SCD Type 3 Implementation using Informatica PowerCenter

  ETL Design , Mapping Tips

Unlike SCD Type 2, Slowly Changing Dimension Type 3 preserves only few history versions of data, most of the time 'Current' and Previous' versions. The 'Previous' version value will be stored into the  additional columns with in the same dimension record.  In this article lets discuss the step by step implementation of SCD Type 3 using Informatica PowerCenter.

The number of records we store in SCD Type 3 do not increase exponentially as we do not insert a record for each and every historical record. Hence we may not need the performance improvement techniques used in the SCD Type 2 Tutorial.

Understand the Staging and Dimension Table.

Slowly Changing Dimension Series
Part I : SCD Type 1.
Part II : SCD Type 2.
Part III : SCD Type 3.
Part IV : SCD Type 4.
Part V : SCD Type 6.

For our demonstration purpose, lets consider the CUSTOMER Dimension. Here we will keep previous version of CITY, STATE, ZIP into its corresponding PREV columns. Below are the detailed structure of both staging and dimension table.

Staging Table

In our staging table, we have all the columns required for the dimension table attributes. So no other tables other than Dimension table will be involved in the mapping. Below is the structure of our staging table.

Key Points 

1. Staging table will have only one days data. Change Data Capture is not in scope.
2. Data is uniquely identified using CUST_ID.
3. All attribute required by Dimension Table is available in the staging table

Dimension Table

Here is the structure of our Dimension table.

Key Points

1. CUST_KEY is the surrogate key.
2. CUST_ID is the Natural key, hence the unique record identifier.
3. Previous versions are kept in PREV_CITY, PREV_STATE, PREV_ZIP columns.

Mapping Building and Configuration

Step 1

Lets start the mapping building process. For that pull the CUST_STAGE source definition into the mapping designer.

Step 2

Now using a LookUp Transformation fetch the existing Customer columns from the dimension table T_DIM_CUST. This lookup will give NULL values if the customer is not already existing in the Dimension tables. 

• LookUp Condition : IN_CUST_ID = CUST_ID
• Return Columns : CUST_KEY, CITY, STATE, ZIP

Step 3

Using an Expression Transformation, identify the records for Insert and Update using below expression. Additionally, map the columns from the LookUp Transformation to the Expression as shown below.  With this we get both  the previous and current values of the CUST_ID.

• INS_UPD :- IIF(ISNULL(CUST_KEY),'INS', IIF(CITY <> PREV_CITY OR STATE <> PREV_STATE OR ZIP <> PREV_ZIP, 'UPD'))   

Additionally create two output ports.

• CREATE_DT :- SYSDATE
• UPDATE_DT :- SYSDATE

Note : If in case there are too many columns to be compared  to build the INS_UPD logic, make use of CheckSum Number (MD5() Function) to make it simple.       

Step 4

Map the columns from the Expression Transformation to a Router Transformation and create two groups (INSERT, UPDATE) in Router Transformation using the below expression. The mapping will look like shown in the image.

• INSERT :- IIF(INS_UPD='INS',TRUE,FALSE)
• UPDATE :- IIF(INS_UPD='UPD',TRUE,FALSE)

INSERT Group

Step 5
Every records coming through the 'INSERT Group' will be inserted into the Dimension table T_DIM_CUST.

Use a Sequence generator transformation to generate surrogate key CUST_KEY as shown in below image. And map the columns from the Router Transformation to the target. Leave all 'PREV' columns unmapped as shown below image.

Note : Update Strategy is not required, if the records are set for Insert.

UPDATE Group

Step 6
Records coming from the 'UPDATE Group' will update the customer Dimension with Current customer attributes and the 'PREV' attributes. Add an Update Strategy Transformation before the target instance and set it as DD_UPDATE. Below is the structure of the mapping.

We are done with the mapping building and below is the structure of the completed mapping.

Workflow and Session Creation

There is not any specific properties required to be given during the session configuration. 

Below is a sample data set taken from the Dimension table T_DIM_CUST. See the highlighted values.

Hope you guys enjoyed this. Please leave us a comment in case you have any questions of difficulties implementing this.

Share with your friends