当前位置：网站首页>Graphic data analysis | business cognition and data exploration

Graphic data analysis | business cognition and data exploration

2022-06-12 02:01:00 【ShowMeAI】

ShowMeAI research center

author ： Han Xinzi @ShowMeAI
Tutorial address ：http://www.showmeai.tech/tutorials/33
This paper addresses ：http://www.showmeai.tech/article-detail/137
Statement ： copyright , For reprint, please contact the platform and the author and indicate the source

The core steps of data analysis are divided into ： Business cognition and data exploration 、 Data preprocessing 、 Business cognition and data exploration Wait for three core steps .
This article introduces the first step —— Business cognition and data exploration .

One 、 Common business objectives

（1） Descriptive analysis ： Analyze and describe the characteristics of the data

Descriptive analysis is a good way to deal with information aggregation , Combined with visual analysis , Can provide a comprehensive data structure , And show the conclusions through the dashboard . Through analysis in enterprises KPI To evaluate performance , Is one of the common applications of descriptive analysis .

（2） Predictive analysis ： Predict future results

Predictive analysis is a complex field , Need large-scale historical data , And with the help of technological progress （ Especially machine learning ） Build a high concurrency prediction model , To gain predictive insight into the future .

（3） Diagnostic analysis ： Diagnose practical problems through data

Through diagnostic analysis , You can think critically about the data , Judge the actual problems , And further optimize management or reduce losses . Logistics enterprises can reduce logistics delay by using diagnostic analysis , E-commerce enterprises can update their marketing strategies with the help of diagnostic analysis , Reduce cost and improve efficiency .

（4） Normative analysis ： The integrator of data analysis

Normative analysis combines all the above analysis techniques , It can help the company make relevant decisions on the basis of data conclusions . It should be noted that , The necessary basis for using canonical analysis is a large enough data scale 、 Quantity use AI technology , Therefore, it is often used in large Internet companies （ Such as Google） And financial institutions .

Two 、 Common business indicators

2.1 Vanity indicators VS Effective indicators

An indicator is an index to measure the quality of things . There are many data indicators , Polaris indicators , Two level index 、 Qualitative indicators 、 Quantitative index 、 Vanity indicators, etc . Good data indicators , Guidance should be provided for the development stage of the product business line , Some data have the opposite effect .

Vanity indicators （Vanity Metrics） Make people feel good , But it is superficial and even deceptive . for example , Only focus on the number of visits to the website （PV、UV）, But ignore the jump rate 、 When users browse data indicators such as time , It's easy to fall into the trap brought by vanity indicators .
Effective indicators （Clarity Metrics） It refers to those indicators that can really bring benefits . for example , Conversion rate 、 Retention 、 Proportion of daily active people, etc , Can better insight into the actual trend of products and user behavior .

2.2 Select the appropriate indicators

（1） Ratio indicators

The ratio indicator has 『 Comparative 』, So as to become the best data index . An example of driving , Mileage reveals distance information , And speed （ distance / Time ） To tell you it's accelerating , Or slow down .
The comparison object of the ratio index , It's different periods 『 Self - 』 contrast , Used to compare the growth trend of things ; Or the comparison between different objects in the same period , To compare the growth of different things .

Usually , User behavior analysis uses the following ratio indicators （ Or one of them ）：

Time related indicators ——『XXX Speed 』, for example , Growth rate of new users （ Number of new users per unit time ）,
Proportion related to quantity ——『XXX rate / Than 』, for example , More active users than （ Proportion of active users in total users ）.

（2） Multi index joint analysis

Put aside the cardinality , The meaning of ratio will be greatly reduced . quote 《 Benefit data analysis 》 Examples in ,“ When your product was first launched , Strictly speaking , Your father registered an account , It can also double your number of users ”.

Be careful ！ Although the ratio is the best data indicator , But it also needs to be supplemented by other qualitative and quantitative indicators . Indicators never exist alone , Instead, it should be comprehensively evaluated from multiple indicators . This requires understanding the coupling between indicators , Design index system , Reveal the hidden facts in the data from various angles .

