The model builds every six hours to
determine the response rates for each profile attribute.
However, the system must be able to respond to faster
changes in response rates and modify its responses within those six-hour
windows. Quick changes like news announcements about interest rates, sales,
television advertisements, and other things can almost instantly affect the
traffic to an automated decisioning campaign. The campaign accounts for this by
tracking the response rate for each campaign option compared to the prediction
from the model. The real-time response rate then influences the model score.
This means that, for example, a content option with a high model score but a
lower real-time response rate has less chance of being selected than if the
system only used the model score.