Reprinted Source: https://blog.csdn.net/qq_35495339/article/details/96203751
In the mobile Internet company, no one should be planned in advance APP data system, it allows online operations, with the data it can be more scientific and more effort to operate. Today, for us to say how to do data analysis of APP.
First, why do data analysis APP
1. Analysis of operating data frame structures
APP is a construction and operations division of work is usually implemented by more than one role, due to the different everyone's focus, focus only on one aspect of data like piece of a jigsaw, unable to understand the operations of the product, the analysis can not be made effective recommendations. Therefore, only operators to build comprehensive data analysis framework, in order to fully measure mobile applications service providers. In addition, a complete data analysis framework also allows the operator product managers and developers to not only know the basic status and usage of product operations, a better understanding of who the user in the end, the depth found in the user's needs.
For example, for a new subscriber growth for mobile applications companies do, everyone will focus on the product, how many users are active users, etc., because these are related to product development are closely related; and investors will be more concerned about the product's user retention rate in order to determine whether to see the development of health products, to evaluate investment value; the same time, the boss is more concerned about how many users pay a total amount of revenue contribution and so on. So we need to build operational data analysis framework.
2. Use data to drive product and marketing iteration
Operational framework based data analysis will be a good show for the overall development of the company's products, but entrepreneurs will be part of more attention to detail.
For example, who use this product? Whether users prefer? How to use? What are features? The user channels which bring higher quality ... We can use the data to answer these questions.
Product designers can be targeted for product usage data analysis to understand the user's use of different functions, behavioral characteristics and use of feedback, which can provide a good direction for product improvement.
Marketers should not focus only on "what channel brings the number of users" should be concerned about is what brings a higher quality of some user channels.
3. Product profit promoter
Profit is the ultimate goal of the company, regardless of whether a product has explored a mature business model, the entrepreneur should make use of all the data to make profitable products have a better process. In the product business of the road, the data can help companies do two things: ① find the critical path of product profitability ; ② optimizing existing profit model .
Two, APP indicators should be concerned about what data
APP Data System is divided into five dimensions, including user scale and quality, participation analysis, channel analysis, functional analysis and user attribute analysis.
1. User scale and quality
User scale and quality of the analysis is the most important dimension of APP, its index is relatively most other dimensions, data analysts to focus on indicators of this dimension.
(1) active user metrics
User start over active user of APP in a certain period, in addition, we can be active within a user-defined statistical cycle consumer products over the core functions of the operation. Active users application index is a measure of the size of the user, usually, a product is successful, If you look at an indicator, then this must be the index of active users.
Number of active users can be divided according to different statistical period daily active (DAU), weeks of active (WAU), monthly active (MAU). Most users will want to open every day applications such as news APP, social APP, APP and other music, KPI assessment indicators of its products are daily active users. But for some low-APP consumer demand such as tourism, wedding photography, you may be concerned monthly active, even in the number of active longer period.
(2) new user metrics
New users are those after installing the application, users first launch the application. According to statistics time span divided into daily, weekly, monthly new users.
New users is a measure of the amount of the main indicators of the most basic indicators of the effect of channel marketing; new user accounts for the proportion of active users, can also be used to measure the health of the product. If a product is too high proportion of new users, it means the product is active promotion come by. This is very interesting, especially concerned about the retention of the user.
(3) the user configuration index
The user is composing weeks or active users monthly active users constitute analyzed to help understand the structure of the new and old customers health of active users. Weekly active users, for example, include the following types of users:
** This week reflux user: ** last week had not started the application, the user starts the application this week;
** n consecutive weeks of active users: ** n consecutive weeks, start over at least once weekly active users application;
** loyal users: ** 5 weeks of continuous active users and above;
Continuous active users **: ** consecutive 2 weeks or more active users;
** Loss of recent users: ** n consecutive weeks (one week equals the large, but less than or equal to 4 weeks) had no user starts the application.
(4) user retention index
User retention rate is the number of new users within a certain period of statistical re-start after a period of time is still the proportion of users of the application. User Retention can focus on the next day, on the 7th, 14th and 30th retention. The next day retention period that is a statistical proportion of new users launch the application again the next day; on the 7th retention rate that is the ratio of the number of new users launch the application again on day 7 of a statistical period; 14 and 30 day retention and so on.
User Retention verify that the product is attractive to users very important indicator. Users can often take advantage retention rate compared with competing products, measure application attractive to users. For one relatively mature version of the application, if a user retention rate of significant change, then the user has a significant change in quality, probably because of the change in mass marketing channels caused.
