Ember Kitchen Ubud
Restaurant · Google Ads Performance Max + Local Actions
Ember Kitchen is a dining destination in Ubud, Bali — a market with high hospitality density and strong competition for tourist attention. The target audience: travelers and expats already in the area, actively looking for dining options.
New account, no historical data, modest daily budget of $15–20. The critical question was not which keywords to bid on — it was which conversion event to optimize toward. Get that wrong, and the algorithm trains on the wrong behavior from day one. Get it right, and a small budget can produce outsized local reach.
For a restaurant without an online reservation system, the conversion hierarchy is clear — but easy to misset. The goal that matters is foot traffic. 'Get directions' is the highest-intent trackable proxy.
User actively navigating to the location — highest intent
User viewed the site — no intent signal for visit
Low frequency in tourist dining context
- →Performance Max with Local ActionsCampaign type
PMax with Local Actions enabled — Google serves ads across Search, Maps, Display, and YouTube, optimizing for in-store direction events
- →Geo radius A/B: 5 km vs city-wideTargeting test
5 km radius produced 31% lower CPA than city-wide targeting — nearby users convert; distant users browse
- →Dual audience strategyTargeting layer
Primary: female 18–30 (high-frequency dining content consumer). Secondary: general 25–45 for broader coverage at lower bid priority
- →Creative differentiationAsset mix
Lifestyle-forward imagery over generic food photography — higher CTR in direction-intent placements where visual identity drives the decision
For a restaurant, the meaningful action is someone walking through the door. 'Get directions' is the closest trackable proxy to that. Optimizing for website sessions, page views, or a generic 'engagement' event would have trained the algorithm to find the wrong people entirely — users who browse but never visit.
A/B testing showed the 5 km radius outperformed city-wide targeting. The reason is behavioral: users within walking or short-ride distance are far more likely to convert on a 'get directions' goal than users across town. City-wide captured impressions from users who might visit 'someday'. 5 km captured users who might visit tonight.
Ubud has high hospitality density — competing venues run similar campaigns with similar budgets. On a $15–20/day budget, the creative has to do significant work. Lifestyle-led imagery targeting the primary female 18–30 segment consistently outperformed generic food photography in the direction-intent context.
$1.90 CPA is achievable because 'get directions' is a high-frequency, low-friction micro-conversion. If the goal had been set to reservation completions or website purchases, the CPA would be structurally higher and the volume structurally lower. The efficiency is a product of choosing the right signal — not of unusually cheap traffic.
A new Google Ads account on an $18/day average budget delivered 406 direction conversions per month at $1.90 CPA — against an industry average of $4–6 for hospitality local actions. Geo A/B test confirmed 5 km radius as the efficient targeting boundary. Primary audience: female 18–30, supplemented by general 25–45 at lower priority.
Campaign: Performance Max + Local Actions · $15–20/day budget
Geo: 5 km radius · 406 direction conversions/month · $1.90 CPA
Client name is a pseudonym. Report ID is internal. Metrics reflect specific account conditions — results vary by market, budget, and local competition density.