Slowly changing dimension

Dimension is a term in data management and data warehousing that refers to logical groupings of data such as geographical location, customer information, or product information. Slowly Changing Dimensions (SCDs) are dimensions that have data that changes slowly, rather than changing on a time-based, regular schedule.

For example, you may have a dimension in your database that tracks the sales records of your company's salespeople. Creating sales reports seems simple enough, until a salesperson is transferred from one regional office to another. How do you record such a change in your sales dimension?

You could sum or average the sales by salesperson, but if you use that to compare the performance of salesmen, that might give misleading information. If the salesperson that was transferred used to work in a hot market where sales were easy, and now works in a market where sales are infrequent, her totals will look much stronger than the other salespeople in her new region, even if they are just as good. Or you could create a second salesperson record and treat the transferred person as a new sales person, but that creates problems also.

Dealing with these issues involves SCD management methodologies referred to as Type 0 through 6. Type 6 SCDs are also sometimes called Hybrid SCDs.

Type 0

The Type 0 method is a passive approach to managing dimension value changes, in which no action is taken. Values remain as they were at the time of the dimension record was first entered. In certain circumstances historical preservation with a Type 0 SCD may occur. But, higher order SCD types are often employed to guarantee history preservation, whereas Type 0 provides the least control or no control over managing a slowly changing dimension.

The most common slowly changing dimensions are Types 1, 2, and 3.

Type 1

The Type 1 methodology overwrites old data with new data, and therefore does not track historical data at all. This is most appropriate when correcting certain types of data errors, such as the spelling of a name. (Assuming you won't ever need to know how it used to be misspelled in the past.)

Here is an example of a database table that keeps supplier information:

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State
123	ABC	Acme Supply Co	CA

In this example, Supplier_Code is the natural key and Supplier_Key is a surrogate key. Technically, the surrogate key is not necessary, since the table will be unique by the natural key (Supplier_Code). However, the joins will perform better on an integer than on a character string.

Now imagine that this supplier moves their headquarters to Illinois. The updated table would simply overwrite this record:

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State
123	ABC	Acme Supply Co	IL

The obvious disadvantage to this method of managing SCDs is that there is no historical record kept in the data warehouse. You can't tell if your suppliers are tending to move to the Midwest, for example. But an advantage to Type 1 SCDs is that they are very easy to maintain.

If you have calculated an aggregate table summarizing facts by state, it will need to be recalculated when the Supplier_State is changed.

Type 2

The Type 2 method tracks historical data by creating multiple records for a given natural key in the dimensional tables with separate surrogate keys and/or different version numbers. With Type 2, we have unlimited history preservation as a new record is inserted each time a change is made.

In the same example, if the supplier moves to Illinois, the table could look like this, with incremented version numbers to indicate the sequence of changes:

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State	Version
123	ABC	Acme Supply Co	CA	0
124	ABC	Acme Supply Co	IL	1

Another popular method for tuple versioning is to add 'effective date' columns.

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State	Start_Date	End_Date
123	ABC	Acme Supply Co	CA	01-Jan-2000	21-Dec-2004
124	ABC	Acme Supply Co	IL	22-Dec-2004

The null End_Date in row two indicates the current tuple version. In some cases, a standardized surrogate high date (e.g. 9999-12-31) may be used as an end date, so that the field can be included in an index, and so that null-value substitution is not required when querying.

Transactions that reference a particular surrogate key (Supplier_Key) are then permanently bound to the time slices defined by that row of the slowly changing dimension table. An aggregate table summarizing facts by state continues to reflect the historical state, i.e. the state the supplier was in at the time of the transaction; no update is needed.

If there are retrospective changes made to the contents of the dimension, or if new attributes are added to the dimension (for example a Sales_Rep column) which have different effective dates from those already defined, then this can result in the existing transactions needing to be updated to reflect the new situation. This can be an expensive database operation, so Type 2 SCDs are not a good choice if the dimensional model is subject to change.

Type 3

The Type 3 method tracks changes using separate columns. Whereas Type 2 had unlimited history preservation, Type 3 has limited history preservation, as it's limited to the number of columns designated for storing historical data. Where the original table structure in Type 1 and Type 2 was very similar, Type 3 adds additional columns to the tables. In the following example, an additional column has been added to the table so as to record the supplier's original state: (only the previous history is stored )

Supplier_Key	Supplier_Code	Supplier_Name	Original_Supplier_State	Effective_Date	Current_Supplier_State
123	ABC	Acme Supply Co	CA	22-Dec-2004	IL

Note that this record--having only a column for the original state and a column for the current state--can not track all historical changes, such as when a supplier moves a second time.

