Most Google Ads audits stop at the first defensible answer. Scale the cheap campaign. Add more negatives. Fix the landing page copy. Each one sounds right. Each one would have made things worse. The difference between a checklist and an investigation is being willing to reject good answers until you find the right one.
Businesses don't experience Google Ads problems. They experience business problems. Google Ads is just where some of them happen to surface.
Most Google Ads audits start with a checklist. Ours start with a complaint.
A B2B SaaS recruitment platform came to us with two frustrations. Their cost per lead was climbing. And their sales team had stopped trusting the leads — most of what came through, they said, weren't real buyers.
Notice what isn't on that list. They didn't mention Quality Scores, match types, conversion tracking, or bidding strategies. Those are the things an audit checklist looks for. They aren't the things a business feels. A business feels symptoms — rising cost, falling quality. The job of an audit isn't to explain the account; it's to explain the business — to trace a symptom back to its cause, and only then to a decision.
Two Audits That Look the Same
A checklist audit and an investigative one look almost identical from the outside. Both end in a list of recommendations. But they get there differently.
A checklist audit runs in one direction: Google Ads account → run the checklist → produce a report. It tells you what exists. Tracking is set up or it isn't. Keywords are broad or exact. It's an inventory.
An investigation runs like this: Complaint → evidence → hypothesis → more evidence → contradiction → new hypothesis → verification → decision.
It doesn't ask what exists. It asks why the business is feeling what it's feeling — and it stays willing to be wrong on the way to finding out.
Here's how that played out on one account.
Step One: Is the Complaint Even True?
The first question isn't "what's wrong." It's "is the thing they told me actually happening?"
We started with the cost complaint, and it checked out. The core of the account — the large Exact Match campaign that does the real work — had gone from ₹2,868.26 per conversion in the three months before the pause to ₹3,509.73 after the relaunch.
But the account-level dashboard made it look milder than that. Blended across everything, the post-relaunch cost per conversion read ₹2,731.18 — lower than the core campaign's ₹3,509.73, because a second, much cheaper campaign was averaging it down. The complaint was true; the headline number was hiding how much had moved, and where.
The lead-quality complaint was real too, and harder to see. Month to month, the share of conversions that were genuine, buy-ready prospects was shrinking, and swinging hard from one month to the next — while the money it took to get them stayed roughly the same. By the sales team's own count, only about one in five or six of the "leads" was someone who'd actually evaluate, negotiate, and buy. The cost of a real prospect was quietly deteriorating; the cost of a raw form-fill rose only modestly. The team felt the first number. The dashboard only showed the second.
Step Two: Find the Driver, Not the Metric
The largest campaign had been ramping up. It was spending around ₹4,000 a day and bringing in conversions at a rate that looked healthy, so they did the logical thing: they raised the budget — about 2.5×, to roughly ₹10,000 a day. When something looks like it's working, you feed it.
But the leads barely grew, and the good leads didn't grow at all. Spend at 250% of where it started; results essentially flat. That gap — paying far more for the same outcome — is the "explosion" they were feeling, even though the blended account dashboard made the cost look milder than it was.
Underneath, the mechanics were straightforward once we looked. Cost per click had risen and conversion rate had fallen at the same time, because the campaign had roughly doubled its share of the available auctions. And it had done that because it ran on a bidding strategy with no cost ceiling — told to maximise conversions and handed more budget, it simply bid higher and reached further down the demand curve into weaker clicks. Nobody had set a limit, so it never found one.
That's a cause, not a metric. The rising cost wasn't the algorithm misbehaving or the competition heating up. It was a setting.
What the Cheaper Campaign Was Actually Buying
That second campaign — the one quietly averaging the blended number down — converted at about ₹1,207.93, roughly 66% cheaper than the Exact Match campaign on the dashboard. The obvious question — in any audit, checklist or not — is what makes it so efficient. You answer that by reading the search terms, so we did.
Close to nine in ten of its conversions came from people searching for tools to help job-seekers — résumé builders, interview-prep aids, and the like. Candidates polishing themselves for the market. None of them were going to buy recruitment software.
