A single-location restaurant is a hyper-local, high-intent, low-repeat-visit business. The typical customer types “italian restaurant near me” on a Thursday afternoon, picks one of the top three results, calls or reserves, and shows up that night. If they don't come tonight or this weekend, the ad spend that acquired them is gone. There's no long consideration window, no cart, no re-engagement path that pays back on a 6-month lag.
Performance Max is built for the opposite business. It assumes a broad audience, a purchase funnel with warm-up touchpoints, catalog-driven creative, and enough conversion signal per week to feed the AI. A single-location restaurant with 3–8 conversions per week and a “book a table” goal simply does not meet PMax's minimum viable signal floor. It runs, it spends the budget, it reports impressive numbers on display placements — and it produces almost no incremental covers.
Why PMax Fails Restaurants Specifically
| Restaurant reality | What PMax needs | Result |
|---|---|---|
| 3–15 reservations per week | ≥ 30 conversions per week to leave learning phase | Permanent learning; strategy never stabilises |
| 5 km radius of intent | Broad audience for asset expansion & lookalikes | Display spends most of budget outside catchment area |
| Value = tonight's cover only | Cross-session attribution over 30-day window | Optimises for someone who won't come until next month |
| Best signal = phone call from mobile | Optimises against “conversion” = anything set as goal | Cheapest signal (menu views) crowds out calls |
| Creative = 4 hero photos + menu | Asset variety for testing across placements | PMax renders low-quality auto-crops in Display |
The killer is the second-to-last row. PMax will optimise for whichever signal you feed it, and cheaper signals — “view menu”, “click directions” — are 30–50× cheaper to trigger than an actual phone call or reservation. So PMax quietly reallocates budget toward the cheapest signal on the page, reports thousands of “conversions”, and produces no meaningful covers. The restaurant owner sees the dashboard and wonders why the dining room is empty.
The Structure That Actually Works
Search — Brand: Restaurant name + variants. Cheap, high CVR, isolate from generic.
Google Business Profile — Ads on GBP: Small budget slice (~15%). Drives calls and directions from Maps.
Retargeting — Display + YouTube: 30-day site visitors, 7-day menu viewers, own audience shape. Small budget, promo-driven creative.
This structure explicitly excludes PMax. If the client insists on running it, we cap it at 15% of budget, give it a single hard conversion goal (a phone call or a completed reservation, never a soft signal), and add the brand + core generic terms as account-level negatives to prevent cannibalisation.
Conversion Tracking That Reflects the Business
Restaurant conversion tracking is where 80% of accounts we audit fall over. The default GTM setup fires a “click phone number” conversion from the mobile menu button; the same click gets fired if the user just hovers to see the number, if the user is on a device that can't call (desktop), and if the user calls but reaches voicemail. The signal is noisy on both directions.
Use Google Ads call extensions with a forwarding number. Google tracks the actual call duration and marks anything ≥ 60 seconds as a conversion. This is the cleanest signal restaurants have; run it on every campaign that can carry a call extension.
OpenTable, Resy, TheFork, SevenRooms — all support conversion postbacks. Wire the confirmation to a Google Ads offline conversion upload keyed on the source click ID. This is your gold-standard conversion signal; treat it as the primary.
If you sell delivery through your own site (not just Uber Eats aggregators), track it as a distinct conversion action and run it in a distinct campaign. Delivery buyers are a different audience with different creative — don't let their signal contaminate the dine-in campaigns.
Directions from GBP are a decent signal but count only as “secondary” in your Smart Bidding stack — see the primary/secondary lever piece. Setting them as primary teaches the algorithm to chase curious tourists who never eat.
Google Business Profile Is Half the Game
The single biggest lever most restaurants ignore is Google Business Profile optimisation itself, before any paid ad. GBP performance influences Local Pack ranking, which influences how efficient every subsequent paid ad becomes. A restaurant with a well-maintained GBP (photos updated monthly, Q&A answered, reviews replied to, menu synced) will pay 30–50% less per call and reservation on paid ads than the same restaurant with a neglected GBP — because the ad reinforces a listing Google already trusts.
We run a monthly GBP-only audit before touching the ad account:
- Photos: ≥ 12 recent, at least one per week
- Menu: current, prices matching in-restaurant, dishes with photos
- Reviews: all in last 30 days replied to within 48 hours
- Q&A: all questions answered by the business (not customers)
- Hours: correct including holiday exceptions
- Attributes: reservations, outdoor seating, dietary options accurate
None of that is glamorous. All of it moves the paid-ad performance more than any bid adjustment.
Creative That Works for Restaurants
Restaurant creative rules are the inverse of e-commerce. The lower-fi it looks, the better it performs — because the buyer is trying to picture themselves inside the room tonight, and studio-lit food shots feel like menu photography, not a place. Three formats consistently win:
| Format | Where it works | Why |
|---|---|---|
| Short room video (10–15s, phone-shot) | Google Display, YouTube shorts, Meta feed | Sells atmosphere; buyer imagines being there tonight |
| Staff-shot dish close-up + steam | Search extensions, GBP posts, Display | Real, seasonal, doesn't feel like agency stock |
| Handwritten menu / today's special | Search sitelinks, GBP posts | Urgency, freshness, hyper-local feel |
Avoid: aggregator-generated hero shots (they render everywhere and the user recognises them), text-heavy graphics with the logo, promotional banners with %-off deals (attracts price-hunters who never come back).
The Local Intent Search Set
Search keyword sets for a restaurant break into four intent tiers. Structure the account by these tiers, not by menu category or cuisine.
| Intent tier | Example | Approach |
|---|---|---|
| Book now (highest) | “book table italian [neighbourhood]” | Exact + phrase, highest bid, own campaign |
| Near me / local (high) | “italian restaurant near me” | Phrase, location targeting 3–5 km, tight negatives |
| Cuisine + city (medium) | “best italian restaurant [city]” | Phrase, broader radius, exclude tourist-only queries |
| Menu / dish (medium-low) | “where to eat carbonara [city]” | Phrase, own campaign, retarget these users heavily |
Split the campaigns by tier because bidding, negatives, and creative differ meaningfully across them. Mixing “book now” queries with “where to eat” queries in one bucket forces Smart Bidding to average across CVRs that differ by 5–10×, which reproduces the PMax problem in miniature.
Audit Checklist for a Restaurant Owner
What Reasonable Performance Looks Like
For a single-location mid-tier restaurant (30–70 covers/night, average cover 30–60 €) running the structure above in a European city, we consistently see: 4–9 € per phone call from Google Ads, 8–18 € per confirmed reservation via booking system postback, 15–35 € per net-new customer including retargeting. On budgets between 500 and 1500 €/month, that produces 40–120 incremental covers/month — a real, measurable share of a restaurant's weekly cover count.
PMax accounts running the same restaurant on the same budget usually report 3–5× more “conversions” but produce zero incremental covers, because the conversions are menu views and Maps directions. If your restaurant is on PMax and you can't point at 40+ new covers a month attributable to it, the structure is wrong, not the budget.
The Google Ads engine is powerful. It just needs the right conversion signal and the right unit of learning. A restaurant is not a broad-audience e-commerce store; the account structure has to reflect that. See how we approach these builds inside restaurants performance media.