One variation of this type is to create the field Previous_Supplier_State instead of Original_Supplier_State which would then track only the most recent historical change.

Type 4

The Type 4 method is usually referred to as using "history tables", where one table keeps the current data, and an additional table is used to keep a record of some or all changes.

Following the example above, the original table might be called Supplier and the history table might be called Supplier_History.

Supplier

Supplier_key	Supplier_Code	Supplier_Name	Supplier_State
123	ABC	Acme Supply Co	IL

Supplier_History

Supplier_key	Supplier_Code	Supplier_Name	Supplier_State	Create_Date
123	ABC	Acme Supply Co	CA	22-Dec-2004

This method resembles how database audit tables and change data capture techniques function.

Type 6 / Hybrid

The Type 6 method combines the approaches of types 1, 2 and 3 (1 + 2 + 3 = 6). One possible explanation of the origin of the term was that it was coined by Ralph Kimball during a conversation with Stephen Pace from Kalido^{[citation needed]}. Ralph Kimball calls this method "Unpredictable Changes with Single-Version Overlay" in The Data Warehouse Toolkit.

The Supplier table starts out with one record for our example supplier:

Supplier_Key	Supplier_Code	Supplier_Name	Current_State	Historical_State	Start_Date	End_Date	Current_Flag
123	ABC	Acme Supply Co	CA	CA	01-Jan-2000	31-Dec-9999	Y

The Current_State and the Historical_State are the same. The Current_Flag attribute indicates that this is the current or most recent record for this supplier.

When Acme Supply Company moves to Illinois, we add a new record, as in Type 2 processing:

Supplier_Key	Supplier_Code	Supplier_Name	Current_State	Historical_State	Start_Date	End_Date	Current_Flag
123	ABC	Acme Supply Co	IL	CA	01-Jan-2000	21-Dec-2004	N
124	ABC	Acme Supply Co	IL	IL	22-Dec-2004	31-Dec-9999	Y

We overwrite the Current_State information in the first record (Supplier_Key = 123) with the new information, as in Type 1 processing. We create a new record to track the changes, as in Type 2 processing. And we store the history in a second State column (Historical_State), which incorporates Type 3 processing.

If our example supplier company were to relocate again, we would add another record to the Supplier dimension, and we would once again overwrite the contents of the Current_State column:

Supplier_Key	Supplier_Code	Supplier_Name	Current_State	Historical_State	Start_Date	End_Date	Current_Flag
123	ABC	Acme Supply Co	NY	CA	01-Jan-2000	21-Dec-2004	N
124	ABC	Acme Supply Co	NY	IL	22-Dec-2004	03-Feb-2008	N
125	ABC	Acme Supply Co	NY	NY	04-Feb-2008	31-Dec-9999	Y

Note that, for the current record (Current_Flag = 'Y'), the Current_State and the Historical_State are always the same.

Type 2 / Type 6 Fact Implementation

Type 2 Surrogate Key With Type 3 Attribute

In many Type 2 and Type 6 SCD implementations, the surrogate key from the dimension is put into the fact table in place of the natural key when the fact data is loaded into the data repository. The surrogate key is selected for a given fact record based on its effective date and the Start_Date and End_Date from the dimension table. This allows the fact data to be easily joined to the correct dimension data for the corresponding effective date.

Here is the Supplier table as we created it above using Type 6 Hybrid methodology:

Supplier_Key	Supplier_Code	Supplier_Name	Current_State	Historical_State	Start_Date	End_Date	Current_Flag
123	ABC	Acme Supply Co	NY	CA	01-Jan-2000	21-Dec-2004	N
124	ABC	Acme Supply Co	NY	IL	22-Dec-2004	03-Feb-2008	N
125	ABC	Acme Supply Co	NY	NY	04-Feb-2008	31-Dec-9999	Y

Once the Delivery table contains the correct Supplier_Key, it can easily be joined to the Supplier table using that key. The following SQL retrieves, for each fact record, the current supplier state and the state the supplier was located in at the time of the delivery:

SELECT
  delivery.delivery_cost,
  supplier.supplier_name,
  supplier.historical_state,
  supplier.current_state
FROM delivery
INNER JOIN supplier 
  ON delivery.supplier_key = supplier.supplier_key

Pure Type 6 Implementation

Having a Type 2 surrogate key for each time slice can cause problems if the dimension is subject to change.

A pure Type 6 implementation does not use this, but uses a Surrogate Key for each master data item (e.g. each unique supplier has a single surrogate key).

This avoids any changes in the master data having an impact on the existing transction data.

It also allows more options when querying the transactions.