2.3 Combined with application scenarios

When analyzing data and designing key indicators , It should be combined with the actual business , Restore to a specific scene . for example , When analyzing user behavior , The stickiness of users to services or products （Stickiness） Is an effective indicator . In different usage scenarios , There are also differences in the measurement of stickiness ：
Stickiness refers to when a customer purchases a product or service , The degree of willingness to buy again or recommend to others .

Jump off rate （Bounce Rate）： For content products , When visitors click on your web page , How many people will shut down immediately , What's the percentage .
Conversion rate （Conversion Rate）： When the product is promoted , The conversion rate is the ratio of the number of registrations to the number of visitors , Indicates how many visitors actually become registered users of the website .
Retention （Retention Rate）： For new users , It is used to test whether new users start using the product within a certain period of time , After a period of time , Proportion of users who continue to use the product .
Activity level It is also an effective measure . For social networking sites , Number of daily active users （Daily Active Users, DAU） Proportion is a key indicator .

2.4 Typical cases

Case study ： At a cloud product presentation （ Activities ） in , The organizer made a comprehensive demonstration and explanation of the major update of the product , I hope that through offline interaction and online live broadcast , To attract more users to use the product .

（1） Clear analysis objectives

If the goal is to evaluate the benefits of the activity , As a data analyst , What will be done ？

Ideas ： For this scene , You can use contemporaneous group analysis , According to whether participants are registered , Break down participants into unregistered people （ Potential users ） And registered personnel （ Old users ）.

After the event , Some unregistered users become new users . There may be a variety of incentives to attract new users to register , The most common is a free trial . Later, you can track the product usage behavior of users from different sources .
Old users are fans of products , It's paid users , Can bring benefits to the company . Paid users may lose , It is also possible to rely more on products 、 Use the product more frequently .
Be careful ！ When analyzing user behavior , Be aware of , Although some users can't bring direct benefits , But it can bring more users , So as to indirectly create income for the company .

（2） Design analysis index

Ideas ： Take human behavior as the starting point , from 『 Quantitative data 』 and 『 Qualitative data 』 Design indicators from two angles . In order to analyze the benefits of the activity , The following key indicators can be developed ：

[1] Participation （Engagement）： Measure how many people participated in the event , Used to evaluate the impact of activities

The total number of 、 Number of old users 、 Number of potential users .
After participating in the activity , How many users have registered （ Free or trial ）.

[2] Observe the behavior of new users ： The conversion rate can evaluate the benefit of the activity , Retention rate can evaluate the stickiness of new users to products
At the meeting 、 Unregistered personnel , Is a potential user of the product . Regardless of the use of 『 Free trial 』 still 『 Buy + Give 』 And other promotion methods , As long as the user registers , It can be considered to have completed a transformation .

Conversion rate of new users （Conversion Rate）： Proportion of new users registered after participating in the activity , Higher conversion rate , It shows that the more attractive the product is to users , The better the benefits of the activity .
The source of new users ： Analyze the source and transformation path of new users , It helps to determine which sources bring more effective transformation .
Conversion path for new users （ conversion funnel ）： Free trial （Free Trial） New users registered , The data indicators of the transformation path are 『 New trial （New Trial）』『 Active the next day （Activated Trial）』『7 The day is active （Activated Trial in 7 days）』.
Retention rate of new users ： In a continuous billing cycle , Proportion of new users still active in the same period .

[3] Observe the behavior of old users ： Loyalty and usage of old users

Regular users refer to before participating in the event , Registered users , These users are loyal fans of the product . Analyze the behavior of old users , Can also evaluate the effectiveness of activities .

Loyalty refers to how often users use products , And the degree of dependence on products and services . If this activity encourages regular users to use the product more frequently （ That is, the consumption of old users increases ）, It shows that holding this event has improved the loyalty of old users , Bring about an increase in benefits .

