Case Study: Big Data ETL

by Pete Carapetyan


Placed as a founding lead on a mission critical, 'first of' big data project for a Fortune 500 corporation, my role was to help interview and select the rest of the 10 person team, help build it, and then lead a portion of the work through to the end of the first deliverable 1 year later. The project was to integrate with a third party vendor, complete series of ETL tasks, and deliver complex analytics through a lightning fast API.


Contractual challenges emerged between the primary third party platform vendor and the purchasing customer. During the startup, the parties failed to consumate a satisfactory marriage. Much organizational tumult ensued. My role was downgraded to developer. Project Manager, Business Analyst, and Project Architect were engaged in an unplanned attempt at a fix.


Rudimentary development tasks were interspersed with code cleanup, introduction of some relatively insignificant testing regimens, and creation of automated ETL code via code generation. The resulting friction made the team uncomfortable, so once the project stabilized one year into the work, I removed myself.


Sometimes 'fit' is just personality or turf driven. Other times, it is important not to introduce systems, design patterns, and metadata driven code generation into an environment where they are threatening to team leads who have other values. We are very small shop, so the most important lesson we can learn is to match our customers values with our own best skills, even when we lose a long term customer as a result.