107 checks, named. What every audit looks for.

A primary conversion action that has recorded nothing lately. Smart Bidding has nothing to learn from it, and a broken tag may be hiding real results.

A lead action set to count every conversion. Google recommends every conversion for sales and one for leads, so a repeated form fill flatters the cost per lead.

Lead or call values that are a number somebody typed, standing in for what a customer is worth. Fine as a start, but every value-based decision rests on it.

A purchase action set to always use its default value, so a €15 order and a €900 order weigh the same. Any value bidding on it is bidding on a count.

Ads sending paid clicks to pages that answer 404 or a server error. Common after a site rebuild, since nobody tells the ads account a URL has gone.

Primary conversion actions whose click window is one day or less, so a customer who comes back the next day is never counted. Where Oppy cannot fix it, the finding says what does.

Two primary actions recording the same result, which may double every conversion. Reported with both names, as a question, because only someone who knows the site can say which one is real.

Every action that fired uses a default value, so conversion value rises and falls with the count. Every ROAS built on it is running on a count.

Conversions that keep arriving days after the click. Lag belongs to the business, from seconds for a call to weeks for a B2B lead. The fault is judging last week as finished.

Form fills, contacts and calls summed into one conversions figure. All real, with different costs and close rates, in a single number every ratio divides by.

Enhanced conversions switched off while the customer data terms are already accepted. The slow legal step is done and the toggle takes a minute; the conversions it recovers are business already won.

How much of the result comes from people who had already bought. On a mostly repeat account, the cost per conversion is the cost of an order, not of a customer.

A shop that records only the purchase. It cannot tell a visitor who never wanted the product from one who quit at payment. Often the events already exist in Analytics and only need importing.

Purchases that arrive with no amount, read over a longer period than most checks. A test cannot be judged on sales value, and value-based bidding has nothing to aim at.

The account optimises for a form, but the page the ads send people to has no form. Every metric looks like a media problem until someone opens the page.

Two actions in the same category that both recorded in the same window. Asked as a question with both numbers, because the same account can report very different sales depending on which action is primary.

When an account has too little traffic or no conversion action, every ratio is noise. The audit says so and reads the setup instead of printing a cost per conversion nobody should trust.

Two definitions of a phone call in one account, such as 180 seconds on the action that drives bidding and 60 on the others. The finding is the disagreement.

A sales or call import that exists, is a primary goal and has received nothing. Somebody built the pipe and nobody sent it a row.

Keywords Google rates well and shows, that have not spent in months. Demand the account is not buying. Always a person's budget decision, never automated.

Keywords that used to convert and are now switched off. Never restarted on its own: a winner can be off for a good reason, so the finding reports and asks.

A Display or Video campaign with no frequency cap, so a few heavy users can absorb hundreds of impressions each.

A Search campaign also serving on Search Partners or the Display network. Both are on by default, convert differently, and blur the campaign's numbers.

Thirty conversions a month split across ten campaigns is three each, and none of them can bid well. Flags a budget divided past the point where any campaign can learn.

A country where people search in several languages, targeted in only one, such as a Belgian campaign in French that never reaches people who search in Dutch. Asked as a question, since one language can be a choice.

A listing group that shuts most of the catalogue out. A product excluded from every group can never show, whatever its margin.

A negative keyword that switches off a keyword you chose. The keyword still says enabled and still shows a Quality Score, and its own searches never reach it. Checked against every negative in force.

A call-driven account sending clicks to a page with no tappable phone number. Most visitors on a phone will not copy a number by hand.

AI Max switched on for a Search campaign, which widens what the campaign matches beyond its keywords. Oppy can switch it off on one campaign once you approve.

An account with enough volume to run a test that has never run one. Only fires when the account is big enough to reach an answer in reasonable time.

Asset groups with no search theme, the one place you tell Performance Max in words what the group is for. Audience signals do not replace it.

Broad match running with no negative keywords at all. Broad match lets Google decide what a keyword means, and negatives are the only place the account can disagree.

Location set to people 'in or interested in' a place, Google's default. Someone abroad reading about your city counts as your market, and with no location at all a campaign serves worldwide.

Competitor names you confirmed that no keyword covers. Only the confirmed list is used, because history alone cannot tell a rival's name from a word your own product needs.

The account counts store visits or directions, yet no ad shows an address. Linking the Business Profile as a location asset is the missing half.

An account that counts walk-ins while its campaigns target a country, a city or a postcode rather than a radius round the shop. A radius is the clearest way to say 'people near us'. Asked, since many shops also ship nationally.