The consumption of old users has increased ： After participating in the activity , The consumption of old users is compared with that of the previous payment cycle , Increase or decrease .
Proportion of users with increased consumption ： After participating in the activity , Proportion of users with increased consumption .

Consider some special circumstances , How many old users have been recalled , How many old users are lost ：

Zombie account activation （New billed Customers）： How many old users left , Re pay for products or services .
Loss of paying users （Churned Customers）： Users in a payment cycle , No longer pay for products or services .

[4] Observe the dosage of the product ： Service type

Amount of product , It is also an effective index to evaluate the benefits of activities . The behavior of users using products , In fact, the services provided in consumer products . increase 『 Service type 』 This dimension , You can assess whether the increase in consumption is related to this activity —— Compare the total dosage 、 Dosage related to the theme of the activity .

If the consumption related to the theme increases greatly , The growth of total consumption is relatively flat , So it shows that this activity has promoted the growth of consumption .
If the consumption related to the theme increases slowly , The growth of total consumption is also flat , So it means that this activity has no effect on the increase of consumption .

Be careful ！ The comparison should be comparable . In the processing of comparison data , The impact of new users on data should be eliminated . Take the time of the event as the dividing point , Only compared with old users N（1-3） Within and after a payment cycle N（1-3） Usage in a payment cycle . In order to calculate the dosage more accurately , The influence of individual factors should be avoided , have access to 2-3 The average consumption in a payment cycle .

Analysis of consumption growth of old users ：

Proportion of old users with increased consumption
After participating in the activity , How much does the consumption of old users increase , What is the proportion of growth
Average usage per user

Consumption growth analysis related to the activity theme ：

Proportion of old users with increased consumption （ Activities related ）
After participating in the activity , How much does the consumption of old users increase , What is the proportion of growth （ Activities related ）
Average usage per user （ Activities related ）

（3） Adjust as needed

it is to be noted that , Data analysis should be combined with business , According to the main problems faced , Set key data indicators , To answer the problems encountered in decision-making . for example ：

When the product is not well known , The number of visits to the website can be used as a key indicator .
When you have a large user base , You can count the number of registered people （ Or number of probationers ） As a key indicator .
When the number of registrants increases to a certain extent , The conversion rate can be （ namely , Convert free users into paid users ） As a key indicator .

2.5 Common data indicators for website analysis

Attach the data indicators commonly used in website analysis , Can you identify which indicators of vanity ？

（1） Website traffic indicators

Page view The amount (Page View,PV）： Visitors are recorded every time they open a page .
Number of unique visitors （Unique Visitors,UV）： The number of unique visitors to the website in a day .
Number of repeat visitors （Repeat Visitors,RV）： In a day , Repeat visitors to the site .
Page views visited by visitors （Page Views per User）： The average number of pages visited by each visitor , If the index is high , It indicates that the user is highly viscous , in other words , Visitors show interest in the website 、 Willing to stay for a long time .

（2） User behavior indicators

Jump out rate ： It means that users only browse one page and leave the website , The jump out rate shows the degree of interest of visitors to the website ： The lower the bounce rate, the more interested visitors are in the website .
Average visit time ： It refers to the length of stay of each visit , The larger the indicator , It means that the longer visitors stay on the web page , The more interested you are in the website .
Average number of pages visited ： It refers to the number of pages viewed per visit , The larger the indicator , It shows that visitors are more interested in the website .

（3） Transformation indicators

Number of conversions ： It refers to the number of times visitors arrive at the conversion target page , Transformation means that the visitor has done what the website administrator wants the visitor to do , Related to the promotion purpose expected by the website operator .
Conversion rate ： It refers to the efficiency of access conversion , Conversion rate = Number of conversions / Number of visits , Higher conversion rate , It shows that the promotion effect of the website is better .
Transformation path ： Path refers to a series of intermediate pages that visitors pass through before arriving at the target page you set , What you expect from a prospect is to complete a transformation , It is closely related to your promotion purpose and the definition of promotion effect . By tracking the transformation path , You can learn about visitors' access to each step of transformation .