(5) days for each active user total index
Each user of the total number of active days index (TAD, Total Active Days per User) is the number of active days in the cycle statistics, the average user in the application. If the statistics are relatively long period, such as statistical period more than one year, then the total number of active days each user can basically reflect the number of days before the user takes the loss in the APP, which is a reflection of the quality or user stickiness, especially very user activity important indicator.
2. Participation analysis
Participation analysis is to analyze the user's activity, including the number of start analysis, the use of long analysis, interval analysis to access the page analysis and use of time.
(1) Starts index
The number of starts is the number of times a user starts the application within a certain statistical period. During data analysis, on the one hand to pay attention to the trend of the total number of starts, on the other hand, you need to focus on the number of starts per capita, when the ratio of the number of starts and the number of active users that is the same statistical period, usually the number of times per start and per capita use long It can be analyzed together.
(2) long using
When using the total length refers to the total duration of all APP from start to finish for use within a certain period of Statistics. When using a long average may also be used from a long, single-use and long when the angle analysis.
Per capita use long
= Use statistics in the same period the total length / number of active users
When using a single long
= Statistical cycles using the same total duration / number of starts
Long-related index is a measure of activity when using the product, an important indicator of product quality. User time of day is limited and valuable, if the user is willing to invest in your product more time to prove that your application is very important to users, such as the now very popular micro letters and other social applications.
(3) to access the page
Access refers to the number of pages the user first starts the number of pages visited. We usually want to analyze the distribution of the number of page visits, or statistical number of active users accessing a certain period (such as 1 day, 7 days or 30 days) page application distribution, such as the number of active users access pages 1-2, 3-5 number of active users page, page 6-9 active users at the same time ... we can cycle through different statistical (but the same span statistics, as of 7 days and more) access to the differences in the distribution of the page in order to find the problem the user experience.
(4) use time interval
Use the time interval is the same user adjacent two starts time interval. Applications within the general user statistics month interval number of active user distribution, such as the time interval within 1 day, 1 day, 2 days ...... At the same time, we can cycle through different statistical (but statistically the same span, as are 30 differences days) of the distribution of time intervals in order to find the problem the user experience.
3. Channel Analysis
Channel analysis is to analyze the various channels in the same investment, the number of users changes and trends in order to evaluate the scientific quality channels to optimize channel marketing strategy. Channel analysis includes new users, active users, the number of starts, and long retention index and other single-use. APP marketing channel mainly for Android and iOS.
Andrews channels: ① market third-party applications, such as Huawei, oppon, millet, 91 aides; ② ad networks, such as Network Alliance, Friends of the Union and so on; ③ manufacturers pre-installed, such as Huawei, millet, vivo and so on; ④ parallel brush machine, brush such as elves and so on; ⑤ social promotion, such as in the community do share, the formation of secondary or even several times the spread in the community, can do to promote, but this is not very good analysis of the data acquisition.
For Andrews, the user points or more to several sources, each of which may respectively to define. Different types of promotion, can do analysis of data from different dimensions. For example, as a third-party application market, many users are downloading APP through this channel, so more data in this area is to look active and retention; ad network like this, is analyzed by the integral wall, more the user completes the task by the order to do the analysis.
The main channels for iOS AppStore is, in principle, all of our data and activation are to get through this channel, but in actual promotion, we want more analysis is carried out through what channels the user to jump to downloads on AppStore , activate the product. This requires us to do the most direct and the underlying technology of the docking interface, docking --API. Specific analytical methods and Andrews are similar, mainly active analysis and retained data.
The above-mentioned preliminary assessment of the quality of channel dimensions, if further studies are needed channels, especially on the level of anti-cheating channels, indicators still need more, comprising: determining whether the user's normal behavior indicators, such as the amount of active critical operations the proportion of the total active, the user activates the APP time is normal; determining whether the real user equipment, such as type, operating system, analysis of the degree of concentration.
4. Functional Analysis
Functional Analysis analysis functions mainly active case, page views and conversion paths.
(1) Function active index
The main concern of a number of active functions, new users, user configuration, user retention. Similar to the definition of these indicators and the previously mentioned "user scale and quality" indicators. But, in this section focus only on a functional module, rather than the overall APP.
(2) page access path analysis
The main users of statistics is to leave open the application from the application of the whole process and the jump page views every step of the case. App aims to reach business goals that guide the user to complete more tasks in different modules of App efficient, and ultimately to promote user fees.
