The main model and algorithm inventory of the electric business industry

= 08 Graduation, unknowingly infiltrated the e-commerce industry, and unknowingly did three years of data analysis, just catch up with the internet E-commerce industry development of the fastest years, is not bad, after all, feeling the future is still very bright. Three years, can say with a lot of colleagues learned a lot of things, need to thank a lot of people, they selflessly taught me a lot of things.

As far as the data analysis profession is concerned, personal feeling is very important for internet companies and it is something that can actually bring real results. For example, the use of data analysis to make members of the subdivision for precision marketing; Use of data analysis to identify existing deficiencies to improve, so that customers have a better shopping experience, the use of CRM system to manage the life cycle of members, improve loyalty to members, to avoid the loss of members, the use of members of the purchase data to tap the potential needs of members, Provide sales, expand influence, etc.

When I first started in the company, I was in the operations department, is responsible for the operation of the report data, at that time the system is still poor, extraction data is very difficult to do the report is difficult, are scraped some data, and then made PPT, remember that the main data is sales, order volume, gross margin, customer unit price, per unit price, Inventory and so on some special basis of data, and then use this data to make some charts. In this stage is basically to do some data extraction work, Excel skills have learned a lot, is a data analysis primer.

Then the company went to the data Warehouse, which had a lot of raw data, the extraction of data is very convenient, and more dimensions, you can follow their own ideas at random combination analysis, that stage is mainly for members of the analysis of shopping behavior, start to contact data modeling, algorithms and some other difficult things, but also learn the most time. I remember doing a lot of analysis, weekly also to the President of the Office to report these reports, the following details of the use of some of the main models and algorithms:

1. RFM model

Model definition: In many customer relationship management analysis models, the RFM model is widely mentioned. RFM model is an important tool and means to measure customer's value and profitability. The mechanical model describes the customer's value status through a customer's recent purchase behavior, the overall frequency of purchases, and the amount of money spent on three indicators. In RFM mode, R (recency) indicates how far the customer's last purchase was, F (Frequency) represents the number of times the customer has purchased in the recent period, and M (monetary) represents the amount that the customer has purchased over the last period of time. The general analytical CRM focuses on the analysis of Customer contribution degree, RFM emphasizes to differentiate customer by customer's behavior. Using RFM analysis, we can do the following several things:

⑴ set up the member pyramid, distinguish each level of members, such as senior members, intermediate members, junior members, and then for different levels of members to implement different marketing strategies, to develop different marketing activities

⑵ found lost and dormant members, through the timely discovery of lost and dormant members, to take marketing campaigns to activate these members.

⑶ in the SMS, EDM promotion, you can use the model, select the best members.

⑷ maintain old customers and improve loyalty of members.

Use method: can give three variables different weights or according to certain rules for grouping, and then combined use, can be divided into many different levels of members.

2. Correlation analysis

The original case of correlation analysis comes from Wal-Mart's "beer and diapers." In the popular sense, is to buy only a commodity people, and a lot of people bought B goods, then we can think of a, b two of goods are relatively high relevance. Many data mining tools have association mining, the main algorithm is the Apriori algorithm, in the process of calculation will mainly examine the item set, confidence, correlation of these three results data, in order to finalize the relationship between the goods. In addition to the Apriori algorithm, there are many other association analysis algorithms, are basically from the Apriori development, such as fpgrowth. I from a few years of experience in data analysis, correlation analysis in the retail industry is not very practical, mining out of the relevance of higher than the goods are generally similar products or the same brand of goods, such as "beer and diapers" this, rarely able to have.

Usage: Group sales or related exhibitions.

