Eight, data analysis methods

ABC analysis (Pareto analysis)

1. Description

1.1 concept

Classification ABC (Activity Based Classification), the full name should be classified as ABC stock control. Barreto also known as the Pareto analysis or analysis, Pareto analysis, due to primary and secondary analysis, ABC analysis, ABC Management Act, and are usually called "80 pairs of 20" rule.

According to the main characteristics of things in technical or economic terms, be classified queuing to distinguish between general priorities and, thus determining the difference management. It is an object to be analyzed is divided into A, B, C categories, three types of items there is no clear demarcation value.

A items are very important The proportion of the number of small, high-value accounting
Class B is more important items A no less important items are, A is interposed, between C
Class C items as important But the proportion of large number of small value accounting

The core idea of ​​classification: a few items contributed most of the value. With style and sales, for example: A number of models accounted for 10% of the total, but contributed 80% of sales.

1578148917128584.png

1.2 Effect FIG.

40.png

2. realization of ideas

The existing data processing, and in descending order, obtaining cumulative amount and cumulative proportion, according to the use and display the instrument panel, and based on the integrated accounting object divided into three categories.

1578484329603640.png

3. Example 1: Self-implemented data set

By making self-service data sets to achieve.

Following the "retail" business package "sales list" and "brand dimension table" for example, sales of major brands were Pareto analysis, and draw the most important brands.

3.1 Add association between tables

Ready to enter the data interface, select "retail" business package, and on "sales list" to establish a "brand dimension table": N association, the association for the brand numbered field, as shown below:

1578140755323923.png

3.2 Creating Self dataset

3.2.1 Add field

Click to add a table, add a self-service data set, select the "Brand dimension table" of "Brand description" field and "sales list" of "sales" field, as shown below:

1578141160147040.png

1578141178838113.png

3.2.2 group summary

1) Click + selected group summary, as shown below:

1578141404488334.png

2) Description brand drag packet block, the block sales drag summarized as shown below:

1578141543783400.png

3.2.3 Sorting

1) Click +, selection sort, as shown below:

1578141657347816.png

2) Click add a sort column select Sales field, as shown below:

1578141810807342.png

3)选择降序排列,如下图所示:

1578141908240595.png

3.2.4 新增求和列

1)点击+,选择新增列,如下图所示:

1578142090340626.png

2)给新增列命名为「销售总额」,选择所有值/组内,取值规则为「所有值」,数值来自「销售额」,统计方式为「求和」,如下图所示:

1578142758826615.png

3)得到销售总额字段如下图所示:

1578142501825132.png

3.2.5 新增累加列

1)点击+,选择新增列,如下图所示:

1578142540296873.png

2)给新增列命名为「累计总额」,选择累计值/组内,取值规则为「累计值」,数值来自「销售额」,点击确定,如下图所示:

1578142699906399.png

3)得到累计总额字段如下图所示:

1578142933371878.png

3.2.6 新增累计占比列

1)点击+,选择新增列,如下图所示:

1578142958240265.png

2)给新增列命名为「累计占比」,输入公式累计总额/销售总额,点击确定,如下图所示:

1578143081297302.png

注:公式中累计总额和销售总额不能手动输入,需要点击数值字段的字段名。

3)得到累计总额字段如下图所示:

1578143211823047.png

 

为自助数据集命名为「帕累托图分析表」,并点击保存,进入数据准备界面,点击更新数据,如下图所示:

1578143438625800.png

 

4. 示例二:仪表板创建计算指标实现

4.1 创建自助数据集

从「零售行业」业务包中选择销售日期、店性质、品类描述、品牌描述、销售额字段。创建自助数据集「帕累托数据」,如下图所示:

1578484860147135.png

4.2 创建仪表板

4.2.1 计算总销售额

1)创建新的仪表板并将其命名为「商品销售帕累托分析」,添加计算指标使用 SUM_AGG() 函数,计算每种品牌的销售总额,如下图所示:

