常见问题解答
关于数据采集和分析的一些问题
- Data collection
- Retail smart gateway
- Rules of intelligent analysis algorithm
- Sales analysis glossary
- 数据采集
- 智能分析算法规则
- 销售分析专业词汇
- 零售智能网关
数据采集设备的兼容性是指三个方面的兼容性：打印机类型、打印机端口类型和打印指令类型。
打印机有多种类型，例如办公室常见的激光打印机和喷墨打印机。用于零售行业销售点的凭证打印机，绝大多数是热敏凭证打印机和点阵凭证打印机，也有个别使用宽幅点阵打印机甚至办公室用激光打印机或喷墨打印机的情况。
酷方支持绝大多数常用的销售点热敏打印机和点阵打印机，并支持宽幅点阵打印机，暂不支持办公室常见的激光打印机和喷墨打印机。
在酷方支持的打印机中，我们兼容几乎所有打印端口，包括：USB接口打印机、并口（高级并口）打印机、串口打印机和以太网口打印机。
打印机的打印指令纷繁复杂，在用协议总量超过40个。酷方支持用于销售点凭证打印机和宽幅点阵打印机常用打印协议10余个，并支持数十家打印机厂牌的扩展指令。
数据采集设备的稳定性，包括对原有设备的干扰程度，设备本身的稳定运行和被采集数据的完整性与安全性。
酷方采用硬件旁路监听方案，从原理上杜绝了对被监听设备的干扰及影响。即便酷方自身被断电，仍不会打断销售点的正常收银业务。
酷方采用工业级的成熟方案，支持7×24不间断工作。
酷方本身具有数据本地保存功能，可以保障在网络故障时，将采集数据缓存到本地，以防止丢失。
酷方包含可选的备用电池，当外部电源意外断开时，备用电池开始供电，保障数据的完整性。
| 硬件采集 | 软件采集 | |
|---|---|---|
| 兼容性 | 常见接口打印机 | 在常见接口打印机的基础上，可以支持一体机 |
| 打印环境变更 | 需重新设置 | 需重新设置，但部分收银机会每日或定期初始化系统，则软件无法支持 |
| 跟打 | 支持 | 支持 |
| 安装便利性 | 不更改收银机任何设置 | 需要管理员权限进行安装，并需要更改收银机相关设置 |
| 网络 | 设备自带网络 | 需使用收银机自带网络，或外接网络设备 |
| 安全性 | 单向传输，不会对收银机有任何影响 | 有安全隐患，软件本身有被攻击可能；如果需要收银机连接公网才能传输数据，则收银机便暴露在公网环境中 |
| 稳定性 | 硬件本身有损坏概率 | 如不使用外接网络设备，则不存在硬件稳定性问题 |
| 计算消耗 | 对收银机无影响 | 利用收银机算力，需要一定的计算资源 |
| 第三方设备接入 | 实施简单，对收银机无影响 | 需占用收银机端口资源，并需做更多调试 |
| 成本 | 较高 | 较低 |
零售智能网关，是一种全功能的销售点边缘网关，它集成了传统的销售点凭证打印功能，并支持无线打印。
零售智能网关，配合我们的多种监听配件，可以采集销售点收银业务的完整数据，包括：凭证打印机、客显、金融手柄、键盘、扫码枪等几乎所有设备。
零售智能网关，本身是一台无线路由器，可以支持2.4G/5G的Wi-Fi接入。
零售智能网关，除了支持Wi-Fi接入外，还支持BLE无线接入和ZigBee无线接入，可以支持多种无线传感器，可广泛用于客流追踪、产品追踪等店级IOT应用。
通过选择专用的智能屏幕，或是无线接入移动设备，可以为消费者提供更多新的营销交互，或为店员移动跟随服务提供有效保障。
交易金额
交易额、营业额、交易流水、销售额
指定时间段内，所有实际完成交易获得的总金额
指定时间段内，报告对象所有单笔交易发生的金额的求和
客单价
平均单价
指定时间段内，单笔交易的平均金额
指定时间段内，报告对象每笔交易金额的加权平均值
交易笔数
销售笔数、水单数
指定时间段内，所有完成交易的次数
指定时间段内，报告对象所有单笔交易的次数之和
客流
客流量
指定时间段内，在某个场所停留过的总人数
指定时间段内，初始人数 + 进入人数
人效
单人销售额
指定时间段内，单个销售人员完成的销售额
指定时间段内，销售额 / 销售人员数量
坪效
单位面积销售额
指定时间段内，每平方米产生的销售额
指定时间段内，报告对象的交易额 / 店铺面积
商户数据覆盖率
–
指定时间段内，具有有效数据的商户数量占所有商户数据的百分比
指定时间段内，有效数据的商户数量 / 所有商户的数量
购买频次
–
指定时间段内，商品（SKU）在交易中出现的次数
指定时间段内，出现统计对象（通常是单一商品）的交易笔数之和
区域
