An illustrated lecture on paid search marketing: how the ads are built, how the auction decides who shows up, and whether any of it causes a sale that wouldn't have happened anyway.
Every day, billions of people type exactly what they want into a box. Paid search is the business of deciding how much to pay to be the first answer.
No other ad channel works like this. A billboard guesses who drives past. A social ad guesses who might care. A search ad waits until someone says "emergency plumber near me" and then shows up. That is why it became one of the largest ad markets in the world, and why it's worth asking hard questions about.
In which a stranger tells you exactly what they want, and you decide how much to pay to hear it.
The searcher typed what they want. You don't guess the audience; the audience announces itself.
They haven't decided yet. A good ad at the right moment can still change where the money goes.
If you're not in that spot, a competitor is. Part of paid search is simply defense.
Hold onto #2. Later we'll ask what happens when someone searches your brand name, when they have, in fact, already decided.
You don't buy searches; you buy keywords, and the match type decides how loosely Google interprets them. Pick one below for the keyword waterproof hiking boots.
The ideal keyword has high purchase intent, high volume and low competition. In practice you get two of three at best, and your job is to pick which one to give up.
| Keyword | Purchase intent | Volume | Competition |
|---|---|---|---|
| running shoes | medium | high | high |
| buy nike pegasus 41 size 10 | high | low | medium |
| how to start running | low | high | low |
| wide toe box trail shoes womens | high | medium | low |
| emergency plumber el dorado hills | high | low | low |
In which you write fifteen headlines, and a machine decides which three anyone reads.
Echo the searcher's words. Matching terms stand out, and relevance feeds Quality Score.
Why you and not the three ads beside you? "Free returns." "Installed tomorrow." "From $19."
Tell them the next step: Shop, Book, Get a Quote, Call Now.
Extra lines like sitelinks, callouts and phone numbers that make the ad bigger and more useful. More in VII.
Responsive search ads (RSAs) began as an optional feature and are now the default way to write a Google search ad. You supply up to 15 headlines (30 characters each) and up to 4 descriptions (90 characters), plus two optional display paths (15 characters). The final URL has no limit, because tracking codes get long.
In which the ad stops looking like an ad: a product photo, a pin on a map, a button that dials a phone.
Text ads are only the beginning. Each format below answers a different kind of search.
In which the highest bidder doesn't always win, and the winner doesn't pay what they bid.
Every search triggers an instant auction. Position is decided by Ad Rank, which, in the textbook model, multiplies what you bid by how good Google thinks your ad is.
How likely people are to click your ad for this keyword.
How closely the ad matches what the searcher meant.
Whether the page delivers what the ad promised, quickly.
Quality Score runs 1–10 and is measured for each keyword, not each ad. And you don't pay your bid. You pay just enough to beat the advertiser below you:
You set a maximum you'll pay for each click.
Also "cost per action." You set a target cost per conversion; Google converts it into a CPC bid for each auction.
Google's algorithms set every bid to hit a goal you choose.
Automated bidding defaults to maximize conversions within your budget. You can instead optimize for conversion value, clicks or impression share. The catch: these algorithms are data hungry. A small advertiser with a handful of conversions a month may wait a long time before automation beats a careful human.
In which eBay turns off its ads to see what happens, and the dashboard turns out to have been taking credit.
Everything so far assumes the ads work. Here's the harder question: did the ad cause the sale, or would it have happened anyway?
That's a hypothetical but realistic eBay report for searches of the word "eBay." It looks like a great deal. Then eBay's economists ran an experiment: they turned the brand ads off.
The searcher typed "eBay." They were going to eBay. The ad just stood in the doorway and collected a fee.
eBay then switched off non-brand ads (think "used camera") in some regions but not others. Look at how the estimated return on ad spend changes depending on the method used to measure it:
Ready to buy, but the site didn't persuade them. This is who conversion rate optimization is for.
Wrong audience. Tighten keywords and add negatives.
Future buyers. Remarketing and RLSA exist to reach them later.
Now flip the eBay problem around. Some ads look unprofitable in the dashboard because the payoff shows up later, under a different name.
Hypothetical hotel-chain results for generic searches: 5,218 clicks, $22,542 spent, 84 conversions at $268 each. Looks awful. But Rutz & Bucklin showed generic ads create later branded searches, which get the credit.
Rutz, Trusov & Bucklin found that 599 of 3,186 traffic-producing keywords led to later increases in direct visits, people typing the URL straight into the browser.
Brand ads get too much credit. Generic ads get too little. The dashboard can't tell you which is which. Only an experiment can.
In the automotive study, the words inside a keyword predicted whether it produced return visits. Read the sign carefully: a negative estimate means more likely to bring people back; positive means less likely.
In which position #1 loses five thousand dollars, and position #4 quietly makes the most money.
Higher positions earn more clicks, but every click costs more. PROSAD (Skiera & Nabout) estimates CPC and click-through rate for all ten first-page positions from just two you've observed, then calculates the profit at each. Change any input.
Four formulas cover everything:
In which you set up a campaign for a real business and bolt ten extras onto a two-line ad.
For each setting, decide as a class, and be ready to defend your choice.
Assets are free to add, make your ad physically bigger, and count toward Ad Rank. Tick each one to add it.
In which six backpacks walk into an auction, and you have to decide which one to fire.
Here's a real-looking keyword report. You can cut one keyword and raise the bid on one. Which? Argue it first, then reveal the cost per conversion. Click a column to sort.