Google's auto-applied recommendations switched on, with what the change history credits to them. Each subscription is listed.

Optimized targeting on, which lets Google go past the audiences you chose. It is on by default, so a carefully built audience and no audience can end up in the same place.

Keywords that have stopped getting impressions. None is paused: Oppy finds why each one never shows. Keywords added recently are named apart, since they have not had time to show.

Landing pages that are slow on a phone, measured on mobile, with what the account paid to send people to each one. Not priced as a saving, since nobody knows how many left because of the wait.

Local Services ads and Search campaigns running together, with both spends side by side. They can compete for the same search, and which should win is a business decision, since one is paid per lead.

No audience list attached to any campaign, even in observation. Nothing breaks, but the account cannot tell a returning visitor from a stranger and bids the same for both.

A campaign with no language set serves to every language Google has an audience for, so you can pay for clicks from people who cannot read the ad.

No list separating people who already bought. Without it you cannot stop advertising to yesterday's buyer, or measure what a new customer costs.

No Search keyword contains the brand, so a competitor can sit above your result, often for less than you would pay. Judged on the Search keywords.

No placement or content exclusions anywhere outside Search. Sensitive content categories are counted too.

Compares what each match type pays for a click inside this account, never against an outside benchmark. A finding to read, since moving a keyword's match type is a removal and an addition.

Performance Max matching queries your Search campaigns already bid on. The credit moves to Performance Max, and Search looks worse for work it used to do.

Performance Max with no negative keyword list attached. The list the account keeps for Search is usually never attached here.

Performance Max with no brand exclusion list. Left alone it wins your own name cheaply and books those sales as its own, while brand Search falls.

Portfolio bid strategies nothing uses, including ones on strategies Google has retired. They cost nothing, which is why they stay, and the next campaign attached to one may get the wrong one.

Products limited by an error or a warning in the feed, such as a missing colour, gender or age group. They still sell, in fewer places, and the regions affected are named.

Products marked not eligible, which never enter the auction and so never appear in any report. Named so someone fixes them in Merchant Center.

Remarketing lists that exist but hold too few people for Google to serve, such as basket abandoners at zero. The tag that fills them may be missing or not firing.

Searches that already convert with no keyword of their own, read over a long period so a search that converts now and then still counts. The one check that says the account is missing something.

A campaign serving around the clock for a business that answers a phone. The unanswered call costs its click and teaches the bidding that the keyword does not work.

Display and YouTube campaigns that exclude no audience list at all, so they can keep reaching people who already bought. Says whether the account has a list it could exclude.

The share of the catalogue that got no impression lately. Reported as a fact: some is discontinued, some overpriced, some the next thing to sell.

The same keyword live in two campaigns, so the account competes with itself for the same search and the reporting splits it between two places.

Who has admin or standard access to the Google Ads account, and since when. The list itself is the finding: an agency with admin may be right; nobody knowing is not.

A ROAS target set without a known margin. Prints the margin at which that target breaks even, so you see at once whether it is ambitious or quietly losing money.

Automated bidding on the account's own name. People searching for the brand have already chosen it, so Smart Bidding bids to reach its target on average, for clicks that cost far less by hand.

A campaign that runs out of budget on most days it runs: a budget decided once and never revisited. Separate from lost impression share, which measures how much is missed, not how often.

A weekday costing much more per conversion than the others. Judged by day, not by hour, and only on days with enough conversions of their own.

A device paying much more per conversion than the rest of its own campaign. The fix is a bid adjustment, not an exclusion, so the traffic and its measurement stay.

Inside a shared budget, a campaign's limit is whatever the others left, so its impression share lost to budget also depends on the other campaigns. Names the campaigns this applies to before anyone acts on their numbers.

A target CPA or ROAS a long way from what the campaign achieves. Too strict throttles the campaign; too loose means someone manages a number that steers nothing.

A CPA or ROAS target set on a campaign with too few conversions a month to learn from. The bidding holds back for a cost it cannot learn yet. The usual fix is to take the target off and let Maximize conversions learn first.

A Shopping campaign bidding the whole catalogue at one price, so a €4 item and a €400 item compete with the same bid and the cheap ones win the traffic.

One campaign capped by its budget while another has money it does not use and performs worse. Proposes moving budget before asking for more, and only when both sides are true.

Maximize clicks on an account that counts calls or sales. It buys the cheapest clicks, which are the least likely to call. Judged once the account has some conversions to aim at.

