Data warehouse model fact table model design

Fact table ： Data detail layer , take ODS Layer data , Conduct ETL after , Mild polymerization , Expand details ！ Then deposit DWD layer .

1、 When expanding details , Expand some dimensions . Such as adding location information , It mainly extends dimension information .

2、 Refer to the modeling strategy of star model , For business processes , Particle size , dimension , The facts . Use 3W Principles .

who: Who did

where: Where did you do it

what: What they did

Type of fact table ：

When designing the fact table , There are actually many different types , Of course, every company has different design methods . We include 3 Kind of .

1. Transactional fact table ： The data in the data warehouse remains unique . Data is generated after a transaction event , The granularity of data is usually one record per transaction record . Once the transaction is committed , Fact table data is inserted , The data will not be changed , Its update mode is incremental update . This fact shows that we only focus on incremental processing .

2. Snapshot fact table ： Associate the full data of yesterday's partition with the detailed dimension .（ Default ODS Partition full data by day ）

3. Cumulative fact table ： The fact is that , Usually, a transaction has only one primary key , The modification of transactions is based on the original data . The processing of this fact table . Although we can also handle incrementally , But it is the partition that needs dynamic partition to cover history , To make the transaction primary key unique .

Order details --- Transactional snapshot fact table （ODS Yesterday was full join）
Payment fact sheet --- Transactional snapshot fact table
Refund fact sheet --- Transactional snapshot fact table
Evaluate facts quickly --- Transactional snapshot fact table , use 3W principle
                 Transactional fact table
Shopping cart fact table --- Periodic snapshot fact table , Full amount of synchronization , So generally, only 7 God

Coupon fact sheet --- Cumulative snapshot fact table , Take the starting time of statistical facts as the partition field . The same coverage .