The short answer
PPC means paying for the click, not the customer, so the whole job is making sure the clicks come from people who want what you sell. Get conversion tracking right, set location targeting to presence, keep search campaigns off the display network, and read the search terms report weekly. Bidding comes last.
What you are actually paying for
You pay when somebody clicks, not when they buy, and that single fact explains most of what goes wrong in paid search. Google is happy to sell you clicks all day long. Whether those clicks came from people who wanted what you sell is a question only your tracking can answer, and in most accounts we inherit, the tracking cannot answer it.
Position in the auction is not bought with money alone. Your bid is weighed against an assessment of how relevant your ad and landing page are to the search, which is why two advertisers bidding the same amount pay different prices for the same slot. Relevance is not a nice-to-have. It is a discount.
Keywords, match types, and the report nobody opens
A keyword is not a search. It is an instruction about which searches you are willing to enter. Broad match takes the widest interpretation of that instruction, phrase match narrows it, exact narrows it further, and all three now include variations Google judges to be close enough.
The search terms report tells you what actually happened. Open it every week. In a new account it is normal to find a third of the spend sitting on searches you would never have chosen: people looking for jobs, for do-it-yourself instructions, for a competitor by name, or for a free version of what you sell. Turn what you find into negative keyword lists. Jobs, salary, training, free, and how to belong in a shared negative list before you launch anything at all.
Two defaults that quietly waste money
The first is location targeting. The default option shows your ads to people who are in your area or who show interest in it, which for a plumber in Monmouth County means paying for clicks from someone three states away reading about a town they used to live in. Change it to the presence option, so you are paying for people who are actually in or regularly in the places you serve.
The second is the display network setting on a search campaign. Search campaigns can be set to also run on display, where clicks are cheap, plentiful and almost never customers. If the campaign is meant to be search, switch that off before it spends anything.
While you are in the settings, look at auto-applied recommendations. Google can apply its own suggested changes to your account automatically, including adding keywords you did not choose. A few are sensible. Deciding for yourself is better, and it takes two minutes to turn off.
Ad copy that does not get skipped
Your ad sits next to three others making identical claims. Quality service, free estimates, family owned since whenever. Everybody writes that, which means it distinguishes nobody. Write the specific thing instead: the response time, the length of the guarantee, the real price band, the brand you are certified to install, the fact that a person answers the phone at six in the morning.
Then match the ad to the search. Somebody typing emergency ac repair and somebody typing new hvac system cost are at completely different stages and should not see the same ad or land on the same page. That mismatch is the most common reason a technically well-built account underperforms.
The landing page is half the job
Sending paid traffic to the homepage is the default mistake. A homepage has to serve everyone, so it serves nobody who arrived with one specific problem. The page should repeat the phrase they searched near the top, answer the question directly in two or three sentences, and put one obvious action in front of them.
On a phone that action is a tap-to-call button, above the fold, not hidden behind a chat widget. A nine-field form on a mobile screen is a decision to collect fewer leads. Name, phone number and one line about the problem is enough to call somebody back. Work on the landing page usually returns more than the same hours spent adjusting bids.
Track conversions once, and correctly
A conversion has to be something you would gladly pay for. A page view is not. A visit to a thank-you page is closer, but still counts anybody who wandered in. Calls should be tracked as calls with a minimum duration, so that a wrong number hanging up after four seconds does not get recorded as a lead and then fed to the bidding algorithm as an example to find more of.
Decide whether to count every conversion or one per click. For lead generation, one is almost always right, otherwise a single anxious customer calling three times looks like three customers. Then confirm the same lead is not being counted separately in the ads platform and in analytics, because that discrepancy will eat an entire monthly meeting every month until somebody resolves it.
Budget and bidding, last
Set the budget from the value of a job rather than from what feels comfortable. If a booked job is worth eight hundred dollars, you close one inquiry in three, and one click in five becomes an inquiry, you can work backwards to what a click can be worth to you. That number, not your nerve, is the ceiling.
Start on a simple strategy only long enough to gather conversion data, then move to a conversion-based one once the count is reliable. Automated bidding on top of broken tracking will optimize toward the wrong outcome extremely efficiently. Bidding is the last lever to touch, because it is the only one that cannot repair a structural problem underneath it.
The mistakes that cost the most
Launching with no conversion tracking, then judging the account on clicks and click-through rate.
Leaving the search terms report unopened for a quarter.
One campaign covering every service, so emergency work quietly subsidizes casual browsing.
Running ads during hours when nobody answers the phone.
Pausing after three weeks. Most local accounts need six to eight weeks before the data means anything.
Follow-up questions
Enough to gather meaningful data in a reasonable time, which usually means at least thirty to fifty clicks a week on your core terms. Below that you will spend months unable to tell whether anything works. If that is out of reach, narrow the targeting rather than cutting the budget.
With care. It works well for ecommerce with a clean product feed. For lead generation it often absorbs traffic that would have arrived through your own brand name anyway and reports it as new. Run search campaigns properly first, so you have something honest to compare against.
Six to eight weeks gives a clear read on most local accounts, assuming conversion tracking was in place on day one. Judging it at two weeks is judging noise. If the account ran a year without tracking, you are starting from week one whatever the history says.
Let’s build something.
Twenty minutes on the phone and you’ll know whether we’re the right fit. Either way you’ll leave with something useful.