Steamed Bread Business School--1, using good data to do fine operation

1. The overall instrument panel is sufficient for rough operation.

2, the number of users dropped, there are several dimensions can be considered: payment method problems, user channels may be blocked, geographical reasons, market reasons.

Data silos:

Through the various business units data: 1, Customer Operations Department: CRM data, customer service Department: Customer service System data, User Operations department: User behavior data, Sales Department: Order system data.

The data pyramid is divided into three tiers:

1, data acquisition: Each data buried 2, modeling: the establishment of easy-to-use table daily rolling update 3, Data analysis: Analysis of data based on business and scenarios.

first, data collection : ( full: A variety of data sources, full-volume data rather than sampling, fine: who,when,where,how,what, many conditions occur. )

there may be problems with :

1, inaccurate: such as network anomalies, or equipment anomalies, statistical caliber 2, incomplete: such as only the acquisition of the client, the back end did not collect. 3, not meticulous: for example, some dimension information is not recorded, different browser version of the user source.

Data Source:

1, visualization of the buried point

Install SDK visualization in Andriod,ios, @1, analyze Pv,uv, and click on the basic indicators. @2, activity/new features quick on-line iteration of the effect evaluation. Disadvantages :

2, code buried point

Client, the service side of the SDK for data collection, recording the corresponding process data: @1, refined analysis of the core transformation process. @2, analysis of different channels and different promotion methods of the effect of delivery.

3. Import Auxiliary Tools

Back-end logging, there are some offline data, can not be collected by the client, with data tools:

Data Modeling : ( organizing data to facilitate the use of data: for example, there are many tables in the database )

Need to build a Data Warehouse : Data that stores user behavior.

Data dimensions: user_id, time, events event device model, user area.

OLAP: Multidimensional Analysis Indicators ( dimensions, metrics ) online processing simply say it. Data extraction for multiple databases, after cleaning. form data cube storage.
Then use OLAP to process the results ...
Cubic is not a multi-dimensional ...

Third, data analysis-Multidimensional time analysis

@1, Events: Place order, cancel orders, register.

@2, Dimension: The name of the user Ah age what

@3, Indicator: is the number.

Steamed Bread Business School--1, using good data to do fine operation