Perform data mining prediction analysis of large data application at the beginning of distribution chain

In this article, learn how to apply predictive analysis to improve the business operations of multiple functional departments of a wholesale distribution enterprise, and understand the IBM product set, including the http://www.aliyun.com/zixun/aggregation/13568.html "> The research of large data technology and the early application of large data technology with the growth of skills and data in the future.

This article focuses on the predictive analysis and its related analysis process, and you may have learned a lot about it in other forums. These articles are not limited to boring technical discussions. You can read the predictive analysis (the terminology for data mining updates and extensions) in IT and technical journals and in the Journal of Business operations and distribution, and even in regular news magazines.

Although predictive analysis is not a mainstream method in wholesale distribution, the application of this method will be more and more extensive. Like many other technologies, this approach was first adopted in some large enterprises, followed by a number of medium sized businesses. As the adoption increases, many commercial and open source tools are also derived. This is a very large number of tools, and if not experts in the field, it is likely to be overwhelmed when selecting the toolset.

Definition of predictive analysis

First, let's take a look at what the definition of predictive analysis does not include:

is not a reporting feature. Providing aggregated information from a transactional database is a useful feature, but this is not a predictive analysis. Predictive analysis uses statistical processes to provide enterprise users with information that cannot be collected through traditional reports.
is not an online analytical processing (OLAP), cube (data cube), or memory database. Although the advent of non relational data storage technologies drives the delivery of information to enterprise users, it is still not a predictive analysis. This is not to degrade the performance of the memory database and OLAP engine, but because putting historical information into these formats alone does not provide more insight to business decision-makers.
Not a spreadsheet. This is part of the backlog. The most popular spreadsheet application does provide a richer statistical function than the average maximum (max), Minimum (min), sum (sum), and average calculation. (However, few people know how to use more advanced statistical functionality). This spreadsheet can perform multiple regression analyses and is useful for predicting future trends. However, the spreadsheet has great limitations in terms of the amount of data that can be processed, the processing speed, and the ability to apply forecasts (that is, to predict new data and notify other parties of the forecast results).
What is predictive analysis depends on your conversation object. My general definition of it is that predictive analysis refers to the process of analyzing data using automated statistical processes and summarizing the results into useful information. Useful information can take a variety of forms, but for distributors, useful information should be manipulated by business decision makers, or encoded into applications, and automatically included in business logic based on enterprise resource Planning (ERP).

Predictive analysis is useful because everyone in your ERP system and other non-ERP databases needs to understand, analyze, and process large amounts of data. Order history data, customer relationship Management (CRM) data, and procurement and inventory data will be entered into the ERP system at a certain speed and accumulate, and this speed will not exceed the server's processing capacity. You will summarize this information through the report, which will be viewed frequently by corporate executives and line of business (LOB) users. However, the historical information itself does not provide predictable or normative advice. And that's where predictive analytics works.

The concepts, technologies, and tools used by large distributors can be successfully applied to the operations and data of medium sized distributors. Let's take a look at where you can use the data in your ERP application through predictive analysis. Then, learn about tools and concepts for implementing and deploying predictive analysis and large data technologies, as well as for using large data technologies for unstructured or semi-structured data.

Example of predictive analysis in distribution field

By quickly searching for relevant terminology on the Internet, you will find many examples of applying predictive analysis to different functional departments. Here are some examples that I like better.

Procurement

The application of predictive analysis in procurement optimization has a long history. medium-sized Distributors typically install systems to look at inventory and order history for a variety of products to provide recommended purchases and procurement schedules. The real effect is to reduce inventory levels.

Seasonal fluctuations are often a hidden cause of excess inventory. Predictive analysis can show seasonal fluctuations. More impressively, some distributors use predictive analysis to identify continuous seasonal fluctuations. For example, holiday decorations distributors through the predictive analysis process can be found that artificial Christmas trees and decorative lights sales show the same seasonal trend, but the peak sales and sales of the continuous period between 5 days.

Financial

Customer credit has always been a difficult field to deal with. When your company builds historical data, you can apply predictive analysis to your CRM and Accounts receivable (AR) files to monitor individual customers and customer groups. Typically, for new customers of the company, the customer's credit will be increased according to the external agent's customer report. The report is rarely viewed unless the customer has problems with AR when the payment is overdue or the total amount of credit increases. The predictive analysis model can look at the history of the customers who have gone bad and find out the warning signs. Some distributors combine AR and CRM files in the predictive model and find that customer callback latency is an important warning signal.

Marketing

Grouping customers is a common way for most businesses to plan and locate. Although heuristic customer groupings are very easy to implement, predictive analysis techniques can be used to create a more optimized partitioning system. This grouping model applies to potential customers and new customers. The specific use includes a quick match between the new customer's order model and the pattern of the long-term performance of the same group.

Another outstanding application of the customer grouping model in the distribution area is for customer lifecycle management. Understanding how customers can become your customers through different stages will help your company develop plans and incentives to retain these customers.

Sales

Some concepts that use predictive analysis in marketing can also be applied directly to the sales department. Understanding the customer lifecycle helps salespeople find out where the distributors are losing their business.

Predictive analysis has many other potential applications in the sales department. One of my favorite applications is the cross-selling model, an automated or semi-automatic system that shows the products that customers are most likely to buy but have not yet purchased. Increasing the order amount is one of the best ways for distributors to increase profits directly. True predictive analysis is not just a display of hot products (top product) that a department uses for cross-selling. Customers usually have purchased these products. The best cross-selling Model can act as a personal shopping assistant, recommending some products with positive correlation but not obvious.

Cross-sales models can also help convert Low-value customers to high-value customers. Imagine that the model allows salespeople to increase the number of items a customer buys at a time. This method is very subtle and works well.