InsideReferenceData

#ToFIS Toronto Financial Information Summit Tuesday, June 7. Still time to register here http://bit.ly/1GkHV2a 

#nafis Q how to avoid redundant models? Impose a tax: separate reviews, monitoring for each one. Making a suite of models simplifies

#nafis pulling data for customer facing, must have several checks. Also true for reporting

#nafis Q accountability for good data? Better know what it is before you start using it.

#nafis have the 400 models all monitored right in front of me was a challenge

#nafis make sure you have right disclosed issues. Not getting away from owner

#nafis solving small data issues? Have feedback loop

#nafis scale across variety of environments forced us to develop our own, w guardrails

#nafis tools in Hadoop and data lakes aren't mature enough (for our needs)

#nafis on average models have 500 elements w 4 or 5 items each. ...

#nafis but aask customer if it's valid by text message. That's 100% right all the time. More effective. So what's value of complex model?

#nafis that's AML. ... how do you learn about fraud and cancel transaction? Our modeling group spent months on early detection model

#nafis can better identify digital interactions & understand customer journeys. Bridge all accounts a customer has

#nafis if you understand data lineage, can create value. How to mine complaint calls to pre identify problems?

#nafis people are creating new data. ... need to get into data life cycle when you go from batch to real time

#nafis open source challenge is no version controls. No readily available audit trails

#nafis caution: open source tools. So many more than the 9 or so available from 2005 to 2010