做经营分析时的数据统计范围，通常以单一楼层做为一个区域
业态
商户的经营特色属性，以销售对象和销售产品特色来进行划分。常用的业态分类有餐饮，零售，生活服务，休闲娱乐，儿童亲子，超市
异动（商户）
指定时间段内，销售金额上升或下降变化幅度最大的商户
重点关注（商户）
指定时间段内，销售金额最高的商户
关联购买（商品）
在同一笔交易中完成购买的商品组合
热卖商品 / 活跃商品
指定时间段内，有过实际销售的商品
新品商品
在交易中首次出现购买的商品（SKU），新品自首次出现30天后即作为正常商品，不再算作新品
特卖商品
指定时间段内，有促销活动的商品
本报告体系中，同比分析的定义为历史同期数据分析，以年为单位，日报为去年同日，周报为去年同周，月报为去年同月。
偏离值处理 – 数据缺失
由于外力因素，数据可能产生缺失，为确保分析的延续性和可读性，需要针对数据进行推算还原，数据推算规则如下：
情景1：推算商户Store历史数据
1. 推算基本对象：商户
2. 推算基本报告时段：日报
3. 数据推算最大时间跨度：90天
4. 可推算指标：销售额，销售笔数
第一步：累计商户一个月的数据Data-Store-MonA，计算当月的数据平均值Ave. Data-Store作为基数。
第二步：以Mall的同一业态同一个月的数据Data-Cat-MonA，计算当月的数据平均值Ave. Data-Cat。
第三步：以所在Mall的同一业态当月的数据平均值Ave. Data-Cat为基数100%，推算Mall的同一业态过去3个月的数据变动指数Cat-Index_N（N取值为1到90）。
第四步：利用购物中心业态的数据变动指数Cat-Index_N，和商户基数Ave. Data-Store反推过去90天商户的各项销售数据。
当商户开业之日少于90天，数据推算到实际开业当天。
情景2：推算商户Store当期偶发缺失数据
1. 推算基本对象：商户
2. 推算基本报告时段：日报
3. 数据推算最大时间跨度：14天
5. 可推算指标：销售额，销售笔数
第一步：计算商户历史的日数据平均值。
第二步：利用商户历史的日数据平均值，作为当期数据。
日数据的平均值以每周同期进行计算，即周一的平均值为历史数据周一的平均值。
商户历史数据使用4周平均。
当历史数据少于4周，大于1周，可按照实际历史数据推算。
当历史数据少于1周，视同情景1，推算商户Store 历史数据处理。
本报告体系对待数据缺失，采用多值插补的方法来规划推算。多值插补的思想来源于贝叶斯估计，认为待插补的值是随机的，它的值来自于已观测到的值。具体实践上通常是估计出待插补的值，然后再加上不同的噪声，形成多组可选插补值。根据某种选择依据，选取最合适的插补值。
本报告体系中，环比分析的定义为相邻两个时段的数据进行分析，日报为前一日，周报为前一周，月报为前一月，年报为前一年。
偏离值处理 – 数据波动过大
由于日常经营活动中，商户的经营数据会产生比较大的变化。因此，本报告体系对数据波动设定阈值，当数据变动阈值超过预设时，即产生提醒，并对异常数据进行校正。
1. 数据校正对象：商户
2. 数据校正基本报告时段：日报
情景1：商户历史日交易笔数超过或等于15笔
当商户交易额超过过去4周同期平均值的200%时，系统将对商户数据自动进行排查。在排查过程中，相关数据将以过去4周同期平均值的200%，在报告中显示。 经过数据排查，商户数据将以实际值显示在系统中。
情景2：商户日交易笔数超过或等于8笔，少于15笔
当商户交易额超过过去4周同期平均值的300%时，系统将对商户数据自动进行排查。在排查过程中，相关数据将以过去4周同期平均值的300%，在报告中显示。 经过数据排查，商户数据将以实际值显示在系统中。
情景3：商户日交易笔数少于8笔
当商户交易额超过过去4周同期平均值的400%时，系统将对商户数据自动进行排查。在排查过程中，相关数据将以过去4周同期平均值的400%，在报告中显示。 经过数据排查，商户数据将以实际值显示在系统中。
对于任意ε>0，有：
。当
时，如果总体为一般总体的时候，统计数据与平均值的离散程度可以由其标准差
反映，因此有：
。
本报告体系根据不同商户的业态类型和经营状况，计算分层标准差基数，用于检验商户销售额的偏离度。
The compatibility of data collection device refers to three aspects: printer type, printer port type and printing instruction type.
