S13 · stress test simulation

Stress-test a parcel delay model before peak season

People, data and outputs in this scenario are simulated. Each stage is recorded when you commit; the reveal comes after.

4stages
35 minabout
0stages committed

The brief

A logistics company (simulated) scores every in-transit parcel each hour with a gradient-boosted classifier that estimates the probability of missing its promised delivery date. The top 10% of scores go to a recovery team that can reroute about 4,000 parcels a day. All data and results in this scenario are synthetic.

Holdout performance (temporal split, last eight weeks): recall in the top 10% of scores 0.71, precision 0.38.

Features include parcel weight, service level, origin hub code, scan events in the last 6 hours, hub daily volume, and weather alerts on the route. Peak season starts in five weeks. The operations director (simulated) wants a go/no-go recommendation.

Stage 1 of 4

You have two days to build stress tests. The pipeline can replay any historical day with perturbed inputs into a staging copy of the scorer; nothing reaches production.

Which tests belong in the first-pass plan? Select all that apply.

Replay days with the scan feed delayed by 3 hours, so recent-scan features arrive empty.Score parcels from a new hub code that did not exist in training.Scale hub daily volume to 2.5 times the training maximum to mimic peak.Inject zero, negative and missing parcel weights.Retune hyperparameters on the eight-week holdout until recall rises.