⚙️Training-Serving Skew
Also known as Train-Serve Skew, Training/Serving Skew, Online-Offline Skew
Live data and training data are made in two ways. So a good model does badly.
Think of it like
It is like practising football on grass. Then the real game is played on ice.
Example
Emma's model was trained on ages in years. The live app sends ages in months, so every customer looks 12 times older and the scores go wrong.
Why it matters
Nothing crashes, so it can hide for months. A model that tested well quietly gives poor answers.
Definition
Common mistake
Key takeaways
- 1Same input, two ways. Training and live code do not match.
- 2Silent harm. No error, just worse answers.
- 3Share the code. One feature step for training and serving.