Estimating Billboard Reach and Impressions

Scenario

An out-of-home (OOH) advertising team wants to estimate both reach (unique people who saw the ad at least once) and impressions (total ad views) for a single highway billboard over a specified time window (e.g., a week).

Task

Design a defensible estimation approach suitable for a data-science interview:

1. Define the metrics (reach, impressions, frequency) and scope (time window, directionality, static vs. digital rotation, daylight vs. illuminated hours).
2. Segment the population and traffic appropriately (e.g., by direction, time-of-day, weekday/weekend, lane, speed), and outline data sources (traffic counts, mobile/GPS samples).
3. Propose an exposure model that converts “passings” into “probability of seeing” the ad (e.g., visibility/time-in-view, unobstructed view, attention/viewability).
4. Derive formulas for impressions and reach. State assumptions clearly.
5. Show a small, numeric back-of-the-envelope example.
6. Briefly discuss validation and how you would quantify uncertainty.

Hints (you may use or adapt)

• Segment the population, collect traffic counts or mobile-location samples, model exposure probability, then multiply by average views per person.
• Use de-duplication or a distributional model (e.g., Poisson/NB) to infer reach from impressions when only pass-by counts are available.