I Always Wanted a PPC Agency Obsessed With My Results. So I Built One.

It wasn't too long ago that it would cost you $10,000 to have a specialized agency set up a dynamic, comprehensive pay-per-click campaign for your business. Once it was built, you'd pay them $1,000 a month to maintain it.

If you landed one of the rare good ones, they'd genuinely optimize your campaign each month based on the data. But let's be honest about the other ninety percent: most agencies quietly hope you never call, and couldn't tell you how well your campaign is actually doing if you did.

I always wanted an agency that was obsessed with my performance and my success. Human nature doesn't really work that way — especially when you're one of thirty accounts somebody has to service before Friday. It isn't malice. A person with thirty clients has to ration attention, and the accounts that get it are the ones making the most noise or paying the biggest fee. Mine was never going to be either.

That's why AI has changed the game in pay-per-click.

Training the agent

I spent many weeks training my agent to become a genuinely comprehensive expert in pay-per-click advertising. We scoured research forums, practitioner websites, and case studies to understand which strategies are actually working right now — not the advice that was true in 2019 and has been getting copied between blog posts ever since.

Most importantly, we did a deep dive into the never-ending parameters and customizations that Google gives you.

That last part is the whole thing, and it's worth sitting with. Google Ads is not a hard product to use badly. You can have a campaign running in fifteen minutes. What Google actually exposes, though, is an enormous surface of settings — match types, bid strategies, geographic targeting modes, negative keyword scoping, ad group structure, asset types, network options, audience layers, budget pacing — and almost every one of them has a default that's fine for Google and mediocre for you.

Nobody fills all of it out. Not because it's secret, but because it's tedious, and because the person doing it is being paid per campaign rather than per dollar returned.

But one of AI's single strongest attributes is being able to digest and analyze massive amounts of data. So we took the almost endless set of parameters Google offers, and we filled out every single one of them, optimized exactly to my business.

What that actually produced

• 21 ad groups, each narrow enough that one ad can answer the exact search. "Romantic Hot Tub Treehouse" is its own ad group, not a keyword buried in a generic bucket.
• 125 keywords, each with a deliberate match type and its own maximum bid, priced against the real market rate for that specific term instead of one blanket number across the account.
• 21 responsive search ads, one per ad group, every headline and description validated against Google's character limits before import so nothing gets silently truncated.
• 104 negative keywords — a 75-term master exclusion list applied account-wide, plus 29 more whose only job is routing, making sure a treehouse search lands in the treehouse campaign and never gets picked off by a broader one.
• Geographic targeting drawn on the real drive-shed — the towns people actually drive from for a weekend, set to "presence" only, so I'm not paying for somebody in Florida idly reading about Ohio.
• The full asset set — sitelinks, callouts, images, call and location assets, so the ad takes up as much of the results page as Google will hand over.
• Search network only, display explicitly off, manual bidding to start with a written plan for when to graduate to automated bidding.

None of those individual choices is exotic. Getting all of them right, at the same time, on a small account, is what nobody does.

The five words that show what analysis buys you

Here's my favorite example of the difference, and it's a single decision out of hundreds.

Two keywords in my account contain the same five words: places to stay in amish country ohio. One is set to exact match, the other to phrase match. Three weeks of real bookings later:

Same words. One version returned 158 times what it cost. The other spent fourteen times as much money to hand back sixty cents on the dollar. Phrase match lets Google decide what "close enough" means, and Google's definition is generous in whichever direction spends your budget.

Multiply that one call across 125 keywords and you have the entire difference between an account returning 19.8x and an account returning 2x.

The exhaustiveness pays off in the other direction too. The single highest-returning keyword in my account is romantic treehouse ohio: $2.76 spent, $736.94 booked — a 267x return. It's a low-volume term. Any human building a keyword list by hand would have skipped it as too small to be worth the row. It's in there because the agent didn't build a selective list, it built a complete one.

Making sure the money is real

Any of this is worthless if you can't verify that the bookings are real and that the right ones are being credited. So before a single campaign was allowed to turn on, I set up custom tagging to follow the entire customer journey end to end.

Analytics on the website and inside the booking engine, with cross-domain tracking configured so a visitor's session survives the jump between the two — that's where most lodging businesses lose the thread, and it's why so many owners think their ads don't work when really their measurement doesn't. The purchase event carries the actual booking value, imported into Google Ads as the primary conversion action, with checkout starts tracked separately as an early signal. Then a live test booking, traced the whole way through, confirmed to land with the right dollar amount attached to the right click.

Every number in this post comes through that pipeline. They aren't estimates or modeled projections — they're reservations, with the revenue attached to each one.

It's also proved itself under fire. When our booking engine vendor's confirmation event quietly stopped firing for a few days, the agent caught it within a day by cross-checking Google's numbers against the actual reservation database, and backfilled the missing conversions from the reservation data so the bidding never learned from a false zero. Not one booking went unattributed.

The results

July 12 through August 1, straight from the Google Ads API:

Three weeks. A thousand dollars in, twenty-one thousand dollars of booked stays out.

Sit with the unit economics for a second, because that's where it really lands. The average booking is worth $680. It costs $34 to win. On the treehouses it's better still — $23 to bring in a booking worth $738. I am buying six hundred and eighty dollar reservations for the price of lunch.

The blended cost per click across the whole account is 26.6 cents.

I consider this a breakthrough, and not a small one. Think about what it does to a business. The old agency math was $10,000 to build plus $12,000 a year to maintain — $22,000 before you've bought a single click. My entire ad spend so far is $1,063, and it has produced twenty-one thousand dollars of bookings.

At this efficiency the constraint on growth stops being the marketing budget. Every dollar I put in comes back roughly twenty times over, which means the real question is no longer "can I afford to advertise" — it's "how much occupancy can I absorb." That is a completely different conversation to be having about your business, and most small operators never get to have it, because they've only ever seen advertising that loses money slowly.

On being early

It's no secret that there are others out there building agents like this. I'm not going to pretend I've invented something nobody else will have. Within a couple of years this will be ordinary, and the agencies charging $10,000 for a build will be running the same kind of tooling behind the curtain.

But the advantage was never the tool. It's the head start.

Ad auctions are competitive markets, and right now I'm buying romantic treehouse ohio for $2.76 because almost nobody else has thought to bid on it. When everyone's agent has read the same research and built the same exhaustive keyword list, that term won't cost $2.76 anymore. The cheap ground gets claimed by whoever shows up first, and every week I run this campaign is a week of data, of conversion history, of quality scores, and of bookings that competitors starting next year will have to buy their way past.

Being early isn't the whole edge. But in an auction, it compounds — and by the time this stops being a novelty, I'd rather be the incumbent than the one paying to unseat one.