The cheapest campaign in the account was cheap because it was buying the wrong people. Its low cost per lead wasn't efficiency. It was the price of irrelevance.
We'd gone looking for the account's best campaign. We'd found its cheapest mistake.
And scaling it — the move the numbers alone would have pointed to — would have looked like a win on every dashboard for months without solving the business problem.
Before moving on, we checked one more thing: was that junk also dragging down the expensive campaign — were the two feeding off the same bad signal? The timeline ruled it out. The big campaign's cost had started climbing weeks before the cheaper one ramped up. Two separate problems, not one. We'd half-expected a link; there wasn't one, so we set that theory aside.
A Hypothesis We Expected to Confirm, and Didn't
If unwanted traffic is converting, the reflex is to block it with negative keywords. So that was the next hypothesis: the negatives were missing.
They weren't. The team had been adding them diligently for months. Blocking wasn't the gap. The gap was structural — a few broad keywords kept generating new variations of unwanted searches faster than anyone could list them by hand. You can't out-type a broad keyword. The fix was the keyword architecture, not a longer block-list.
The Cost Had One More Layer
After the search terms, the obvious next stop is the keywords. In the large campaign they were clean — relevant, high-intent, exact-match terms. Nothing to fix there. But their Quality Scores sat between 2 and 5 out of 10, and that's where it got interesting.
Quality Score is easy to dismiss as a vanity number. It isn't. Google effectively charges you more per click when it's low — by most estimates a 2/10 keyword can cost several times what the same click would at a 7 or 8. So even on perfectly relevant keywords, low scores were quietly inflating the cost per click, and with it the cost per conversion on the dashboard. Another layer under the rising cost, invisible unless you went looking for it.
The natural question was why. Our first guess was ad relevance — maybe the ads weren't matching the searches well enough. Google's own breakdown ruled that out: ad relevance was fine. The component flashing red was landing-page experience, below average on every one of the poor-scoring keywords. That narrowed it. The trail pointed at the pages.
And it wasn't account-wide. The client also ran a small brand-defence campaign, and those keywords scored 9 and 10. The capability was clearly there; whatever was dragging the non-brand keywords down was specific, not systemic — usually a good sign you're close.
The Pages Were the Surprise
So we opened the landing pages, half-expecting weak copy. The copy was good. The messaging was clear, the keywords showed up naturally in the page, and each one read like a genuine continuation of the ad that brought you there. On message, they'd done the work — which made the low scores more puzzling, not less.
Then we ran the pages through PageSpeed Insights, and the mobile experience was bad enough to fail Google's thresholds outright. There it was.
The fair objection: why does mobile matter when 81% of the spend and about 70% of conversions came from desktop? Because Google leans heavily on the mobile experience when it judges landing-page quality — so a slow mobile page drags the Quality Score, and inflates the cost per click, on all the traffic, desktop included. After getting the keywords, the messaging, and the ad-to-page match right, one technical gap was quietly taxing both the cost and the lead quality across the account. The fix wasn't to rewrite anything. It was the page speed.
Why It's Easy to Stop Too Early
The honest part is that at several points, we thought we had the answer. Each time, we could have stopped, written a perfectly reasonable recommendation, and been wrong.
- Stop after the campaign data → scale the cheaper campaign. Wrong — it's buying job-seekers.
- Stop after the search terms → add more negatives. Wrong — they already exist.
- Stop after the pages → rewrite the copy. Wrong — the copy's fine; the page is slow.
Each one is defensible on its own. Each is what a thorough checklist would produce. And each is incomplete. None of them feels wrong while you're writing it — which is exactly why stopping early is so easy to do.
Looking Beneath the Symptoms
Step back far enough and a familiar pattern shows up — the kind a doctor uses when a patient says "my chest hurts." The complaint is real, but it's the top of a ladder, and you don't treat the sentence. You work down it. A campaign issue? Partly. The bid strategy? Partly. Traffic quality? Partly. Keep going.