462.png

4.2.2 计算累加总销售额

点击添加计算指标,命名为累计销售额,输入公式 ACC_SUM(SUM_AGG(销售额),0),点击确定,如下图所示:

其中 ACC_SUM()表示根据当前维度字段对指标进行跨行累计计算。

75.png

4.2.3 计算累计销售额占比

点击添加计算指标,命名为累计销售额占比,输入公式 ACC_SUM(SUM_AGG(销售额)/TOTAL(SUM_AGG(销售额,0,"sum")),点击确定,如下图所示:

其中 TOTAL() 表示根据当前维度字段对指标进行跨行汇总计算。SUM_AGG(销售额)/TOTAL(SUM_AGG(销售额,0,"sum")表示不同商品销售额占比。

1578539227660643.png

4.2.4 制作图表

1)点击创建组件按钮,命名仪表板为「帕累托图分析法」,点击确定,进入仪表板编辑界面,如下图所示:

1578143553920491.png

2)将字段拖入对应横纵轴,如下图所示:

412.png

3)选择组合图,并设置销售额为柱形图,累计占比为折线图,如下图所示:

361.png

4)对累计销售额占比字段设置值轴,如下图所示:

1578146121770502.png

详细设置参见:图表设置轴 。

5)对销售额字段进行降序排列,如下图所示:

1578551753166648.png

6)添加计算指标,命名为 ABC 划分,输入公式 IF(累计销售额占比<0.8,1,IF(累计销售额占比>0.9,3,2)),其中 1 代表 A 类商品,2 代表 B 类商品,3 代表 C 类商品,如下图所示:

29.png

7)将 ABC 划分字段拖入图形属性>销售额下的颜色框,并设置需要的颜色,如下图所示:

1578556020357719.png

8)得到帕累托图如下图所示:

309.png

同时可以设置动态帕累托图,增加过滤组件和其他需要的组件类型,如下图所示;

40.png

5. 结论分析

创建仪表板得出以下结论:

  品牌名称 数量占比 销售额占比
A 类商品 ZIPPO(之宝)、PAW IN PAW、NEW BALANCE(新百伦)、HANG TEN 40% 80%
B 类商品 SINOMAX(丝梦露)、O.C.T.MAMI(十月妈咪) 20% 10%
C 类商品 WHO.A.U、RACB JJQN、LESPORTSAC、X.ZHINING 40% 10%
 
----------------------------------------------------------------------------------------------------------------------------------------

RFM 分析法

1. 描述

1.1 概念

RFM 分析是美国数据库营销研究所提出的一种简单实用客户分析方法,发现客户数据中有三个神奇的要素:

最近一次消费(R):客户距离最近的一次采购时间的间隔。

消费频率(F):指客户在限定的期间内所购买的次数。

消费金额(M):客户的消费能力,通常以客户单次的平均消费金额作为衡量指标

这三个要素构成了数据分析最好的指标,RFM 分析也就是通过这个三个指标对客户进行观察和分类,针对不同的特征的客户进行相应的营销策略。

1.2 效果图

自助数据集效果:

383.png

仪表板效果:

26.png

2. 实现思路

1578228119140164.png

3. 示例

以「样式数据」业务包下的「RFM明细数据」为例,对客户消费明细进行分析,将客户进行分类。

3.1 计算消费金额、最近消费距离、消费频次

3.1.1 选字段

1)进入数据准备界面,选择「样式数据」业务包,点击添加表,选择添加自助数据集,如下图所示:

1578214746295815.png

2)进入自助数据集编辑界面,选择「RFM明细数据」并添加表下的所有字段,给自助数据集命名为「RFM分析」,如下图所示:

1578214941508102.png

3.1.2 计算客户总体平均消费金额

1)点击+,选择新增列,如下图所示:

