EXAMPLE - PREV Function

This example describes how you can use the PREV function to analyze data that is available in a window in rows before the current one.

Functions:

The following dataset contains orders for multiple customers over a period of a few days, listed in no particular order. You want to assess how order size has changed for each customer over time and to provide offers to your customers based on changes in order volume.

Source:

Transformation:

When the data is loaded into the Transformer page, you can use the PREV function to gather the order values for the previous two orders into a new column. The trick is to order the window transform by the date and group it by customer:

PREV(OrderValue, 1)

CustId

Date

PREV(OrderValue, 2)

CustId

Date

Manual rename

window

'OrderValue_1'

Manual rename

window1

'OrderValue_2'

You should now have the following columns in your dataset: Date, CustId, OrderId, OrderValue, OrderValue_1, OrderValue_2.

The two new columns represent the previous order and the order before that, respectively. Now, each row contains the current order (OrderValue) as well as the previous orders. Now, you want to take the following customer actions:

To address the first one, you might add the following, which uses the IF function to test the value of the current order compared to the previous ones:

Single row formula

IF(OrderValue >= (1.2 * (OrderValue_1 + OrderValue_2)), 'send rebate', 'no action')

'CustomerAction'

You can now see which customers are due a rebate. Now, edit the above and replace it with the following, which addresses the second condition. If neither condition is valid, then the result is send holiday card.

Single row formula

IF(OrderValue >= (1.2 * (OrderValue_1 + OrderValue_2)), 'send rebate', IF(OrderValue <= (1.2 * (OrderValue_1 + OrderValue_2)), 'send coupon', 'send holiday card'))

'CustomerAction'

Results:

After you delete the OrderValue_1 and OrderValue_2 columns, your dataset should look like the following. Since the transformations with PREV functions grouped by CustId, the order of records has changed.