3. Cluster analysis

The cluster analysis of retail industry mainly refers to the group segmentation of customers with similar shopping behavior, in order to support fine marketing activities, bring more marketing effect and save cost. The clustering analysis in SPSS mainly has two kinds of k clustering and system clustering. can also be in the data warehouse in accordance with customer purchase properties of the member cluster analysis, where there is no need for algorithm support, only to be based on the system has been the classification of goods, the purchase of the same commodity categories of customers together. This approach may be closer to the company's business. Cluster analysis is a member of fine management, fine Marketing Foundation, do a good job of cluster analysis, enterprises will have great benefits.

4, "The" Word analysis method

This method is mainly to have a very clear membership group, and then through the analysis of these members of the purchase behavior, extraction of these shopping behavior similarities, and then through these similarities to return to the entire data, from which to extract a larger membership group to develop accurate marketing.

Later, the company also on the SAP, and go to the BW Group to do report development, do report development this piece can contact more business knowledge, although do data mining is less, but the data is ultimately to guide business, so this is very beneficial to my growth. The business side has learned a few big chunks:

1, Inventory management-Inventory management This is mainly the management of authentic inventory, unsalable inventory, high inventory of goods, such as various types of inventory how to define and how to manage. For example, to manage the inventory of suppliers will be based on authentic inventory and unsalable inventory and inventory normal turnover days to calculate whether the supplier's inventory is at a reasonable level, whether the purchase or to reduce inventory.

2, promotional management-promotional management is to increase sales for the purpose of attracting and stimulating consumer consumption of a series of planning, organization, leadership, control and coordination of management work. Data is mainly aimed at different promotional methods to calculate different ways of income, different promotional methods can bring different effects, so in the use of promotional time to carefully choose to achieve the desired effect.

In addition, there are financial statements, procurement processes and many other aspects of things, these contacts are less to write.

BW Project team, also often help the site to do some analysis work, they also self-study two books on the Web site data analysis, feeling learned some fur, below say it:

1. Website Traffic Analysis

Site traffic is more important KPI indicators have browse volume, the number of visits, independent visitors, jump rate, conversion rate, page stay time, access page number, traffic source, flow source ROI, etc. This data can be a comprehensive reflection of the overall situation of the site. which can be used to measure the quality of the page, flow source and conversion rate can measure the market and marketing work. Data analysis of the site, the need to firmly grasp the conversion rate of this indicator, and then from the changes in this indicator to find other relevant data changes, and finally find out why, to do the corresponding strategy, improve our work.

2, the website analysis subdivision

Data analysis industry has a sentence-no subdivision, rather death, enough to see the subdivision of the data analysis significance. This is especially true for data analysis of Web sites. The amount of traffic data on the site is very large, from the overall see there will be no problem, so must be subdivided. For example, the conversion rate that marketers need to see, must be subdivided into each channel, and then see these channels to the members of the click, they have seen those pages, what interested in, skipping rate is how much, browsing time, the final transformation of how much, so as to see the problem.

3, the website SMS Promotion and EDM

In this era of electronic commerce generally burning money, how much money can be spent to bring real benefits? At the same time to seize the market, how to achieve the maximum ROI is an urgent need to solve the problem. The company almost every day to send tens of thousands of or even hundreds of thousands of of promotional text messages, SMS feedback rate is basically 2%, how to improve the conversion rate, which requires a more accurate user positioning, the money spent to the most likely to bring benefits to the place. Therefore, the site's short-term promotional and EDM promotions, must be based on the refinement of the members of the subdivision, not only to meet customer needs, but also to dig out their needs.

Write here basically write almost, through summing up to find oneself is very little know, there are many need to learn, such as mathematical modeling knowledge is not enough, statistics software is not good enough, business understand not deep, the whole e-commerce industry development grasp is not clear, These are the places that need to be strengthened later. Recently, on the blog of a senior data analyst, he has a little bit of a need for data analysts to love data. Feel that they are not enough, peacetime work is not enough to invest, always feel that is working for the company, not in the interest of their work, in fact, a person to do every day, we must all as a do for themselves, even if it is not for their own to do, but also to learn from a number of things to become their own things, for their own services.