High Data Drift Question 148 of 221

Your PSI alerts fire every Monday. How do you decide if that is noise, seasonality, or a real incident?

MLOps track · Speak this in 60–90 seconds · Faridabad & Delhi NCR

PICTURE THIS: DATA SPLIT

Train 70%Val 15%Test 15%

Fit on train, tune on val, report on test once.

Simple meaning

I would compare against last Monday and a seasonal baseline, not only versus the training month.

WHY — Data Drift instead of guessing?

Why interviewers care about Data Drift:

Data Drift questions separate

people who only read docs from people who shipped.

Keep it short, concrete,

and tied to MLOps work.

Stay structured

Name the idea, why it exists, then one short example.

Close cleanly

End with when you use it and one common pitfall.

STEPS — What happens step by step?

Before you speak the answer, walk the interviewer through these steps:

1
I would compare against
last Monday and a seasonal baseline, not only versus the training month.
2
If delayed labels and
calibration are stable, I would retune the reference window or suppress known calendar effects.
3
If a new traffic
source appears in slices, I would treat it as an incident and check pipelines and product launches.
4
Give an example
One tiny concrete case you can say aloud.
5
Common mistake
What juniors usually get wrong.
6
Close
When you pick this over the alternative.

EXAMPLE — See it in action

Here's a short line you can speak, broken into clear beats:

Say this line

“If delayed labels and calibration are stable, I would retune the reference windo”

Break into beats

Ifdelayedlabelsandcalibrationare

Speaking order

2987408337471632900

Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.

Key takeaway

I would compare against last Monday and a seasonal baseline, not only versus the training month. If delayed labels and calibration are stable, I would retune the reference window or suppress known calendar effects.