1578215832453916.png

2)给新增列命名为「客户总体平均消费金额」,选择所有值/组内,取值规则为「所有值」,数值来自「MONEY」,统计方式为「求平均」,点击确定,如下图所示:

1578226648204213.png

3.1.3 计算每个客户的消费频次、每个客户每次消费的平均金额、最近一次消费时间

1)点击+,选择左右合并,如下图所示:

1578216088415223.png

2)将 CONPANY 、CUSTOMERNAME 、CUSTOMERTYPE、客户总体平均消费金额字段拖入分组框,将 DATE 拖入汇总框并设置为最晚时间,将 MONEY 拖入分组汇总框并设置为求平均,将 ACCOUNT 拖入汇总框并设置为记录个数,如下图所示:

1578217499126486.png

3)点击+,选择新增列,如下图所示:

1578215832453916.png

4)设置新增列名为最近一次消费距离时间,选择时间差,设置时间差=系统时间-DATE ,计量方式为天,点击确定,如下图所示:

1578217627352495.png

3.2 计算客户总体消费指标的平均值

3.2.1 计算客户总体消费频次的平均值

1)点击+,选择新增列,如下图所示:

1578215832453916.png

2)设置新增列名为客户总体消费频次的平均值,选择所有值/组内,取值规则为「所有值」,数值来自「AMOUNT」,统计方式为「求平均」,点击确定,如下图所示:

1578218502388499.png

3.2.2 计算客户总体最近一次消费距离时间的平均值

1)点击+,选择新增列,如下图所示:

1578215832453916.png

2)设置新增列名为总体最近一次消费距离时间的平均值,选择所有值/组内,取值规则为「所有值」,数值来自「最近一次消费距离时间」,统计方式为「求平均」,点击确定,如下图所示:

1578219558995783.png

3.3 特征向量化

根据是否大于总体的平均值水平,将客户特征进行向量化。其中在 IF(xxx>总体平均值,1,0) 中,小于总体平均的设为 0,大于总体平均的设为 1 ,使得 1 都是保持正向特征,0 保持负向特征)

1)点击+,选择新增列,如下图所示:

1578215832453916.png

2)给新增列命名为消费金额向量化,输入公式 IF(MONEY>客户总体平均消费金额,1,0),点击确定,如下图所示:

1578219856433171.png

3)同理,增加消费频次向量化和最近消费向量化字段,如下图所示:

1578220401493370.png

1578227016795527.png

 

3.4 客户特征分析

1) customers have been characterized to value, customers can be divided classified according to the following table:

Customer profile
Customer classification
Important value customers (111) Recent spending time near, frequency of consumption and the amount of consumption are high (VIP)
Important to develop customer (101) Recent spending time close to the high amount of consumption, but not high frequency, loyalty is not high, a lot of potential users, must focus on the development.
Important to maintain customer (011) Recent spending time pay far, the amount and frequency of consumption are high.
Important to retain customers (001) Recent consumer distant time, frequency of consumption is not high, but the high amount of consumption of the user, may be lost or the user will have to be lost, should be based on retention measures.
General-value customers (110) Recently spending time in recent, high frequency but low amount of consumption, needs to improve its customer unit.
General development client (100) Recent spending time close, the amount of consumption, frequency is not high.
General customer retention (010) Recent spending time far higher frequency of consumption, but the amount is not high.
Generally retain customers (000) It is not high.

2) Click Add new column, using the CONCATENATE () function to the quantized value RFM spliced together, as shown below:

1578227188171708.png

3) add a group summary, the following fields are dragged into the packet block, as shown below:

44.png

4) Set RFM field defined packets from the packet based on section 3.4 customer type analysis table, click OK, as shown below:

1578227846173419.png

5) self-storing and updating the data set, the following results:

83.png

At this point, the self-service data sets can also be classified customer related data through a dashboard visual display.

 

Guess you like

Origin www.cnblogs.com/XxXx-YA/p/12395224.html