3.1 Characteristic analysis of Target customers
The first step in a data-based operation (the most basic step) is to find your target, target audience, and then the appropriate operational solutions, personalized products and services.
In the typical characteristic analysis of the target customer, the business scenario can be the virtual feature exploration before the trial operation, or it can be the analysis, mining and refining based on the real operation data after the test operation, the goal is the same, but the thinking is different, the data source is different, and the analysis technology has a certain difference.
3.2 Forecast (response, classification) model for Target customers (core is response probability)
The prediction (response, classification) model includes the loss early warning model, the paid forecast model, the renewal forecast model, the operational activity response model, etc.
Based on the size of the actual response scale in the modeling data, the response ratio is less than 1% as a rare incident response model, and the other as the general response model
3.3 Activity definition of the operating group
Activity definition two basic points:
- The composition indicator of activity should be the most important behavioral factor in the business scenario.
- The important judgment of whether the definition of activity is appropriate or not is based on its ability to respond effectively to the ultimate goal of business needs.
The definition of activity has two major statistical techniques:
- Principal component analysis: Bar multiple core behavioral indicators into one or a few major components, and ultimately into a comprehensive score, as the definition of activity (the characteristic root of principal component analysis and the cumulative variance contribution rate)
- Standardization of data
3.4 User Path Analysis
There are two types of analysis techniques commonly used in path analysis: One is algorithm-supported, and the other is to traverse the main path strictly in step order
3.5 Cross-Sell model
- By association Technology (Shopping basket analysis)
- Use the response model to establish a predictive model for important commodities, filter the potential consumers through these specific predictive models, and then target the most likely top 5% consumers for accurate marketing campaigns.
- Draw on the predictive response model and get the key product 22 portfolio to identify potential customers who are most likely to consume
- Discover specific rules based on specific data resources through tree-like rules with clear decision trees
The corresponding modeling techniques mainly include correlation analysis, sequence analysis, i.e., on the basis of association analysis, the consideration of increasing sequencing, and the prediction (response, classification) model technology, etc. 3.6 information quality model
.....
3.7 Service Guarantee Model
.....
3.8 User (buyer, seller) tiered model 3.9 seller (buyer) trading Model 3.10 Credit Risk Model 3.11 Product Recommendation Model 3.12 Data Product 3.13 Decision Support
Chapter III: Common Data Analysis project types in data operations