APP page access path analysis needs to consider the following three aspects APP user: ① identity : the user may be your members or potential members, there may be your colleagues or competitors, and so on; ② goal : to use APP purpose of different users different ; ③ access path : even if a similar capacity, using a similar purpose, but is also likely to be different access path. Therefore, we are doing when APP page access path analysis, segmentation needs to be done to APP user, and then the APP page access path analysis.
The most common method is subdivided in accordance with the intended use APP to conduct user classification. Such as automotive APP users can be subdivided into focus type, type of intent to purchase the user type, and each type of user different access path analysis tasks, such as type of user intent, he compares what are the paths of different models, What is the problem; another method is to use the algorithm, cluster analysis based on the user all access paths, users are classified according to similarity of access paths, and then analyzed for each type of user.
(3) Conversion
Conversion rate is the number of (or page views) and the ratio of the number (or page views) of the current page into the next page. Typically used to funnel model, it can analyze conversion products in the critical path to determine the flow of the product design, user experience issues.
For example, a user from entering the electric business website - see the products - the products into your cart - the payment is completed, every aspect has a lot of churn. By analyzing conversion rate, we can more quickly locate a different path using the product users, to analyze whether there is a problem, and suggest how to optimize improvements, usually we only need a day on the conversion rate of the continuity of monitoring can be.
5. Analysis of user attributes and portrait
User attribute analysis mainly analyzes the terminal equipment used by the user, and network operators, geography and user portrait angle. Dimensional analysis of terminal equipment have model analysis, analysis and resolution of operating systems analysis; alone and network operators to analyze users are networking and telecom operators, mainly from different provinces in the region and the country to analyze.
Image analysis features include user demographic analysis, analysis of the user's personal interests, business user interest analysis. Demographic characteristics, including gender, age, education, income, expenditure, occupation, behavior; personal interest refers to the analysis of the personal life interests, such as listening to music, watching movies, fitness, pets and so on; refers to users of commercial real estate interest, analysis of interest in the automotive, financial and other consumption. Data users need to be part of this portrait, portraits related to data collection, a more detailed analysis of the portrait can support. Before the junior partner interested can view the article " dry goods | Data Analysis user portrait practice and methods "
6. Revenue Analysis
Profit is the ultimate purpose of the product, so the total revenue, the number of paid users, pay rate, ARPU four indicators often used. Total income, number of paid users reflect the revenue and subscribers of scale; pay rate, ARPU represents the quality of user pay, user fees reflects the breadth and depth. We focus on transforming the order number and amount of the last part of the funnel.
Third, how to set up the APP Data System
In many products, many indicators mentioned above basically can not see, eventually leading data analyst because no data can not be analyzed. Mainly because there is no data on the development of statistics before the product line.
Usually this part of the work is mainly done by the product manager, but the data analysts in advance communication and coordination with product managers, planned indicator system data they need to drive product development and related data collection and follow-up of operations, dynamic optimization and rich data system. We start to understand a concept of "Buried."
1. What is Buried
Buried macro goal is to obtain data on the overall index to verify that the product is smooth business logic, some basic assumptions before is established? This time involving need to verify the data may include: product direction, market operation and business logic of the three aspects. Buried in fact, is itself the product of a visual health check, through logic and data, throughout the life cycle of a product, the product gradually achieve the best state. To give guidance for the future direction of optimization products.
Of course, different target buried point, final data validation results will be different. Such as user behavior on the line and the new version features performance data verification (several scenes): ① new features and whether the user's approval? The new version adds new features, how user click-through rate? ② user is smooth on the core using the path? Because there is no interactive experience design and function button leads to an increase in invalid clicks? ③ carried out for a particular date within the ad banner to promote products or promotions, how effective the campaign operations?
2. Specific steps
(1) understand the product form
Refers to the operation of the logic of the entire product, is concerned that the role of the user, and information channels, as well as transfer the relationship between them is what, like a blueprint and framework products.
(2) understand the business logic
Refers to the implementation of a business, user roles need path traveled, what role involved, what function modules (or subsystems) involved, the correlation between the flow between the modules, is what kind of data .
(3) operational flowchart
Is based on the business logic of functional decomposition down, such as business logic recorded songs to sing it, there will be a module that generates a score, this module that generates a score, her specific business processes is what kind of, what the details process, abnormal process, tips and so on.
(4) the node of the service code of
This step is mainly to add statistics and statistical parameters important event nodes (nodes need to be calculated) listed.