Only after all of that did a pattern begin to emerge — a single thing sitting underneath the rest. The account had gradually become optimised, at almost every level, to produce form submissions — and it was doing that well. The bidding maximised them. The keywords cast wide to catch them. The conversion signal treated every one as identical. No one set out to build it that way; a lot of individually reasonable decisions had simply accumulated into a system pointed at the wrong outcome. The two complaints — rising cost, falling quality — were just what that looks like from the outside.
The Output Wasn't Findings. It Was Decisions.
Diagnosis has no value on its own. Its only job is to improve the decisions that follow. An audit that produces findings has done half the work — findings are observations. The shift happens when each one turns into a decision, and the decision is often the opposite of what the finding first suggests.
| Finding | Why It Mattered | What We Decided |
|---|---|---|
| The largest campaign scaling with no cost ceiling | More budget just bought pricier, weaker clicks | Put an efficiency ceiling on it — don't just cut it |
| A second campaign with a far lower cost per lead | The low cost came from buying job-seekers, not buyers | Don't scale it |
| Unwanted traffic still converting | A broad keyword spawned variants faster than blocking could keep up | Rebuild the keyword architecture, not the block-list |
| Non-brand keywords at Quality Score 2–5 | Low scores quietly inflated the cost per click on relevant terms | Treat it as a cost problem, not a vanity metric |
| Pages converting poorly | The copy was fine; the slow mobile page dragged QS on all traffic | Fix the page speed, leave the copy |
| Zoho's brand term on the negative list | Plausibly the most efficient real demand in the account | Re-examine it before keeping it blocked |
| Every form-fill counted as success | The system was optimised for the wrong outcome | Redefine the success signal around qualified leads |
Read the left column and you have an audit. Read the right column and you have a plan. The line between them — where diagnosis becomes decisions — is the part of this work that doesn't automate.
What We Didn't Recommend
It's worth being plain about what we didn't suggest, because it's the mirror image of that table. We didn't add more negative keywords. We didn't move budget to the cheapest campaign. We didn't chase a lower cost per lead. We didn't rewrite the landing pages. We didn't pin the rising cost on the algorithm or the competition.
Every one of those would have been an easy line to put in an audit. The things you decide not to do are often where the actual thinking shows up.
The Part That Mattered
Looking back, the moments that mattered weren't the findings. They were the moments the evidence made us put down an answer we were already holding — the cheaper campaign we nearly scaled, the negatives we nearly added, the copy we nearly flagged. The findings were just what we gathered along the way. The judgment was in being willing to change our minds.
All of it points to one idea: you shouldn't measure marketing by the number of forms submitted. You should measure it by the number of businesses it actually creates.
We try to hold ourselves to that. The company whose audit you've just read first found us through one of our own Google Ads campaigns.
AI has made reports abundant. It has made analysis inexpensive. It has made checklists nearly free. What it hasn't made is judgment.
Judgment is deciding which evidence matters. It's knowing when to abandon a good hypothesis. It's recognising that the cheapest campaign was the wrong campaign. And it's knowing which recommendation not to make.
That's the work.
Frequently Asked Questions
- What's the difference between a checklist audit and an investigation?
- A checklist audit inventories what exists in an account (tracking setup, keyword match types, Quality Scores). An investigation traces business symptoms (rising costs, poor lead quality) back to their root causes through hypothesis testing and evidence gathering.
- Why does mobile page speed affect desktop ad costs?
- Google heavily weights mobile experience when calculating Quality Score. A slow mobile page drags your Quality Score down, which increases cost-per-click across all traffic—desktop included—even if most of your conversions come from desktop.
- Should I scale campaigns with the lowest cost per lead?
- Not automatically. Low cost per lead can mean efficiency, or it can mean you're buying the wrong audience. Always check the search terms and lead quality before scaling based on cost alone.
- How do I know if I'm optimizing for the wrong conversion signal?
- Compare your conversion volume in Google Ads to qualified leads in your CRM. If Ads reports 200 conversions but Sales says they got 12 qualified prospects, your conversion definition is treating all form-fills as equal when they're not.