Introduced on day 19
Training-serving skew
A difference between the feature values a model saw in training and those it receives in serving for the same underlying facts, caused by differing code, data or timing.
In detail
Typical causes are two implementations of one feature, a joined table that changes between training and serving, different missing-value handling and stale tables. It is present from the first request, unlike drift, which develops. Prevent it with one shared feature definition and by logging served features to compare with training.