Someone who types "pizza delivery near me" at 7:40 p.m. is going to order within twenty minutes. The only question is whether that order goes to you or to whoever caught the click first. That urgency is what makes food searches some of the best-converting traffic you can buy, and it's the same reason so many restaurants still lose money on them. Usually the budget leaks in small, boring ways: bidding on words that don't convert, letting aggregators sit on your brand name, and sending paid clicks to a menu PDF.
Three keyword tiers, three very different prices
The top tier is hunger right now: "sushi delivery near me", "pizza delivery [neighborhood]", "thai food open now". This is the highest intent you can buy, and in the US it runs about $1.50–4 per click depending on the city. Start here, and start on phrase match rather than broad. Broad match will happily spend your money on "free pizza recipes".
Below that comes dish plus city: "best ramen austin", "birria tacos delivery dallas". The clicks are cheaper and the intent is a bit slower, but they're worth bidding on once Tier 1 volume runs dry.
Tier 3 is pure research, like "best restaurants for date night". On a small budget, skip it. Those people aren't hungry, they're browsing, so let the marketplaces overpay to reach them.
Build a negative list on day one: "jobs", "recipe", "free", "calories", and every neighborhood you don't deliver to. Ten minutes of that work routinely trims 15–20% off a food-delivery budget.
Defend your brand name: aggregators are already bidding on it
Search your own restaurant's name in an incognito window. In most US cities you'll see DoorDash, Uber Eats, or Grubhub sitting in a paid slot above your organic listing. A customer who looked for you by name clicks the top result, orders on the marketplace, and you hand over 15–30% commission on a sale that was already yours.
Branded clicks usually cost $0.30–0.80, because your quality score on your own name is about as high as it gets. Paying fifty cents to protect a $30 direct order from a $9 commission is the easiest math in this whole article. Run a small always-on branded campaign before you spend a dollar anywhere else.
We'll show you what a 15%-converting ordering site does to your cost per order — with your numbers.
Your landing page sets your cost per order
Take $1,000 of budget at a $2 average CPC and you've bought 500 clicks. Everything that matters now happens after the click, and the spread is brutal:
Same budget, same keywords, and a 15× swing in cost per order. This is why "Google Ads doesn't work for us" almost always turns out to be a landing-page problem that got blamed on the ads. Before you scale spend, make sure the clicks land on an ordering site where the menu, cart, and checkout all live on one fast page, not a homepage with a "View menu" PDF and a phone number.
Settings that actually save money
Target your real delivery zones instead of a city-wide radius, because every click from outside the zone is a guaranteed zero. Run ads only when the kitchen is open and weight the schedule toward your peaks, since paying for "open now" searches at 3 a.m. helps nobody. Track completed orders and revenue, not clicks or phone calls, or you'll be optimizing blind while Google's automated bidding chases whatever cheap action you handed it. Then pull cost per order into the same dashboard where you watch margins, and grade campaigns each week on a single number: what an extra order costs against what it earns.
Leave Performance Max and Display alone at the start. Both can earn a place later, but they need conversion data to learn from, and on a cold start they'll burn budget on low-intent impressions. Get that data from search campaigns first.
There's one more measurement trap worth naming. Grade campaigns on contribution rather than top-line revenue. A $13 cost per order sounds fine until you remember the order also carries food, labor, packaging, and delivery costs. If a $30 order contributes $12 before marketing, a $13 acquisition cost loses a dollar on the first purchase, which only works if that customer comes back through a channel you don't pay for again.
Ads buy the first order. The app keeps the rest.
Even at $13 an order, buying every repeat order from Google is a treadmill. Restaurants that win on paid search treat it as a front door instead. The first order comes in through the ad, the confirmation screen offers a branded app with cashback on the next order, and from then on repeat orders move to push notifications that cost nothing to send. Apps convert as high as 35%, because opening the app is already a decision to order. Do it right and Google buys you a customer once, not every week.
Do this next
This week, search your restaurant's name in incognito. If a marketplace ad is sitting above you, launch a branded campaign at $5–10 a day. Then look at where your ad traffic lands, and if it isn't a page that can take an order in under a minute, fix that before you touch a single bid.
Ordering site, branded app, and marketing analytics — launched in about two weeks.