A campaign spending real money with no conversion at all. Each keyword can sit under the bar while the campaign is far past it. Zero can mean broken tracking, a page that fails or too little demand, so nothing is guessed.

Hours of the day that take money and return nothing, on accounts with enough conversions to read hour by hour.

The share of spend on searches nobody added or excluded, the ones Google matched on its own. Not waste in itself: one percentage to watch month to month.

How much of the budget buys the account's own name. Brand converts several times better than the rest, so a blended cost per lead describes nothing you can grow. Reported, not judged.

Keywords that spent past the point of proof with nothing back. The bar comes from the account's own numbers, not from one fixed rule. One finding per keyword, so you can agree to four of six.

Impression share lost to budget, on campaigns that convert well enough for the missed searches to be worth buying, with what buying them would cost.

Display money going to phone apps nobody chose, mostly games where a tap is an accident. Names what those apps cost in the period read.

Negatives that let their plural through, because negative keywords do not match close variants. The plural is proposed beside each one.

Countries that took real money across enough clicks to judge and returned nothing, inside campaigns that convert elsewhere. Same arithmetic as the keyword waste check.

Quality Score on non-brand keywords only. Brand terms sit at 9 or 10 by nature, so an account average hides an expensive generic side. The money is attached to the non-brand part.

The three Quality Score parts (expected CTR, ad relevance, landing page experience) below average on the keywords that carry the spend, once they have enough impressions to grade.

Spend resting on searches for other companies' support lines or public services. Budget raises wait until it is cleared.

Searches that took money and returned nothing, judged on the account's own numbers. Your brand is never named, and searches for competitors you listed are proposed as negatives on every Search campaign.

Small searches for things the products never mention, added up by theme. Each search is too small to notice; together they are a share of the spend.

Spend piled into a few products, usually the cheapest ones, because a low price wins more auctions. Shows where the money went and what came back.

One keyword bought in two match types at two very different costs per conversion.

Where a Performance Max budget went across Search, Shopping, Display, YouTube, Gmail, Discover and Maps. Flags a network that took most of the money, so you can check what it brought back.

The placements where Display and video money went, turning 'Display is not working' into a list of places you would or would not have chosen.

A campaign whose results are phone calls, with no call asset on the ad. Judged on what the account counts, so an account bidding for forms gets no advice about phones.

A promotion asset still showing after its offer ended, so the ad keeps advertising a sale that is over.

Ad groups running on one ad. Pausing that ad would take the group off air, and nothing is being tested. Oppy writes the second ad as its own step, never inside the finding.

Ad groups where every ad but one has stopped getting impressions. Google picked a winner and nothing new was offered, so no test is running.

Ads Google has refused to show, and the worse case behind them: an ad group whose only live ad is disapproved, so the group is off air while everything above it says enabled.

Ads using fewer headlines and descriptions than Google can combine, for example ten headlines and two descriptions. Usually the reason behind a weak Ad Strength.

An ad and its landing page that have nothing in common. Only the clear cases are flagged, so taste is never reported as a fault.

An ad test with a real winner, by the arithmetic rather than by eye. Silent on most accounts, which is why it can be trusted when it speaks.

Asset groups Google rates below Good, usually because they lack headlines, images or videos Google could use.

Asset groups with fewer headlines, images or videos than Google can use. A group with four images cannot appear where six are needed.

Responsive display ads holding fewer images or sizes than Google can use. A group with no square image cannot appear in a square slot, and no report shows the gap.

Every video in one length band. Bumpers, skippable ads and longer cuts reach different people, so one band leaves part of the audience unreached.

Callouts, structured snippets and images a Search campaign does not use. Sitelinks have a check of their own. Each makes the ad bigger at no extra cost per click. One row per campaign, not a pile of findings.

Exactly one asset pinned to a position removes every combination Google would test there. Pinning a legal line is fine; this finds the pins that freeze an ad.

Responsive search ads Google rates Poor or Average. The rating measures how much distinct material the ad gives Google to combine, so a low one usually means fewer combinations to show and test. New ads still being rated are left out.

Search campaigns with fewer sitelinks than Google can show. Sitelinks cost nothing extra and make the ad bigger, and Oppy proposes them to your real pages, on the campaign or the account.

Videos Google built from your images and text because the asset group had none, running on YouTube under your name. Read from the asset's source, not guessed.

Videos where too few of those who watched the first quarter reach the end, judged inside the campaign's own numbers.