(5) the delivery of development adjustment DRD
Can communication and coordination with product managers and developers, and deliver the data needed for the index system.
(6) Data analysis
Late database with statistics corresponding node, then you can bring analyzed.
Fourth, the commonly used analysis method
1. Life Cycle Analysis
Product Life Cycle (PLC model) was proposed by US economist Raymond Vernon, that is a new product from development to market to the whole process out of the market. Product life cycle can be divided into start-up, growth, maturity, decline, at every stage of the product, the weight data analysis work and focus of the analysis differ.
(1) start-up period
Start-up focus is to verify the core value of the product, or the product of hypothesis testing: you can solve a problem by some people for specific products or services. Critical data then we need to focus on the target population portraits and retention.
** target population portrait: ** start-up portrait can be understood by the user community early access to some third-party application monitoring SDK, from side to verify that the target user group characteristics consistent with the hypothesis user groups, it is common demographic attribute (gender, age, education, region).
Retention rate **: ** when the current user meets the characteristics of the target audience, the core concern retention of these users, long / frequency, user stickiness and other indicators when used for very many (7, biweekly dimension of retention, on the 30th, etc.), based on product features to select, if the product itself meet the needs of a small minority low frequency, retention are advised to select bi-weekly or even on the 30th; retention rate, on behalf of the user on the value of the product recognized and dependence, to general , suppose will be able to be verified, usually less than 20% retention is a more dangerous signal.
(2) rapid growth
After the initial stage of product polishing, the product has a better retention rate, and this time the product entered the spontaneous growth. Spontaneous growth period can be the focus of attention in the management of the entire life cycle of a user, wherein the new subscriber growth, activation, triggered "Aha moments" to the entire funnel analysis based active users.
(3) maturity
With the rapid growth of users, and constantly improve the product, before and after the product matured, the focus of attention data analysts start from the first half of the user lifecycle (attract, activate, retained) back half (loss reflux) start offset , while concerned about the commercial conversion path.
** ① Loss and Return: ** loss of attention during reflow, the data will reveal a change of the current user of the dish and detailed analysis of the causes of loss can refer to the process below:
That is the core idea, a return visit by + qualitative data validation as the main means to determine the cause of the loss, changes in product operating strategies to prevent loss of customers or users back to promote reflux. In addition, for some stability of delivery channels, improve the general approach may enhance the conversion limited at this time may be a more refined analysis of the channels to optimize the upgrade ROI.
** ② Business Analysis of Conversion Rate: ** mature operators need to focus on high quality for the user, by means of low-quality products and improve user operations, to make high-quality user migration. A combination of different product forms and business models, general data analysis core indicators comprising: the product user per the number of days (weeks, months units observed); length (in days was observed when the product user per use, average duration higher the user-dependent, the stronger, the greater commercialization of space); average purchase price (in months to see a user's purchases, the higher purchase price of electronic business platform meant to be a high net users, operators need to focus on ); the number of purchases per capita (in months, high-frequency, low-priced high-quality user is also a user)
(4) recession
Ultimately, the product entered a recession, not repeat them here.
2. Conversion funnel analysis
Funnel model is a method of data analysis used more often, which is suitable mainly for the scene after a series of user action to complete the task, and the task need to monitor and analyze the finished effect, and every step possible problems.
The core idea of the funnel model, is to start from the ultimate goal, to find the user every step of the conversion or wastage, accompanied every step of the conversion rate, or churn metrics to monitor performance, and ultimately enhance conversion rates, or reduce churn by rate in order to optimize the final index and deliver business value.
When performing the actual analysis of the funnel model, a combination of different types of products and business scenarios, the funnel can be roughly divided into the following model:
① user acquisition model : AARRR start from the entire user lifecycle, including user acquisition Acquisition, Activation user conversion, Retention user retention and activity, Revenue user-generated income, to initiate propagation Refer.
** ② consumption funnel model: ** Consumer funnel generally used for page structure and content of the more complex operations, from the perspective of the consumer and flow to user content, macro-level consumption for the user to answer what, for the micro-level impact analysis What is the problem user consumption yes.
** ③ electricity supplier funnel model: ** user to purchase merchandise belonging to decision-making behavior, to pay for goods from the browser transformed every aspect of the order.
④ function optimization funnel model : funnel analysis also applies to the product features its own optimization, starting from the ultimate goal, to split business links, extraction and optimization of core indicators, so as to enhance the overall function of the conversion rate.
