PPC Consulting

Where the audit stops and the management starts

These are two different purchases and it is worth being blunt about the line between them, because a lot of paid search selling depends on blurring it.

An account audit is a one-off diagnostic. It has a defined start and end, nothing in the account changes because of it, and it finishes with a findings document you own outright. You can hand that document to whoever manages the account now. That is the whole point of it.

PPC consulting is the ongoing work: hands in the account, decisions made week to week, structure rebuilt where it needs rebuilding, negatives added, budgets moved, campaigns paused and tested. It is a retained engagement because the work it consists of never finishes. The search terms report refills every week whether anyone reads it or not.

Most buyers should start with the audit. It is cheaper, it ends, and it tells you whether the account needs a manager or just needs three things fixed once. If you already know what is wrong — you have the diagnosis, you simply have nobody to act on it — the audit is a purchase you can skip.

The most expensive problem in most accounts is not the bidding

It is that the conversion data the bidding runs on is wrong.

Automated bidding optimizes toward whatever you have told Google is a conversion. It has no independent way to know whether that thing is worth money to you. If the conversion action is wrong, every bid decision built on it is wrong, every reported return is fiction, and the more budget the system is given, the more confidently it spends into the error.

The three failures that keep appearing:

• Double counting. A conversion tag hardcoded on the thank-you page and the same event imported from analytics, so one form fill reports as two. Cost per conversion halves overnight and nobody questions it because the number moved the right way.
• Firing on the wrong event. Counting a page view, a visit to the contact page, a form open rather than a form submit, or a phone number click that nobody ever completed. These generate volume, which makes the automation look well fed, and none of them are outcomes.
• Never verified after a change. A site migration, a new consent banner, a redesigned checkout, a tag manager container someone cleaned up. Tracking rarely announces its own death. It just starts reporting a smaller or larger number that gets read as performance.

So the first month of any engagement is spent proving the conversion actions represent things you would be happy to pay for, that each one fires once, and that the definition matches what the business actually sells. Until that is true, no bidding conversation is worth having.

What a month actually consists of

Not all of this every month. The account decides the order.

• Search terms against keywords. What the account is actually being bought against, versus what you think you bought. This is where waste is visible in dollars rather than in theory.
• Negative keywords. Coverage at account, campaign and ad group level, and whether the lists are maintained or were written once at launch and abandoned.
• Structure. Whether campaigns and ad groups are still segmented the way the budget needs them segmented, and whether the same query is eligible in three places at once.
• Budget pacing. Campaigns exhausting budget before the hours that convert; campaigns underspending against a cap that was set for a different season.
• Responsive search ad assets. Headline and description coverage, and whether the assets are pulling in enough directions to be worth testing.
• Performance Max configuration. Asset groups, search themes, brand exclusions, final URL expansion.
• Geographic and schedule targeting. Where and when the money goes, which is the setting most often left at whatever it was on day one.
• Landing page and offer feedback. Paid search buys the visit. It cannot fix what the visit lands on, but it can tell you precisely where the page is losing people.

Broad match, and the trade you are making when you use it

Broad match is the default match type in Google Ads. Google's documentation is direct about what it does: ads "may show on searches that are related to your keyword, which can include searches that don't contain the direct meaning of your keywords."

The signals it uses are worth knowing because one of them surprises people. Google names the user's recent search activity, the content of the landing page and assets, and other keywords in the same ad group, which it uses to infer what the ad group is about. That last one is the mechanism behind a practical rule: a broad match ad group has to be coherent in intent, because a stray keyword changes how every other keyword in that group matches.

Google also states plainly that "it's critical to use Smart Bidding with broad match," because each query differs and needs bidding that reacts at auction time. Reference: Google's documentation on broad match.

The honest reading: broad match paired with automated bidding does work, and it is what Google recommends. It also transfers the selection of queries from you to Google, which leaves the search terms report as the only remaining lever you hold. Broad match without automated bidding, or broad match without somebody maintaining negatives every week, is the textbook wasted-spend pattern and always has been.

