to analyze the detailed behavior of filling out the form, which we call micro-transformation.

For example, the length of time it takes to fill out the form, the loss of which field is filled out by the user who filled out but did not submit the form, the blank rate of the form field, and other form filling behaviors.

Figure: fill in the conversion funnel in the form

Figure: how long it takes to fill out the form

Through the micro-transformation analysis of the above form, the conversion rate from the beginning to the registration is 85%, while the flow to the completion is only 8%. It can be concluded that the biggest leak point affecting the transformation is the fill rate. Then how to improve the fill rate is the core of our registration conversion. Effective content and accurate channels are the core factors that affect filling in. We have already talked about channel factors in customer acquisition analysis, which leads to the fourth tool of our micro-transformation analysis: user attention analysis.

5)。 User attention analysis method

The user's interaction with the content of the page, such as clicking, browsing, staying on the page elements and scrolling the screen, all represent the user's attention to the information to be displayed by the product and whether it can attract the user's attention.

Business data can be visualized, so how can behavioral data be visualized? Several geeks transform the above behaviors into five kinds of heat maps: split-screen touch rate heat map, link click map, page click map, browse heat map and attention heat map. through the cross-analysis of the five kinds of heat maps, we can effectively analyze the content that users pay most attention to.

Figure: attention heat map

Only when we can master the interactive behavior analysis of micro-transformation, can we improve the conversion rate more effectively. All the analysis tools that can not effectively improve the conversion rate of the platform are a waste of human and time resources, which is also the fundamental reason why many enterprises do not benefit from user behavior analysis.

3. Fine operation model

In the past, the operation can only be aimed at all users, if you want to do accurate operation behavior for some target customers.

Figure: user group portrait

For example, when we want to carry out precision marketing for users who have registered with iphone in a certain area but are not active for three days or have not formed a transaction conversion, we need the cooperation of operators, product personnel and technicians to access data and formulate operating rules, which involves a lot of manpower and time investment. The new generation of user behavior analysis can use user clustering, user profile, custom user activity and retention behavior, and accurately locate users, so as to achieve fine operation.

Figure: creating user clusters

4. Qualitative analysis model

User experience is the top priority of an enterprise. It is necessary to master the real process of user experience in many aspects, such as product design, user research, research and development, operation, marketing, customer service and so on. However, how to optimize the user experience has always been a lot of internal controversy, the main reason is that it is difficult to describe specifically and vividly. When abnormal user behavior is revealed through behavior analysis, whether you can reproduce the specific scene when the user uses your product is very important to optimize the product experience.

In the past, when I was on Taobao, the user experience department would optimize the experience by inviting users to interview the company and do usability experiments, but this method requires a lot of time and cost, and the sample is not necessarily representative. In order to solve this problem, digital geek has developed a screencap tool for user behavior, which does not need to invite users to record in the company to save costs, and intuitively and efficiently restore the real operation of users in the form of video. so that all positions in the enterprise can grasp the first-hand information of user experience and help product research and development to improve user experience.

Figure: user behavior screenshot playback interface

Summary: analyze the whole user life cycle through the AAARRR model; improve the product conversion rate through the conversion analysis model; improve the operational effectiveness through refined operation; optimize the user experience through qualitative analysis; if the above four aspects are done well, you can certainly achieve business growth through user behavior analysis.

5. What is the future direction of user behavior analysis?

A lot of people ask me, why do you want to set up a few geeks when there are already several companies that do user behavior analysis? I think the goal of data analysis is to apply the analysis results to optimize business efficiency, but the main analysis tools at home and abroad only stay at the analysis level, and there is still a lot of room for efficient application. Therefore, several geeks should not only be more professional and effective at the analytical level, but also achieve new breakthroughs at the application level. The results of data analysis reflect two main types of problems: operations (including marketing) and products. Therefore, it is necessary to provide targeted solutions to these two types of problems.

1. Automation of operation

As we said earlier, fine operation can be achieved through user behavior analysis, but specific applications also require manual formulation of operational strategies, which can only be applied through products, R & D and development, and when the strategy changes, the corresponding tools need to be redeveloped. this also takes up a lot of time and affects operational efficiency. Digital Geek has developed automated operation tools, where operators directly set rules, and the system automatically pushes accurate activity information to qualified users according to the rules, directly improving the efficiency of operators. Operators can shift their focus to planning instead of wasting too much time in repeated implementation, and automated operation can save enterprises a lot of operating costs.

Figure: create automated operation rules

two。 Scientific decision-making in product and operation (marketing)

The analysis of user behavior data is often carried out after the behavior occurs, while the products and operations make decisions through experience and pat on the head. once the decision is wrong, it will cause irreparable results. Therefore, if the product and operation plan can be verified in a small range through the user shunt Amap B test before the product and operation scheme is launched, and choose the best scheme to release, it can greatly improve the scientific nature of decision-making.

Google generates 10 billion dollars in revenue for the company by running tens of thousands of tests to optimize its products and operations every year.

The method of Ahand B test is very effective, but it is not widely used by domestic Internet companies, which is mainly related to the complexity of applying Ahand B test.

Geek has a complete Aamp B test tool, and business staff can use visual test editing tools on the website and APP to create and run tests. Through automatic interpretation of test reports, the threshold for Amax B testing is greatly reduced.

Figure: website-side visual editing test tool

3. Automation of analysis

Data analysis has a certain degree of professionalism, not only need to master different analysis methods, but also be familiar with the business, combined with business in order to give valuable analysis results. If it can be like 360security guard, only need to load SDK, it can automatically diagnose and analyze, and give the solution, this is the future direction of data analysis, several geeks have also made positive attempts in this respect, and have achieved preliminary results. At present, it has the functions of automatic data early warning, automatic report and so on.

User behavior analysis is a science, and being good at obtaining, analyzing and applying data is the basic skill for everyone to do a good job. Every enterprise should strengthen the application of user behavior analysis big data and find out the rules from the data. use data to drive enterprise growth.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.