The ten items below are what we open every Google Ads account with, in this order, and the order is load bearing: if conversion tracking is lying, every number underneath it is also lying, and there is no point arguing about match types yet. Each item names the report or screen where you look, what a working account looks like there, and what goes wrong when it is broken. All of it is visible with read-only access.
Nothing here needs a tool you do not already have. It needs you to read the account instead of the dashboard summary, which is a different activity and takes longer.
1. Conversion tracking integrity
What to look for
Open the conversion actions table under Tools, then Conversions. Read every row, not the account summary. Four things: which actions are set to Primary, because only primary actions feed bidding; what each action actually fires on; whether two actions are counting the same event; and whether the count setting is Every or One. A thank-you page view is a fair proxy for a form submission. A form load, a phone link click that never connected, a contact page view, a scroll depth: not sales, and not primary. Then check whether Enhanced Conversions is on and reporting matches, whether calls are counted and at what duration threshold, and whether anything comes back from your CRM through offline conversion import.
Why it matters
The bidder optimizes toward the primary conversion actions and nothing else. If form loads are primary, it buys form loads, and it will find the cheapest traffic in the auction that produces them. Duplicate counting halves your apparent cost per conversion, so the account grades itself as winning while the phone stays quiet. And when no offline outcome ever comes back, the platform cannot tell a tire kicker from a signed job, so it treats them as identical and buys more of whichever is cheaper. That is almost always the tire kicker. Fix this item first or the other nine are being measured with a broken ruler.
2. Brand and non-brand split
What to look for
In the campaigns view, find out whether searches for your own company name are served by a campaign of their own. Then pull the Search terms report, sort by conversions descending, and count how many of the top rows contain your business name or a misspelling of it. If brand and cold traffic share a campaign, they also share a budget, a bid strategy, and a performance number.
Why it matters
Brand traffic converts because the person had already decided. It is cheap, it closes, and blended into one account number it will carry a non-brand half that has bought nothing for months. We have opened accounts where the account-level cost per conversion looked defensible and the cold half had produced zero. Brand clicks also drag the average CPC down, which hides how expensive a cold click has actually become. Give brand its own campaign and its own budget, and read it on its own line.
3. Search term waste mapping
What to look for
Search terms report, over a window long enough to have volume, with cost, conversions and cost per conversion as columns. Then go row by row. Every query lands in one of three buckets: keep, because someone searching that could buy from you; negative, because they never could; or new ad group, because they could buy but are being answered by a keyword and an ad written for something else. Sum the cost column of the negative bucket. That dollar total is the finding.
Why it matters
Waste stated as a percentage is an abstraction. Waste stated as a dollar figure attached to named queries gets fixed the same week. The mapping also produces the two things the rest of the audit needs: the negative list for item 4 and the ad group gaps for item 6. Note the limit while you are in there. The report will not show you every query you paid for; a share of spend sits in aggregated rows you cannot inspect. That is an argument for tighter match types, not for trusting the part you can see.
4. Negative keyword structure
What to look for
Shared library, then Negative keyword lists, and then the campaign and ad group level negatives underneath. Ask whether there is a shared structure at all or just negatives added one at a time wherever someone happened to be looking. Check that your brand terms are excluded from the non-brand campaigns so the two cannot cannibalize each other, and that the brand campaign is not quietly absorbing cold demand. Then check the negatives for damage: a broad negative that eats a converting query, a single word like “free” blocking “free estimate” on a service that offers exactly that.
Why it matters
A structure is a decision about which campaigns inherit which exclusions, and it is the only version that survives a new campaign being added. One-off negatives do not propagate, so the same waste reappears in another ad group next month. And without a brand guard on the non-brand campaigns, the non-brand campaign wins brand auctions, takes credit for conversions it did not create, and the read you did in item 2 becomes false again.
5. Match type drift
What to look for
Search terms report again, this time with the Keyword column showing, so you can read each query against the keyword that triggered it. Measure the distance. Exact match no longer means identical; close variants cover misspellings, plurals, and paraphrases with the same apparent intent, and the system decides what “same intent” means. Phrase drifts further. Broad drifts furthest, and it drifts toward whatever your conversion signal has taught it to like. Also check the keyword table for broad match added quietly over time.
Why it matters
This is item 1 showing up a second time. Broad match is a request for the platform to expand your reach using its model of your customer, and that model is built from the conversion signal you send. A clean signal makes broad useful. A signal built on form loads makes broad an efficient way to spend more money faster in the wrong direction. Drift is gradual, which is why it goes unnoticed: no single day looks wrong, and then a quarter later half the spend is on queries nobody would have approved.
