Sign in to track progress, unlock prep layers, and browse your pipeline.Sign in
Build an uplift model for targeting
Flu-shot Campaign: Treatment-Effect Modeling and Targeting Policy
You are tasked with designing a treatment-effect modeling and targeting approach for an upcoming flu-shot campaign. The objective is to determine whether to contact a customer via SMS or Email.
Historical campaign logs from the previous season are available, including randomized holdouts.
Data Available
- Features (pre-treatment only for modeling): Demographics, past visits, prior vaccinations, engagement history (prior opens/clicks), distance to store, appointment history.
- Labels:
y = 1if vaccinated within 30 days;0otherwise. - Treatments:
T ∈ {control, SMS, Email}, assigned at random with known propensitiesp_t. - Exposure indicators (post-assignment): Delivery status, opened. These are for diagnostics/mediation only; avoid leakage in Intent-To-Treat (ITT) models.
- Costs:
c_SMS = 0.02,c_Email = 0.001. - Operational Constraint: May contact at most 40% of eligible customers.
Tasks
- Modeling:
- Choose and justify an approach among: separate response models + two-model uplift, direct uplift/meta-learners (T-/S-/DR-learner), or multiclass treatment modeling.
- Address leakage (post-treatment features), class imbalance, and probability calibration.
- Evaluation:
- Define offline evaluation metrics: uplift/Qini curves and Area Under the Uplift Curve (AUUC).
- Compute incremental Return on Investment (ROI) including channel costs.
- Use policy evaluation with Inverse Propensity Weighting (IPW) or Doubly-Robust (DR) estimators.
- Policy:
- Given the 40% contact budget, describe how to rank customers by predicted incremental effect.
- Explain how to choose the optimal channel per customer (e.g.,
argmaxof channel-specific uplift minus cost scaled by value). - Explain guardrails such as do-not-contact lists, fairness across age/state, and frequency caps.
- Online Validation:
- Propose a gated rollout comparing model-based targeting versus uniform random targeting (both constrained to a 40% contact rate).
- Define success metrics and stopping rules for the online experiment.
- Diagnostics:
- Describe how to detect and mitigate harmful persuasion (negative uplift) segments.
- Explain how you would handle these segments in the targeting policy.
AI interview prep
Core dilemma
Maximizing the incremental impact (uplift) of a multi-channel marketing campaign under strict budget and operational constraints, while ensuring ethical and fair targeting.
Mid-level signals3
Senior signals3
Red flags3
3 tradeoffs and 3 stress tests locked
Interview prep
Unlock the interviewer X-Ray rubric, tradeoff matrix, stress tests, clarifying questions, and key concepts.
Build an uplift model for targeting
Flu-shot Campaign: Treatment-Effect Modeling and Targeting Policy
You are tasked with designing a treatment-effect modeling and targeting approach for an upcoming flu-shot campaign. The objective is to determine whether to contact a customer via SMS or Email.
Historical campaign logs from the previous season are available, including randomized holdouts.
Data Available
- Features (pre-treatment only for modeling): Demographics, past visits, prior vaccinations, engagement history (prior opens/clicks), distance to store, appointment history.
- Labels:
y = 1if vaccinated within 30 days;0otherwise. - Treatments:
T ∈ {control, SMS, Email}, assigned at random with known propensitiesp_t. - Exposure indicators (post-assignment): Delivery status, opened. These are for diagnostics/mediation only; avoid leakage in Intent-To-Treat (ITT) models.
- Costs:
c_SMS = 0.02,c_Email = 0.001. - Operational Constraint: May contact at most 40% of eligible customers.
Tasks
- Modeling:
- Choose and justify an approach among: separate response models + two-model uplift, direct uplift/meta-learners (T-/S-/DR-learner), or multiclass treatment modeling.
- Address leakage (post-treatment features), class imbalance, and probability calibration.
- Evaluation:
- Define offline evaluation metrics: uplift/Qini curves and Area Under the Uplift Curve (AUUC).
- Compute incremental Return on Investment (ROI) including channel costs.
- Use policy evaluation with Inverse Propensity Weighting (IPW) or Doubly-Robust (DR) estimators.
- Policy:
- Given the 40% contact budget, describe how to rank customers by predicted incremental effect.
- Explain how to choose the optimal channel per customer (e.g.,
argmaxof channel-specific uplift minus cost scaled by value). - Explain guardrails such as do-not-contact lists, fairness across age/state, and frequency caps.
- Online Validation:
- Propose a gated rollout comparing model-based targeting versus uniform random targeting (both constrained to a 40% contact rate).
- Define success metrics and stopping rules for the online experiment.
- Diagnostics:
- Describe how to detect and mitigate harmful persuasion (negative uplift) segments.
- Explain how you would handle these segments in the targeting policy.