The border will spend your budget if you let it
Set a radius around a business in Tijuana and a large part of that circle lands in San Diego County. The map looks reasonable. The spend is not.
Two things go wrong at once. You compete in auctions against advertisers bidding dollars, which raises what you pay for every click you do win. And the clicks you win come from people who are not going to cross an international border to use a local service.
There is a second, subtler setting behind this. Google's location options historically default to including people who show interest in your targeted locations, not only people physically in them. On a border that quietly widens your audience in the exact direction you did not want.
The fix is unglamorous: target the city and municipality explicitly rather than a radius, set the location option to presence rather than presence-or-interest, and add exclusions for the areas across the border. Then read the geographic report weekly for the first month, because the report is where you find out what the settings actually did.
Businesses that genuinely want cross-border customers, and in Tijuana many do, should run that as a separate campaign with its own budget, language and copy. Mixing both audiences in one campaign means neither gets optimised.
Language targeting is not translation
Google's language targeting works on the user's interface and content-language signals, not on the language of the query. Which means a Tijuana resident browsing with an English interface may not see your Spanish-only campaign even while searching in Spanish.
In a bilingual city that is a real gap, and it is invisible in the account because you cannot see impressions you never entered the auction for.
Running both languages is usually right here, but as separate ad groups or campaigns with their own keywords and copy. Translating your Spanish ads into English produces English ads that sound translated, and they underperform against advertisers writing natively.
Keyword research has to happen twice for the same reason. The English terms your buyers use are rarely literal translations, particularly in the categories where cross-border demand is strongest.
Most local conversions are phone calls
For a local service business the conversion is usually a call, not a form. A campaign optimised on clicks or form fills in that context is optimising the wrong end.
That means call extensions, call-only campaigns where they fit, and call tracking that reports which campaign and which query produced the call. Without tracking, the platform reports its best guess, and its best guess flatters whichever campaign got the most impressions.
Then close the loop backwards. What matters is not the call, it is whether the caller became a customer. Someone has to record that outcome against the source, and it is the step that gets dropped in almost every account we inherit.
Judge the account on cost per customer, not cost per click. Those two numbers can point in opposite directions, and the second one is the one agencies put in reports.
Where the waste usually is
Broad match with no negative keyword list. Broad match will find you traffic; without negatives it finds you traffic for things you do not sell.
Search partners and display expansion left enabled on a search campaign, which quietly moves budget onto inventory that converts very differently.
Every ad pointing at the homepage. If someone searched for a specific service and lands on a general page, they have to find it again, and a share of them will not.
No conversion tracking at all, which is more common than it should be and makes every other optimisation guesswork.
Automated bidding running without enough conversion data to learn from. Smart bidding needs volume; switched on too early it optimises toward noise.