6. Responsive search ad quality
What to look for
Open each active responsive search ad and read the asset table, all fifteen headline slots and four descriptions. Three separate checks. Pinning:count how many headlines are pinned to position 1. If nearly everything is pinned, you have a static ad wearing an RSA’s clothes and there is no test running. If nothing is pinned, work out whether a required message, a license number, a service area, a claim that has to be there for the ad to be true, can be dropped from any given serve. Dynamic keyword insertion: look for {KeyWord:Default} with a bare or missing default, and for the same insertion used in several headlines so the served combination repeats itself. Ad strength: note the rating, then set it aside.
Why it matters
Over-pinning freezes the experiment you are paying to run. Under-pinning ships combinations that leave out the thing you legally or commercially had to say. A keyword insertion with no usable default either shows a customer the raw keyword text or fails to serve, and neither is recoverable after the fact. As for ad strength: it is a completeness score for the asset set, measuring quantity and variety, not sales. Chasing “Excellent” pushes writers toward headlines vague enough to combine with anything, which is the opposite of the specificity that closes. Rate the ad on whether it answers the query and earns the click. Let the strength label land where it lands.
7. Landing page match
What to look for
Take every active final URL and click it the way a customer would: on a phone, on cell data, cold. Does the first screen name the thing the query asked for, in the words the query used? How many actions does the page offer, and is one of them obviously the point? Check the Final URL column for redirects, dead pages, and ads pointing at the homepage. Then line the URLs up against the query map from item 3 and ask whether each query is landing on the page it deserves.
Why it matters
The auction charges you for the click. The page decides whether the click was worth buying. A page carrying five competing calls to action converts on none of them. A homepage answering a specific query hands the visitor the job of sorting, and on a phone, on cell data, most of them decline. This is the cheapest item on the list to fix and the one most often skipped, because it lives outside the ads account and nobody owns it.
8. Budget pacing
What to look for
Campaigns view, status column, looking for Limited by budget. Then segment by day of week and hour of day, and look at spend across the days of the month. Read where the budget actually sits: how much of the daily total is landing on brand and how much on cold demand.
Why it matters
Limited by budget on a campaign that converts is money you declined to make. On a campaign that does not convert, the same label is a mercy. Budget concentrated on brand is buying customers who were already coming, which makes the account look efficient and grows nothing. Uneven day-parting shows up as leads arriving when nobody is there to answer the phone, which reads as a lead quality problem and is not one. And a budget that runs dry before month end means the last stretch of every month is invisible in your own data.
9. Bidding strategy fit
What to look for
Campaign settings, the bid strategy column, alongside conversions per campaign per month. For target CPA and target ROAS, compare the target against what the campaign has actually delivered. Note any campaign whose strategy has been changed recently, and how recently.
Why it matters
Smart bidding is a function of two inputs: the quality of the signal and the volume of it. Maximize conversions pointed at a broken signal buys bad conversions faster and more reliably than any human could. When a campaign produces a handful of conversions a month, target CPA has almost nothing to learn from and its behavior swings. In that case manual CPC, or maximize clicks over a tightly mapped query set, is the honest choice until real volume exists, and saying so out loud is usually unpopular. Changing strategy also resets learning, so a history of strategy churn is itself a finding and sends you straight to item 10.
10. Change history review
What to look for
Change history, widest date range the account will give you, filtered by user. Who edited this account, when, and what did they touch. Look specifically for changes attributed to the platform rather than a person, and check the auto-apply setting on recommendations. Then read the sequence rather than the individual rows.
Why it matters
The account you are looking at is the sum of every edit anyone made to it, and change history is the only screen that explains why it is shaped this way. Auto-applied recommendations add keywords, broaden match types, and raise budgets without anyone deciding to, which quietly undoes items 4 and 5 while you are working on them. A long run of edits by several people with no stated reasons tells you the account drifted rather than being managed.
How we run it
Inside Ad Rescue this list is not a checklist someone works through when there is time. It is a fixed 14 day procedure with the items assigned to days: instrumented pull first, then query mapping, then brand separation, then the negative structure, then RSA rebuilds, then the landing pages and the findings review. The price is $2,500, published before the first call. The defect list is yours on day 14 whether you hire us afterward or hand it to somebody else.
That is the argument for doing item 1 before anything else: a dashboard number measures whatever you told the platform to count.
Where to start if the list is long
Run items 1, 2 and 3 first, in that order. Between them they tell you whether the account measures the right thing, whether its headline number is real, and how many dollars are going somewhere no customer lives. If you would rather have someone else read them with you, the 15-minute triage call is a senior analyst working down this same list on your account. Read-only access is optional and gets you live findings while we are on the phone. There is no pitch deck, and you leave with two or three prioritized fixes whether or not anything else follows.