Performance Max, and the campaign that quietly eats brand traffic

Performance Max runs one campaign across the full Google network — Google documents the inventory as YouTube, Display, Search, Discover, Gmail and Maps. Creative is organized into asset groups, described as "a collection of creatives centered on a theme or target audience." Audience signals are inputs to that, not targeting: Google combines them with its own real-time read of intent.

The mechanic that causes the most trouble is the priority rule. Exact match keywords in Search campaigns take priority over Performance Max. Phrase and broad match keywords are prioritized comparably to Performance Max search themes.

Follow that through. If your Search campaigns are built mostly on phrase and broad match — which is what most modern accounts are told to do — then Performance Max is competing on equal footing for your highest-intent and branded queries, winning a share of them, and reporting those conversions as its own. The campaign looks like the best performer in the account because it has been handed traffic that was already going to convert.

Brand exclusions exist specifically to control this, and campaign-level controls also cover search themes and final URL expansion. The test is straightforward: compare branded query volume and cost before and after Performance Max launched, then set brand exclusions and watch where the conversions reappear.

Structure is a cost control, not a tidiness exercise

Match type sprawl is the condition where the same query is eligible across several campaigns and several match types at once. The account competes with itself, the reporting stops attributing cleanly, and no single campaign's numbers describe what actually happened.

The related failure is automated bidding running on too little data. These strategies need conversion volume to learn from; below a workable threshold they are optimizing noise, and the account develops the characteristic pattern of large swings in cost per conversion with no change in anything real. Sometimes the correct recommendation for a small account is manual control and a much simpler structure, which is not what most people expect to be told.

Structure work is unglamorous and it is where recoverable money usually is. Consolidating campaigns that were split for reporting rather than for budget control, separating brand from non-brand so the two are not averaged into one meaningless figure, and putting the query in exactly one place where it can be bid on deliberately.

How this is bought, and what makes one engagement cost more than another

Retained, monthly, no fixed term. You should be able to leave, and an engagement that depends on a lock-in to survive was not working.

What drives the number, so you can locate yourself on it without a published figure:

• Account and market count. One account in one country is a different job from six accounts across four languages with separate budgets.
• Campaign types in play. Search alone is the simplest. Shopping adds a product feed, which is a data quality problem as much as an advertising one. Performance Max adds asset production and configuration oversight.
• Conversion complexity. An e-commerce transaction is self-evident. A lead that has to be qualified downstream, or a sale that closes offline weeks later, requires a measurement build before the bidding can be trusted.
• Whether anyone can change the landing pages. If the pages are frozen, part of the diagnosis is unusable and the account is being asked to carry more than it can.
• Reporting expectations. A monthly written review is standard. Board-facing reporting, or reconciliation against a CRM, is additional work.

Not included: creative production at scale, landing page development, and anything that requires committing to a position or a volume in an auction nobody controls.

When you should not put someone on retainer, and how you will know it worked

Do not buy this if the spend is small enough that the fee consumes the waste it would recover — at that size, buy the audit, act on it yourself, and revisit in a year. Do not buy it if the account is a single well-behaved campaign that has converted steadily for two years; you need somebody to look at it twice a year, not every week. And do not buy it if the real problem is the offer or the landing page, because paid search will faithfully deliver traffic to a page that does not convert for as long as you fund it.

Judging it: cost per qualified outcome is the number, and it should be read against the volume of those outcomes, not on its own. Clicks, click-through rate and impression share are diagnostic instruments, not results. Expect the search terms report to get cleaner and the share of spend against queries with no plausible intent to fall — that one is directly measurable and is usually the first thing to move.

Paid search does not need months to produce numbers the way organic does. It does need enough conversions after a structural change for the comparison to mean something, which in most accounts is four to six weeks. Anyone reading a three-day trend as a result is reading noise.

Frequently Asked Questions

Published 2026-08-11