At Mellangrove Borough Council, inputs about a household — arrears history, household details and a tenancy record that is silently missing — drop into a closed machine nobody can see inside, which prints a confident priority score of 61 that really means somewhere between 56 and 66. Beside it, the threshold at 60 sorts two nearly identical households, scored 59.6 and 60.2, into different weeks; a score of 34 closed a case without discussion; and the fix is that borderline cases go to a person, because the model advises and people decide.

CY-BEGINNER-057Beginner · Model Risk Management

A Number You Did Not Make — Relying On What A Model Tells You

A score a machine produces, a decision about a real person, and a box you cannot see inside. What the number is and is not, when it has quietly stopped being right, and why the human still has to decide.

11 hours8 chapters32 lessons225 activities8 labs

What you'll learn

Say what a model is in one plain sentence, tell a precise number from an accurate one, and read a score as a band.
Find out what a score was built for, refuse to use it outside its purpose, and ask how old it is and when it was last checked.
Spot the silent wrong number a bad input produces, and stop a score from ending the conversation it should start.
Keep the deciding and the accountability with a person, and run the four checks that keep a relied-on model honest.

Price:৳5,000