From data to information to decision-making

As the saying goes, if you forget history, you will betray yourself. It is no better to start this article today.

Based on a virtual story, this article describes how to use historical data to help a salesperson discover regular information and help him make decisions.

 

Tim, the main character of this article, is in a sales department. Recently, the Department decided to create a new sales plan. Then, according to the end of the plan, each Salesperson's sales performance is assessed by KPI.

After Tim's Department determined the sales task, others quickly invested in the store's sales work, while Tim ran to the company's IT maintenance department, A historical customer data is sent to the IT department.

At this time, some people have already blamed Tim for saying that you do not want to run the business outside, but how did you go to the IT department to "do not do business"? Moreover, they wanted to focus on the previous customer data and did not care about new customers.

 

In fact, Tim also has a friend of it, James, who is a data analyst and has some experience in the retail industry. Tim's first thought was to give James some suggestions.

James told Tim countless times about business intelligence while chatting with two people. Although Tim is engaged in sales, he is often instilled by James and has some impressions on his knowledge, such as data mining, of course, this concept is simply tianshu for Tim who is engaged in sales. It is called the separation of lines, but he understands it very well, that is: from data to information, that is,Obtain information from data.

So at the beginning of the sales task, Tim ran to the IT Department and asked for such a data to see what information James could get from here, so that Tim can know more accurately what kind of recommendation product is more reliable.

After James got Tim's data, he roughly browsed it:

The data structure is as follows:

From this data, James shows that it contains the customer's gender, marital status, annual income, family-related and education information. The last key column is whether the customer has purchased the product. If you have purchased it, It is counted as 1; otherwise, it is counted as 0.

James took the Excel file and first made a key factor analysis.

Based on this tool, first specify the column of interest, that is, whether the customer has purchased the product Tag:

Here, select bikebuyer.

Then click choose columns to be used for analysis.

Here, James specifies the columns to be analyzed based on experience. Obviously, datafirstppurchase is useless. James decisively removes this column to avoid affecting the accuracy of the analysis.

Then the system automatically processes the historical data according to James's settings.

After processing, the system generates a report:

James sent an email to TIM:

Dear Tim,

I analyzed the data you provided to me and obtained several rules from the data.

First, pay attention to users who do not have a car, have a child, and come from pacpacific, and who are not too far away from work at ordinary times. They are likely your potential customers.

In addition, for customers with two cars, do not recommend them. From your business records, it is unlikely that such customers will purchase products.

There are also a large number of children, working too far away, more than 65 years old to become your customer is also very unlikely.

Above.

Best wishes!

James.

A year, a month, a day

Tim was very happy after receiving this email, because it could immediately let him determine whether a new customer would buy the product, therefore, customers of this type will not spend too much time on their own, so that they can devote their energy to more target customers.

But soon, Tim had another problem, that is, simply relying on such a judgment is too general and it is easy to lose some very special customers, so Tim hopes to make a more detailed judgment based on the customer's situation.

After receiving this request from Tim, James created a mining calculator in Excel.

First, click the prediction computing tool.

Set the columns to be predicted in the tool.

Click Run and Excel to process data through SQL Server's Analysis Service.

After data processing, several reports are generated in Excel:

In the first report, James gets the list, which identifies the extent to which each attribute affects an unknown customer's purchase of the product.

In another report, the analysis data contains a dynamic operation table.

The value of each attribute is changed to a drop-down list, and the impact values of each attribute are combined to get a score. If the score reaches a certain height, this indicates that this customer is likely to purchase the product.

So James sent the Excel file to Tim, so that Tim could select the items based on the collected customer information, and then determine whether the user is a potential customer through calculation.

This file helped Tim a lot and identified some customers accurately. However, Tim often complained that the computer was not always around when he went out to run the business, so it was often difficult to make judgments in a timely manner.

James knows Tim's troubles and tells him that you can print out the table in the third analysis report.

This table lists the values of each attribute and their corresponding scores. After Tim is printed, You can manually computation on it.

After calculating the total score, compare the score below. That is to say, a potential customer may have to earn at least 601.

As a result, Tim, who is not very familiar with IT system operations, can make potential customer judgments for new customers each time.

 

 

In this story, James does not use any complicated data. In the whole article, he only uses one software, that is, Excel, starting from MySQL 2007, Excel can use SQL Server's function expansion to implement simple data mining. It uses SQL Server Analysis Services to generate a temporary mining model, using Sample Data and mining models andAlgorithmTo find information such as regularity and relevance in the data.

Through Excel's encapsulation of SQL Server Data Mining functions, You can implement the data mining function even if you do not know the specific data mining algorithms, so that you can perform mining, prediction, and analysis to assist in decision-making, you don't even need to know what kind of Mining Model algorithms are suitable for solving the problem. You only need to pay attention to the table analysis tools in Excel to perform simple prediction analysis.

In short, data mining is not exclusive to data analysts. You can also use Excel.

 

 

To use data mining in Excel, You need to download SQL server data mining tools add-ins for office.

The download page for the corresponding office 2010 version is:

Http://www.microsoft.com/en-us/download/details.aspx? Id = 29061

Different Languages and x86 and x64 versions are available respectively.

In addition, you can view Microsoft's official Excel Data Mining video. Although it is in English, it has Chinese subtitles:

Http://msdn.microsoft.com/zh-cn/library/dd299412 (V = SQL .100). aspx

At the same time, for the Application Scenario of shopping basket, I have another article:

Use Association Rules of SQL Server Analysis Services data mining to implement commodity recommendation

Http://www.cnblogs.com/aspnetx/archive/2013/02/25/2931603.html

The first three articles in this series use SQL Server Analysis Services to describe how to implement a product recommendation function in front-end applications.

The last two articles are about how to implement this recommendation function in Excel.

In addition, if you are interested, you can also download the Excel data sample file mentioned in the article and try this function by yourself:

Http://files.cnblogs.com/aspnetx/DMINF.zip

These sample data are from the official Microsoft demo database adventure works.