Here is the Supplier table using the pure Type 6 methodology:

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State	Start_Date	End_Date
456	ABC	Acme Supply Co	CA	01-Jan-2000	21-Dec-2004
456	ABC	Acme Supply Co	IL	22-Dec-2004	03-Feb-2008
456	ABC	Acme Supply Co	NY	04-Feb-2008	31-Dec-9999

The following example shows how the query must be extended to ensure a single supplier record is retrieved for each transaction.

SELECT
  supplier.supplier_code, 
  supplier.supplier_state
FROM supplier
INNER JOIN delivery
  ON supplier.supplier_key = delivery.supplier_key
 AND delivery.delivery_date >= supplier.start_date
 AND delivery.delivery_date <= supplier.end_date

A fact record with an effective date (Delivery_Date) of August 9, 2001 will be linked to Supplier_Code of ABC, with a Supplier_State of 'CA'. A fact record with an effective date of October 11, 2007 will also be linked to the same Supplier_Code ABC, but with a Supplier_State of 'IL'.

Whilst more complex, there are a number of advantages of this approach, including:

1. If there is more than one date on the fact (e.g. Order Date, Delivery Date, Invoice Payment Date) you can choose which date to use for a query.

2. You can do "as at now", "as at transaction time" or "as at a point in time" queries by changing the date filter logic.

3. You don't need to reprocess the Fact table if there is a change in the dimension table (e.g. adding additional fields retrospectively which change the time slices, or if you make a mistake in the dates on the dimension table you can correct them easily).

4. You can introduce bi-temporal dates in the dimension table.

5. You can join the fact to the multiple versions of the dimension table to allow reporting of the same information with different effective dates, in the same query.

The following example shows how a specific date such as '2012-01-01 00:00:00' (which could be the current datetime) can be used.

SELECT
  supplier.supplier_code, 
  supplier.supplier_state
FROM supplier
INNER JOIN delivery
  ON supplier.supplier_key = delivery.supplier_key
 AND delivery.delivery_date >= '2012-01-01 00:00:00'
 AND delivery.delivery_date <= '2012-01-01 00:00:00'

Both Surrogate and Natural Key

An alternative implementation is to place both the surrogate key and the natural key into the fact table. This allows the user to select the appropriate dimension records based on:

• the primary effective date on the fact record (above),
• the most recent or current information,
• any other date associated with the fact record.

This method allows more flexible links to the dimension, even if you have used the Type 2 approach instead of Type 6.

Here is the Supplier table as we might have created it using Type 2 methodology:

Supplier_Key	Supplier_Code	Supplier_Name	Supplier_State	Start_Date	End_Date	Current_Flag
123	ABC	Acme Supply Co	CA	01-Jan-2000	21-Dec-2004	N
124	ABC	Acme Supply Co	IL	22-Dec-2004	03-Feb-2008	N
125	ABC	Acme Supply Co	NY	04-Feb-2008	31-Dec-9999	Y

The following SQL retrieves the most current Supplier_Name and Supplier_State for each fact record:

SELECT
  delivery.delivery_cost,
  supplier.supplier_name,
  supplier.supplier_state
FROM delivery
INNER JOIN supplier 
  ON delivery.supplier_code = supplier.supplier_code
WHERE supplier.current_flag = 'Y'

If there are multiple dates on the fact record, the fact can be joined to the dimension using another date instead of the primary effective date. For instance, the Delivery table might have a primary effective date of Delivery_Date, but might also have an Order_Date associated with each record.

The following SQL retrieves the correct Supplier_Name and Supplier_State for each fact record based on the Order_Date:

SELECT
  delivery.delivery_cost,
  supplier.supplier_name,
  supplier.supplier_state
FROM delivery
INNER JOIN supplier 
  ON delivery.supplier_code = supplier.supplier_code
 AND delivery.order_date >= supplier.start_date
 AND delivery.order_date <= supplier.end_date

Some cautions:

• If the join query is not written correctly, it may return duplicate rows and/or give incorrect answers.

• The date comparison might not perform well.

• Some Business Intelligence tools do not handle generating complex joins well.

• The ETL processes needed to create the dimension table needs to be carefully designed to ensure that there are no overlaps in the time periods for each distinct item of reference data.

Combining Types

Different SCD Types can be applied to different columns of a table. For example, we can apply Type 1 to the Supplier_Name column and Type 2 to the Supplier_State column of the same table, the Supplier table.

See also

• Change data capture
• Log trigger

References

• Bruce Ottmann, Chris Angus: Data processing system, US Patent Office, Patent Number 7,003,504. February 21, 2006

• Ralph Kimball:Kimball University: Handling Arbitrary Restatements of History [1]. December 9, 2007