There are many types of printers, such as laser printers and inkjet printers that are common in offices.
Most of the receipt printers used in POS of retail industry are thermal receipt printers and dot matrix receipt printers, and some use wide-format dot matrix printers, even laser printers or inkjet printers used in offices.
Among the printers supported by Counect CUBE, we are compatible with almost all printing ports, including USB interface printer, parallel (advanced parallel) printer, serial printer and Ethernet printer.
The printing instructions of the printer are very complex, and the total number of protocols in use is more than 40. Counect CUBE supports more than 10 common printing protocols for POS receipt printers and wide-format dot matrix printers, and supports dozens of printer brand expansion instructions.
The stability of data collection device includes the degree of interference to the original device, the stable operation of the device itself and the integrity and safety of the collected data.
Counect CUBE adopts hardware bypass monitoring technology to eliminate interference and influence on monitored device in principle. Even if CUBE itself is powered off, it will not interrupt POS’s normal cashier business.
CUBE itself has the function of data local saving, which can guarantee to cache the collected data to the local in case of network failure, so as to prevent data loss.
CUBE contains optional battery backup. When the external power supply is disconnected accidentally, the backup battery starts to supply power, thus ensuring the integrity of data.
|
Hardware data collection |
Software data collection |
|
|---|---|---|
|
Compatibility |
Printers with common interfaces |
In addition to printers with common interfaces, united printers can be supported |
|
Print environment change |
Reset required |
It needs to be reset. However, some cash registers will initialize the system on a daily or regular basis, which is not supported by the software |
|
Follow-printing |
Support |
Support |
|
Installation convenience |
No change of any settings of cash register |
Administrator permission is required for installation, and relevant settings of cash register need to be changed |
|
Network |
Device’s own network |
Need to use the network provided by the cash register or external network equipment |
|
Safety |
Unidirectional transmission, no impact on cash register |
There are security risks. The software itself may be attacked. If the cash register needs to be connected to the public network to transmit data, the cash register will be exposed to the public network environment |
|
Stability |
Probability of hardware damage |
If the external network device is not used, there is no hardware stability problem |
|
Computational consumption |
No impact on cash register |
Using the computing power of the cash register, certain computing resources need to be occupied |
|
Third party device access |
Simple implementation, no impact on cash register |
Need to occupy cash register ports, and need to do more debugging |
|
Cost |
High |
Low |
A retail smart gateway is a full-featured POS edge gateway. It integrates the traditional POS receipt printing function and supports wireless printing.
Through the cooperation of smart retail gateway and our various monitoring accessories, the complete sales data of POS can be collected, including: receipt printer, customer display, keyboard, scanner and almost all other devices.
A retail smart gateway is a wireless router that can support 2.4G/5G Wi-Fi access.
In addition to supporting Wi-Fi access, the smart retail gateway also supports BLE wireless access and ZigBee wireless access, as well as multiple wireless sensors. It can be widely used in store IOT applications such as customer flow tracking and product tracking.
By choosing a dedicated smart screen or wireless access to mobile devices, we can provide more new marketing interaction for consumers, or provide effective guarantee for mobile follow-up service for employees.
Sales Value
sales amount, sales, turnover or sales volume
- The total amount of all transactions actually completed within a specified time period
- Sum of the amount of all single transactions of the reporting object in a specified time period
Per Customer Transaction
ATV (Average transaction value)
- Average amount of a single transaction in a specified time period
- Weighted average value of each transaction amount of the reporting object within a specified time period
Sales Transaction
transactions, transaction amount
- The number of times all transactions have been completed in a specified time period
- Sum of the times of all single transactions of the reporting object in a specified time period
Customer Flow
traffic
- The total number of people who have stayed in the mall or store in a specified period of time
- Initial number of people + number of people entering in a specified time period
Sales per Sales-person
SPS
- Sales completed by a single salesperson in a specified time period
- Sales value/number of sales personnel within a specified time period
Sales per unit area
- –
- Sales per unit area in a specified time period
- Sales value / store area of the reporting object within a specified time period
Tenant Data Coverage
- –
- The percentage of the number of tenants with valid data in the number of all tenants in the specified time period
- Number of tenants with valid data / number of all tenants in a specified period of time
Purchase Frequency
- –
- The number of times a SKU appears in transactions in a specified time period
- The sum of the number of transactions of the statistical object (usually a single SKU) in the specified time period
Zone
The statistical range of data for operation analysis, usually with a single floor as a zone.
Category
The business characteristics of a tenant, usually divided based on sales target and the characteristics of products. Common categories include F&B, retail, life service, leisure & entertainment, kids, and supermarket.