（4） Retention indicators

Number of retained persons ： Registered users after their first visit to the website , Number of people who continue to use the website in the next cycle .
Retention （Retention Rate）： Number of people who continue to use the website in the next cycle , Proportion of registered users in the current period , The higher the retention rate , It shows that the higher the stickiness of users to the website .

（5） Source analysis

source ： How do visitors access the website , Direct access , Or search engines .
Search term analysis ： What search terms did you find and visit the website on various search engines .

（6） Visitor attribute analysis

The age of the visitor
Visitor's area
Old visitors / New visitors

3、 ... and 、 Exploratory data analysis

The basis of data analysis is the cognition of data , First there's data , And then there's the analysis . There's no credible data , The results of data analysis will be castles in the air . Variables and data are commonly used in data analysis ： Variables to describe the characteristics of things , Data is the specific value of a variable （ Also known as observations ）.

3.1 Variable

A variable used to describe a member of a population . In data analysis , Variable （Variable） Can be associated with properties 、 dimension （Dimension）、 features （feature） Use interchangeably . Common variables are gender 、 Age 、 height 、 Income, etc .
According to the function of variable value , Variables can be divided into qualitative variables and quantitative variables .

（1） Qualitative variables

Qualitative variables refer to the use of text to describe the characteristics of objects . Qualitative data is usually an angle of data analysis , Increase the dimension , Look at things from a different perspective , Ability to segment indicators , Increase the depth of analysis . Qualitative variables are mainly divided into three categories ：

Nominal variable ： Also called category variables , Used to classify data objects （Category）. such as , The color of hair 、 occupation .
Binary variable ： There are only two categories of variables , If the two states of the binary variable have the same value or have the same weight , Then consider that the binary variable is symmetric , for example , Gender ; Asymmetry means that the results of two states are not equally important . for example , Whether or not smoking is related to the effect of treatment , The weights are different .
Ordinal variable ： The order of variables is meaningful , Usually used for rating . Usually , Ordinal variables are qualitative text , such as , Official position 、 Consumer satisfaction . however , Ordinal variables can also be obtained by dividing numerical variables into different intervals , such as , age group .
- In ordinal variables , There is an important class of variables , It's called a time variable , Some common methods of analysis , Such as timing analysis , Periodicity analysis is based on time variables .

（2） Quantitative variable

Quantitative variables refer to the use of numerical values to describe objects , You can compare the size , Are quantifiable variables . Quantitative variables usually contain dimensions . for example , The dimension of height is cm, And the salary dimension is yuan . Data of the same dimension can be compared in size ; Data of different dimensions , After dimensioning by normalization , Comparison of size makes sense . Quantitative variables are mainly divided into two scales ：

Interval scale ： Measurable value , Expressed as an integer or a real number . such as , Age 、 salary
Ratio scale ： Proportional value . such as , Speed 、 Retention

3.2 data

Data is the specific value of a variable .

According to the type of variable , The data can be divided into ： Classified data 、 Sequential data and numerical data .

According to the purpose of data analysis , The data can be divided into ： test group （Treatment） And reference groups （Control）.

According to the type of data , The data can be divided into ： Text data 、 Numerical data and date time data .

3.3 Basic statistical description of data

Statistics is a good assistant for data analysis , View the basic statistical description of the dataset , Can help us understand the whole picture of the data , Identify the distribution of data . Because quantitative data is inherently computational , The distribution of data is usually a statistical description of quantitative data .
Basic statistical description mainly refers to the centralized trend of data 、 Discrete trends and distributions to understand data . Each statistical description , Use specific statistics to measure .

Mathematical basis of data analysis

Data and code download

The code for this tutorial series can be found in ShowMeAI Corresponding github Download , Can be local python Environment is running , Babies who can surf the Internet scientifically can also use google colab One click operation and interactive operation learning Oh ！