Project Planning & Deployment Flashcards

A data program has a Cost Performance Index (CPI) of 0.8. What does this mean?

Answer: The project gets $0.80 of work value for every $1 spent

CPI = EV/AC, so 0.8 means cost overrun: each dollar spent produces only 80 cents of planned value.

Which approach to rolling out an enterprise data governance program across a large company usually works best?

Answer: A phased rollout starting with a pilot domain, then scaling with lessons learned

A phased, pilot-first rollout tests the operating model, shows value, and improves the approach before enterprise scale-up.

In change management for a new self-service analytics tool, what does the 'A' in the ADKAR model stand for first?

Answer: Awareness

ADKAR stands for Awareness, Desire, Knowledge, Ability, and Reinforcement.

Before promoting a data pipeline to production, a team adds automated checks that block deployment if null rates or schema changes go past set thresholds. What is this control called?

Answer: Data quality gate

A data quality gate is an automated checkpoint that stops promotion when quality rules fail.

In a stakeholder power/interest grid, how should a CDO handle a stakeholder with high power and low interest in the data project?

Answer: Keep satisfied

High-power, low-interest stakeholders should be kept satisfied so they do not become blockers.

An ML model has been deployed and its prediction accuracy slowly falls as customer behavior changes. Which MLOps capability should spot this?

Answer: Model drift monitoring

Drift monitoring tracks changes in input data and prediction quality, then triggers retraining or review.

Which deployment technique runs a new model on live production inputs without its predictions affecting users, so its outputs can be compared to the current model?

Answer: Shadow deployment

Shadow deployment mirrors production traffic to the new model and logs its outputs without serving them to users.