Is there always a segmentation engine feeding SFMC

Question

In profiling a dozen or so companies that have implemented SFMC for orchestrating marketing campaigns or other communications, there seems to be a pattern in that SFMC seems to be using pre-segmented data even though it is capable of creating segments and target audiences using tools like MC's own audience builder, SQL, recency/frequency/monetary filters to group contacts etc.

The pattern I refer to is where the segmentation (grouping of target customers) seems to be done before-hand within say, a CDP or sometimes, in a segmentation platform like Unica or epiphany, and then brought into SFMC in batch SFTP or API and stored in data extensions. SFMC then uses its capabilities in Email Studio, Journey Builder, Content Builder to segment, design, preview and communicate/fulfils the messaging via activation channel.

QUESTION : I wanted to check if this observation holds true always ? Or are there instances have, where SFMC actually also stores and models all of its customer, transactional, analytical data in its contact model and then uses Automation Studio, Email Studio, Journey Builder to direct personalized and targeted outbound communications via activation channels?

Answer 1

Short answer:

Yes, there are times that SFMC does modeling and all ETL on data, but at any large volume, frequency or complexity - it will stumble if not fall.

Longer Answer:

SFMC was designed with the intention of being a messaging and marketing tool, not a data tool. So therefore a lot of the hosting, infrastructure, etc are designed and optimized for that and not for heavy ETL or data handling. Although SFMC has some awesome ETL features and capabilities, think of these as add on capabilities not main purpose. For example, SFMC offers SQL - but limits it to just SELECT statements and sets a 30 minute timeout window. This is not because SFMC is trying to be mean, but to ensure that the server is optimized for its main purpose of sending messages. Using ETL and other data modeling causes huge draws on the stacks and can slow it to a crawl, which then causes SFMC to throttle or end your activities- which is why at any sort of volume, frequency or complexity that could create this draw - it will fail or at the least slow to a crawl, causing major issues to your projects, etc.