S02 · code review simulation

Review a nightly no-show scoring script before it ships

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 synthetic airline (simulated company) scores every booking departing in the next three days for no-show risk. A colleague (simulated) opens a pull request: "Nightly no-show scoring v2: faster feature handling, adds risk bands for the call team. Output goes to the reconfirmation queue." The data are synthetic.

import pickle
import pandas as pd

FEATURES = ["days_to_departure", "fare_paid", "party_size", "is_member"]

def score_batch(model_path, bookings: pd.DataFrame) -> pd.DataFrame:
    model = pickle.load(open(model_path, "rb"))          # model_path comes from the job config
    bookings = bookings.fillna(0)
    risks = []
    for _, row in bookings.iterrows():
        x = row[FEATURES].to_frame().T
        risks.append(model.predict_proba(x)[0][1])
    bookings["noshow_risk"] = risks
    bookings["risk_band"] = (bookings["noshow_risk"] * 10).astype(int)   # 0..9, 9 = highest
    bookings["fare_usd"] = bookings["fare_paid"].astype("int32")
    return bookings[bookings["risk_band"] >= 8]

The job runs on about 400,000 bookings a night. model_path is a string in a YAML config that a deploy script writes from a pull request to the config repository.

Stage 1 of 4

Review the code above before reading any hint. Assume the table has the columns named in FEATURES and that a missing value is possible in any of them.

Which issues would you block the merge on? Select all that apply.

`pickle.load` runs code embedded in the file, and the path comes from a config that other pull requests can change.`fillna(0)` on the whole frame turns a missing `days_to_departure` or `fare_paid` into a plausible-looking zero before scoring.`astype("int32")` truncates `fare_paid` to whole currency units, so any total built from `fare_usd` is biased low.No random seed is set, so the scores will differ between two runs on the same input unless the seed is fixed in the config.`predict_proba(...)[0][1]` takes the second column; the script should call `predict` to get a label, which is what the call team needs.