What is point-in-time correctness, and how does violating it look in practice?
hardAnswer
- It means every feature value used for a training example reflects only what was knowable at that example's timestamp.
- Violating it produces leakage that is invisible offline: the model looks excellent in validation and disappoints in production, because at serving time the future information it relied on does not exist yet.
- The classic mechanisms are joining against a table that stores only the current value of an attribute, aggregating over a window that extends past the prediction time, and backfilling a column after an outcome was known.
- Detection is mostly structural, by reviewing every join for whether it filters on event time, and the pragmatic test is a suspiciously strong feature: an importance ranking dominated by one column usually means leakage rather than luck.
Check yourself — multiple choice
- It refers to clock synchronization
- Features must reflect only what was knowable at the example's timestamp; violating it leaks future information, giving great offline scores and poor production results
- It only matters for streaming
- It is about time zones
Point-in-time joins prevent the leakage that makes offline metrics unrepresentative.
#features#data-quality#pipeline
Practise MLOps & Data Quality
215 interview questions in this topic.
Related questions
- How do you prove there is no training-serving skew?
- What are common sources of data leakage in an ML pipeline?
- How do you handle missing values in a production pipeline?
- How does imputation cause data leakage?
- How do you handle high-cardinality categorical features?
- How does time-based leakage happen and how do you prevent it?