PPC management is somebody looking after the account week to week: what it is buying, what it is wasting, what the targets should be, and whether the answer this month is to change something or to leave it alone. Smart bidding took the manual adjustments away and moved the leverage somewhere else. This page is what a month of that work actually contains.
8.81 x
Average return on ad spend across the accounts we manage
7
Years the average client stays, with no long-term contract to hold them
2003
The year our first pay-per-click post went up. It is still online.
20 +
Years buying paid search, from Overture and early AdWords onward
Not the pitch version. This is the shape of a normal month once an account is running, including the weeks where the right answer is to change very little.
Every month, what the account actually bought as opposed to what it asked for. It produces negatives to add, and more usefully it produces phrasings real buyers use that nobody in the building would have written down. On a shopping account the equivalent job is product-level performance, because the feed is doing what a keyword list used to.
Negative keywords from the search terms, placement exclusions on anything running against display inventory, and audience or location exclusions where the data supports them. This is the cheapest work in paid media and the least often done, because nothing about the account looks broken without it.
Bidding targets reviewed against what the account is actually achieving, budgets moved between campaigns that are earning and campaigns that are learning, and anything seasonal put in place early enough to have finished learning before it is needed. Sales, promotions and feed changes all need lead time, and the lead time is not optional.
New ad variants where there is enough traffic to tell them apart, asset refreshes before fatigue rather than after it, feed titles and attributes on retail accounts, and a note on any landing page the data says is losing the click after we paid for it.
Spend, revenue, conversions, cost per acquisition and return, each against last month and against the same month last year, plus impression share against the competition. When a number goes the wrong way it is in the report with a reason, and when the reason is that we made the wrong call it says that.
Not a list of best practices. This is what is genuinely there when we get read-only access to an account somebody else has been running, and the order is roughly how often we see them.
Duplicated tags counting one sale twice, a thank-you page firing on a step nobody completes, form submissions counted as sales, or a tag that stopped working during a site change eight months ago. Every optimization decision made after that point was made on fiction, and automated bidding will have been faithfully optimizing toward it.
Not a thin list — very often no list, and a search terms report nobody has opened in a year. This is why an account can drop spend and gain impression share in the same month, which reads like a trick and is just subtraction.
The ad promises a specific thing and the click lands on a general one. It hurts the quality score, so it raises the price, and it loses the visitor, so it lowers the return. Both ends of the same mistake.
A return target chosen at setup and left there through two years of changing margins and competition, or a budget cap that has been throttling the best campaign in the account since a quiet quarter in 2024. Automated bidding will hold whatever line it was given, indefinitely and without complaint.
We do not publish a rate card, because a public price either fits nobody or anchors the conversation to the wrong number. What we can be plain about is the shape of it.
What you pay Google is yours and goes to Google. Our fee is separate and does not scale as a percentage of it, which matters because a percentage fee quietly rewards an agency for spending more of your money.
One number, with the mix of work inside it moving as the data changes rather than as it was scoped in month one. Some months that is mostly feed and product work, some months it is mostly structure and exclusions.
The average client stays around seven years, and the reason we can say that is that none of them have to. A notice period is the industry's substitute for retention, and an agency that needs a year of lock-in is telling you what it expects month nine to look like.
It stays in your name with your billing, and we work in it. If the relationship ends you keep the account, the history and the learning in it. An agency that runs your ads inside its own account is holding the data hostage even if it never intends to use it.
Two passages, quoted as published. Both describe work that is still exactly the work, which is the argument for reading the old material rather than the new.
Paying for irrelevant clicks in your pay per click (PPC) ad campaign is frustrating, but is there anything you can do to minimize the chances of this happening?
The question that opens the second item in the month above. Sixteen years later the answer is still mostly subtraction, and it is still the part of the job that gets skipped when nobody is looking.
Setting up negative keywords by placing a minus sign in front of them lets you filter out identical terms that can have very different meanings.
The interface has changed and the mechanic has not. What has changed is the scale of the problem: broad match and automated targeting reach much further than they did in 2010, so the cost of an unmaintained negative list is considerably higher now than when this was written.
To buy search engine ads, advertisers bid on relevant keyword phrases, whereas content sites usually charge a fixed price per click.
Worth reading next to the first problem on the list above. We were writing about conversion tracking as a management task in 2010 and it is still the single thing most likely to be broken in an account we inherit.
Around 700 posts, published since 2003
Read the archiveWorth reading in this order: how the thing works, who you would be handing it to, and what else is in the paid budget.
The hub: what decides the price of a click, the five paid surfaces, and one real month of an account reported in full including the numbers that went the wrong way.
/ppc/How to choose one, including how to rule us out. The questions we published in 2010 that still work.
/ppc-agency/The per-lead panel above the ads: what verification requires, what really decides your rank, and the leads you should be disputing.
/local-services-ads/Shaun O'Brien
Owner, Selby Acoustics
Shaun was John's client
Phil Leahy
Founder, MedCart Marketplace & Retail Global Events
Phil was John's client
Matthew Kligerman
Building Dirt — AI land development platform
John was Matthew's client
Read-only access is enough, and we will tell you what is in there: whether the tracking can be trusted, what the search terms report says you are actually buying, where the impression share is going, and whether the return you are getting is the return the account is capable of. If the answer is that it is being run well, we will say that too.
