Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Queries on Customer Analytics Solution (Segmentation and Cross Sell Model)

Posts: 0

Queries on Customer Analytics Solution (Segmentation and Cross Sell Model)


We’re from one of the Foreign Bank in Southeast Asia. We have actually purchased SAS product for Data Warehouse and Customer Analytics Solution (Segmentation and Cross Sell Model). Now, we have embarked to Customer Analytics project, and we’ve few questions as below:

1) Understand that the mart is design for a monthly load instead of daily load. What are the reasons, the mart is being design this way?

2) Will this N-1 month data (N is current month) caused wrongly targeted customer for marketing campaigns? Example: We’ve wrongly targeted to those customer who is no longer with the bank at current month. How can we address this issue?

3) For modeling purposes, do we look the data as at position (30Apr2010) or on monthly basis (1Apr2010-30Apr2010)? And why?

On the other hand, we're very sorry if we've posted in an inappropriate thread in this forum, appreciate if you could guide us where it should be posted in this forum. But hopefully someone can help us to answer this queries that we had.

We’re much looking forward for your prompt response. Thank you.
Occasional Contributor
Posts: 19

Re: Queries on Customer Analytics Solution (Segmentation and Cross Sell Model)

Hi Dia, you will probably get about a million responses to your questions so I said to myself "why not, I'll be one of them."Smiley Happy My answers below are verrrrry brief.
1. Many data warehouses or data marts start out in their infancy as monthly snapshots in time. They typically included summarized (totaled) data, such as total ATM withdrawals, average daily balance, and status codes on an account for the entire previous month. Moving to a weekly or even daily load takes quite a while for IT and database people to set that up and it requires much more storage space for the data.
2. Typically promotions (direct mail drops) occurr a number of weeks beyond the time the mail house gets the data. By that time the data could be 1,2 or even 3 months old. I suggest you send a "purge file" of customers with closed accounts, delinquencies, opt outs, new employees, or those with any other undesirable behavior (and those that have opened an account since the data was pulled) at the last minute so your mail house can remove them from the file. This should be a common practice.
3. When building models you typically need to pull historical data from before the customers opened the accounts, ie., an EquityLine, or did some behavior you're trying to model, like signing up for eStatements. This may require you to go 3 to 4 months back from the current time period. You want the data to reflect the customers as they were before they responded to the promotion. Month-end snapshots would work fine for predictive modeling.
Hope this helps!
Ask a Question
Discussion stats
  • 1 reply
  • 2 in conversation