Operating case study · Restaurant advertising
From $11.11 to $4.78 per modeled store visit.
Jon Harmeling used AI to review Google Ads and support campaign decisions inside his Miami Shores restaurant. Following the changes, the campaign's August advertising cost per Google-modeled store visit was approximately 57% lower than in its May nonbrand test.
By Jon Harmeling · Data retrieved September 6, 2026
The measured result.
The comparison uses the same nonbrand Google Ads campaign, called Chick-fil-A Volume. Cost per modeled visit equals campaign spending divided by Google's attributed store visits.
| Measure | May 11–312026 | August 1–312026 |
|---|---|---|
| Advertising spend | $4,024.56 | $2,212.81 |
| Google-modeled store visits | 362.14 | 463.25 |
| Cost per modeled store visit | $11.11 | $4.78 |
| Days with advertising | 18 | 26 |
The approximately 57% reduction is calculated from unrounded values. The periods contain different numbers of advertising days, so total visits are not presented as a same-duration growth result. Google-modeled visits are estimates, which is why the report includes fractional counts.
Start with the business question.
The operating question was straightforward: how efficiently was the restaurant's advertising budget producing reported store visits?
Jon used an AI assistant to examine campaign performance, question the setup, and recommend improvements. He authorized the work and retained responsibility for the decisions. That distinction matters when software can both analyze an account and help change it.
What actually changed.
The July relaunch changed the bidding approach, narrowed advertising hours, and revised two headlines about menu timing. A July 31 adjustment then capped the maximum price of a click at $0.90, with a $50 daily budget. That budget increased to $75 on August 8 and $100 on August 26 while the click cap remained.
Several sound settings were already in place, including store-visit goals and nonbrand targeting. Proposed additions such as a separate breakfast ad group and a new Performance Max campaign had not been built. They are not credited for the result.
Lower click costs are a plausible part of the improvement. This before-and-after comparison does not isolate the contribution of any one change.
What another frontline leader can apply.
- Choose the measure before making the change. Decide which result matters to the business and understand how it is counted.
- Separate a recommendation from an executed action. Keep a record of what changed, when it changed, and who authorized it.
- Own the judgment. Use AI to investigate and support implementation while staying accountable for spending, interpretation, and the next decision.
This case demonstrates operating efficiency and management judgment. Jon's Frontline AI Leadership keynote carries the question further: what will leaders do with the capacity AI creates to coach, develop, and invest in people?
How to read this case.
Source: Google Ads campaign and store-visit reporting retrieved September 6, 2026, checked against the implementation record. The baseline is May 11–31; the comparison is the full month of August 2026. August conversion reporting remains subject to revision. Google explains how store visits are reported.
Google's modeled visits do not establish how many additional customers the advertising created. This was not a controlled experiment and does not prove AI alone caused the change. Revenue, profit, time saved, and employee-development outcomes were not measured in this comparison.