Changes (Tenant)
Tenants with the largest increase or decrease in sales value in a specified period of time.
Key tenants
The tenant with the highest sales value in the specified time period.
Relevant purchase (SKU)
SKU combination purchased in a same transaction.
Hot/Active SKU
Products with actual sales in a specified time period.
New SKU
SKUs purchased for the first time in a transaction. New SKU is regarded as normal goods 30 days after its first appearance, and is no longer regarded as a new product.
Special offer
SKUs with promotions within a specified time period.
In this report system, year-over-year analysis is defined as historical data analysis in the same period. The daily report is the same day of last year, the weekly report is the same week of last year and the monthly report is the same month of last year.
Deviation value processing – data missing
Data may be missing due to external forces. In order to ensure the continuity and readability of the analysis, it is necessary to calculate and restore the data. The data calculation rules are as follows:
Scenario 1: Calculate historical data of tenant’s store
1. Basic calculation object: Tenants
2. Calculation of basic reporting period: Daily Report
3. Maximum time span of data calculation: 90 days
4. Calculable index: sales value, sales transactions
Step 1: accumulate one month’s data of tenants (Data-Store-MonA), and calculate the average data of that month as the base (Ave. Data-Store).
Step 2: calculate the average data (Ave. Data-Cat) of the same month based on the data of the same category of the shopping center (Data-Cat-MonA).
Step 3: take the average data of the same category of shopping center in the current month (Ave. Data-Cat) as the base 100%, calculate the data change index (Cat-Index_N) of the same category of shopping center in the past three months (n value is 1-90).
Step 4: use the data change index (Cat-Index_N) of shopping center’s category and the tenant base (Ave. Data-Store) to deduce all sales data of tenants in the past 90 days.
If the opening date of the tenant is less than 90 days, the data will be calculated to the actual opening date.
Scenario 2: calculate the occasional missing data of tenant’s store in the current period
1. Basic calculation object: Tenants
2. Calculation of basic reporting period: Daily Report
3. Maximum time span of data calculation: 14 days
4. Calculable index: sales value, sales transactions
Step 1: calculate the average historical daily data of the tenant.
Step 2: use the average historical daily data of tenants as the current data.
The average value of daily data is calculated based on the same period of each week, that is, the average value of Monday is the average value of Monday historical data.
Tenant historical data usage is 4 weeks average.
When the historical data is less than 4 weeks and more than 1 week, it can be calculated according to the actual historical data.
If the historical data is less than 1 week, it is considered as scenario 1 to calculate the historical data processing of tenant’s store.
For the lack of data, this report system adopts the method of multi value interpolation to plan and calculate. The idea of multivalued interpolation comes from Bayesian estimation, which considers that the value to be interpolated is random, and its value comes from the observed value. In practice, it is usually to estimate the value to be interpolated, then add different noises to form multiple groups of optional interpolation values, and then select the most appropriate interpolation value according to a certain selection basis.
In this report system, the definition of link analysis is to analyze the data of two adjacent periods. The daily report is the previous day, the weekly report is the previous week, the monthly report is the previous month, and the annual report is the previous year.
Deviation value processing – excessive data fluctuation
In daily operation, the business data of tenants will change a lot. Therefore, this report system sets a threshold value for data fluctuation. When the threshold value of data fluctuation exceeds the preset value, a reminder will be generated and abnormal data will be corrected.
1. Data correction object: Tenants
2. Basic report period of data correction: Daily Report
Scenario 1: the number of historical daily transactions of tenants is more than or equal to 15
When the sales value of a tenant exceeds 200% of the average value of the same period in the past four weeks, the system will automatically check the tenant’s data. During the checking, the relevant data will be displayed in the report at 200% of the average value of the same period in the past four weeks. After checking, tenant data will be displayed in the system with actual value.
Scenario 2: the number of daily transactions of tenants is more than or equal to 8 and less than 15
When the sales value of a tenant exceeds 300% of the average value of the same period in the past four weeks, the system will automatically check the tenant’s data. During the checking, the relevant data will be displayed in the report at 300% of the average value of the same period in the past four weeks. After checking, tenant data will be displayed in the system with actual value.
Scenario 3: the number of daily transactions of tenants is less than 8
When the sales value of a tenant exceeds 400% of the average value of the same period in the past four weeks, the system will automatically check the tenant’s data. During the checking, the relevant data will be displayed in the report at 400% of the average value of the same period in the past four weeks. After checking, tenant data will be displayed in the system with actual value.
For any ε>0, there are: . When , if the population is a general population, the degree of dispersion between the statistical data and the average value can be reflected by its standard deviation, so there are: .
This report system calculates the hierarchical standard deviation base according to the categories and operating conditions of different tenants, which is used to test the deviation degree of tenant sales value.