Determining
when a process is working properly can be a difficult
task. Unlike a manufacturing plant where a product can
be sampled for defects, defective underwriting is typically
realized only in account performance, which may not
manifest itself for months or years. The cost of poor
process control can be large.
To address these issues, we developed a sampling method
to directly assess the effects of variance in a process.
Specific process steps are identified and repeated samples
are drawn from historical data. The distribution in
key metrics is compared to actual account performance
and process tolerances are established. Output from
a live process is compared to the historical distribution
of key metrics. Performance outside the sampled tolerances
generates a warning. Distributions and tolerances are
easily updated when new historical data become available.
|
| |
|