3.AARRR model
AARRR model is an analytical framework applicable to mobile App, also known pirate index, is the core model "growth hacker" in driving subscriber growth. AARRR model the user behavior indicators are divided into five categories, namely: get the user to stimulate active, improve retention, increase revenue and transmission of the virus.
From the user gets to the spread of the virus, each link has important indicator that we need to focus on, through the systematic dismantling of five major categories AARRR model user behavior, it allows us to more clearly know every aspect of the need to focus on priorities index.
Electricity supplier to business, for example, the following figure based on AARRR model, to build a full user lifecycle operational context and each node needs the focus indicators:
(1) Acquisition get the user
Acquiring user stage, we hope to get more attention to the potential users of our products, our exposure to the promotion page by basic ways: ① get paid: media advertising, SMS, EDM, flow trading / replacement; ② Search Marketing: Search engine optimization (SEO), Search engine marketing (SEM); ③ word of mouth: event invitations between users, H5 virus dissemination.
After the user visits a page, you can navigate, active search, the algorithm is recommended to learn about our products. Hit the current needs of users will register behavior, and be the first meeting with a real sense of the user. Then we should focus on the promotion page UV, an important indicator of click-through rate, the amount of registration, enrollment, user acquisition costs.
(2) Activation stimulate active
Is there a registered user to learn more about our products? Which involves product functionality, design, copywriting, incentives, credible, and so on. We need to constantly tune, guide the user to the next step behavior, allowing new users to become a long-term active users:
We can interface / copy optimization, beginners guide, offers incentives and other means, a user activation process of transformation upgrade. Monitoring browse product pages, add to cart, submit orders, the order to complete the conversion funnel.
This process, we want to focus on activity, if the definition Add to Cart to active users, then they would have to observe the registration ADD TO CART conversion funnel, split by dimension, a common feature of high-quality analysis of the conversion funnel / operational strategies to enhance policy coverage, optimizing the overall conversion effect.
(3) Retention improve retention
After the user completes the initial purchase process, we will continue to be used? The loss of customers can continue to come back to use our products?
Lack of product viscosity can lead to rapid loss of customers, we can build a life cycle node marketing plan, through push, SMS, subscription number, e-mail, customer follow-up and all the appropriate way to remind users continue to use our products. And on this basis by integration / hierarchy, build loyalty, improve user stickiness.
Focusing on metrics retention, re-purchase rate, the number of purchases per capita, the recall rate.
(4) Revenue increase revenue
We get an average cost per user how much money? Average per user can contribute much value to us, to whether the user's behavior, even in other ways to make money?
Basic electricity supplier business to focus on user acquisition costs CAC, customer lifetime value, based on this incentive by operating activities users to purchase, improve user price, frequency, frequency, and ultimately enhance the user lifecycle value contribution.
Focus on user acquisition costs, customer lifetime value, marketing ROI and other indicators.
(5) Referral virus spread
Whether users will spontaneously promote our products? By incentive whether to allow more loyal customers to promote our products?
In today's highly developed social networks, we can go to the product spread through a variety of new ways: the invited users, community operating in the field of vertical, H5 old with the new marketing communications activities to keep old customers to promote our products to attract more many potential users.
Focus on the number of people invited to initiate new subscribers each cycle of transmission of the virus, inviting conversion rate, propagation coefficient.
V. Summary
We already know that the intention is not to analyze the data the data itself, but to build a data feedback loop.
In doing data services into one App, we hope that through the design basis data indicators, multi-dimensional cross-analysis of different indicators to data screening question, then reverse the role of the product, eventually forming a closed-loop data-driven product design.
In fact, App data analysis is not so sacred, commonly used data indicators are not difficult to master. The key is to design data indicators based on two facts: ① business models and business background; ② data analysis motives and purposes.
Doing business index or regular analysis of data, as long as regular attention to those indicators can affect the company's core business, in order to do business quickly determine performance, can greatly improve efficiency, quickly identify problems.
Therefore, when the data for the app to do the service, as long as starting from the point of view on it.
Sixth, tools and books recommended
1. Tools
(1) vertical areas of statistical tools platform: ios: seven wheat data, Zen master, appduu; Andrews: Cool Biography
(2) third-party platform statistics: League of Friends, TakingData, AppAnnie, Tencent cloud analysis, Baidu mobile statistics, Shence data, Zhuge IO, GrowingIO
2. Books
"Growth hackers" "Lean Data Analysis", "easy to understand data analysis," "Web site analysis of actual combat," "secret -Google Analytics website traffic analysis and optimization techniques."
