Predictive Advertising in Multifamily: Questions to Ask Before You Trust the Model

Predictive advertising comes with an appealing promise: better targeting, better leads, less wasted spend. Before calling it a game changer, ask a harder question. Does the model improve a decision without creating a fairness, privacy, or measurement problem? A prediction is not proof. It is an input that requires scrutiny.

Ask what the model is actually predicting

Is the system predicting a click, an inquiry, a tour, or a lease? Those are different outcomes. Ask what data informs the prediction, how recent it is, what factors are excluded, and whether the vendor can explain the limits. A black box with a strong sales pitch is not a measurement plan.

Review housing-advertising risk before launch

Housing advertising has fair-housing obligations. Do not use a model or audience strategy that could exclude, steer, or treat people differently based on protected characteristics or proxies for them. Involve the appropriate legal, compliance, and advertising owners before testing new targeting methods.

Test against a real comparison

Set a specific hypothesis and compare like with like. Use the same message, offer, landing-page quality, and conversion definition where possible. Review qualified actions and lead quality, not just low-cost clicks. Document the result so the next decision is based on evidence.

Keep a human responsible

Automation can help a team process information, but it does not own the outcome. Assign an accountable person to approve the audience, inspect the creative, review the data, and stop the test if the result is unclear or inappropriate.

The smartest model in the room still needs an accountable human asking what it got wrong. Criterion B can help your team evaluate paid-media tests with clearer measurement and responsible guardrails.