For a hospitality brand with locations across the country, marketing accountability is unusually hard. Guests don't convert in a clean digital funnel — they see a TV spot, hear a radio ad, get retargeted online, and eventually walk through a door. Attributing that visit to any single touchpoint misses the picture, but spending without knowing what's working isn't a strategy either.
Building the Model
We built a marketing mix model on a full year of weekly data across all media channels, calibrated against two outcomes: total guests and new guests. The distinction matters — some channels are better at driving trial from first-timers, while others sustain frequency among existing guests. Treating those as the same outcome produces misleading efficiency estimates.
The model used geographic variation as the primary source of causal identification — studying how guest trends in markets with different media weights moved relative to each other, rather than relying on national time-series correlation that's confounded by design.
National-level correlations between spend and visits are almost always positive — you spend more when you're trying to drive visits. The question is whether markets where you spent more actually outperformed markets where you spent less.
What the Model Found
to paid marketing
to paid marketing
driven by paid activity
Paid marketing accounted for 28% of new guests and 15% of total guests — meaningful, but it also means the majority of traffic came through organic channels, word of mouth, and brand recognition that predated any given campaign.
Direct digital was the largest contributor to total guest volume. Paid social drove the most new guests — consistent with the hypothesis that social reaches first-timers more effectively. One finding stood out: programmatic display, while less efficient per impression than direct digital, had a disproportionate influence on total guests relative to new guests, suggesting it was driving re-engagement with lapsed visitors more than first-time acquisition. Useful distinction for planning.
The Out-of-Home Question
OOH ran in only two markets that year — small sample. Both showed positive lift during their flight period relative to comparable markets that didn't. Directional rather than conclusive, but consistent enough to warrant a broader rollout designed for proper measurement.
What Came Next
The immediate deliverable was a scenario simulation tool — a workbook letting the planning team test budget allocations and project guest impact before committing spend. What happens if we shift 10% of video into paid social? What's the projected impact of a 20% cut? Longer term, the framework was designed to be updated as new data accumulated, with geo-level testing built into the plan to continuously sharpen the estimates.
Want to understand what's actually driving traffic and conversion in your marketing mix? Say hello.