Every underperforming PPC account looks fine from the dashboard. Spend is going out, clicks are coming in, the graphs move. The problem is that clicks are not the product. Customers are.
A PPC audit is how you find the gap between the two. It is not a rebuild and it is not a rescue mission. It is a structured read of where your money went, which parts of it did any work, and which parts you are paying for out of habit.
This guide is the audit we run before we touch anyone's account, in the order we run it. You can do most of it yourself in an afternoon. Three checks explain most accounts on their own: whether the tracking is honest, what the search terms report says, and which location setting you are on.
What a PPC audit is, and the three things it always finds
A PPC audit is a review of an advertising account against one question: which spend produced a business outcome, and which did not. When Google is the only platform running, people call the same job a Google Ads audit. It covers tracking, keywords, campaign settings, bidding, creative and the pages the traffic lands on. Everything that follows is PPC optimization in the honest sense: finding what is broken before changing what is not.
In practice almost every audit finds the same three things. Conversion tracking that is measuring the wrong action, or nothing at all. Search terms nobody meant to pay for. And decisions being made on metrics that look like performance but are not.
That is the whole article. Everything below is how to find them.
The auction, in one minute
Before auditing anything, it helps to be clear about what you are buying. A pay per click campaign is not a price list. Every search runs an auction, and you only pay when someone clicks.
What decides whether you appear, and what you pay, is Ad Rank. Your bid is only one part of it. Google also weighs how relevant your ad and your landing page are to the search. So two advertisers bidding the same amount do not pay the same amount. The more relevant one pays less for the same position.
This is the most useful thing to understand before an audit, because it explains why so many fixes are not bidding fixes at all. Making the ad and the page match the search lowers your cost without touching the bid.
One report worth opening while you are here. Auction insights shows which other advertisers turn up in the same auctions as you, and how often they outrank you. If your costs climbed and nothing changed on your side, look here first.
Start with tracking, not keywords
There is no point auditing performance until you trust the numbers. This is the step people skip because it is boring, and it is the step that invalidates everything after it.
Open your conversions and ask three questions:
Is anything being counted? An account with no conversion actions has nothing to aim at. Every optimisation the platform makes is aimed at clicks, because clicks are all it can see.
Is the right thing being counted? Counting a page view of the contact page is not counting an enquiry. Counting every form submission includes the spam. Google's conversion tracking setup explains the difference between the types, and it is worth ten minutes.
Is it counting once? Duplicate tags are common after a website rebuild, and they quietly double your reported conversions, which makes bad campaigns look profitable.
Two more things while you are in the same screen. Check your attribution model, because last click and data driven credit the same sale to different keywords, and that changes which keyword looks worth keeping. And remember conversions are recorded against the day of the click, not the day of the sale. A change you made three days ago has not finished reporting yet, so judging it today will mislead you.
Fix this before reading further into the account. If tracking is wrong, the rest of the audit is fiction. Our guide to conversion rate optimization covers how to decide which actions are worth counting in the first place.
The metrics that lie to you
Once you trust the tracking, the next job is refusing to be flattered by the wrong numbers. Four do most of the damage.
Click through rate. A high rate means your ad is attractive to whoever saw it. It says nothing about whether those people can buy from you. Ads offering something free reliably win on click through rate and lose on revenue.
Cost per click, or CPC. Cheap clicks feel like winning. Cheap clicks from the wrong searches are the most expensive traffic you can buy, because you pay for all of them and convert none.
Return on ad spend, or ROAS. Useful, but it is calculated on revenue, not profit. A campaign at 4 times return can lose money if your margin is narrow, and a campaign at 2 times can be excellent if your margin is wide.
Impression share. This one looks like an opportunity and usually is not. A low share tempts you into raising budgets to capture the rest of a market that may not want what you sell. Chase it only after the money you already spend is working.
The number that matters is cost per qualified lead, or cost per sale. Everything else is a diagnostic, not a scoreboard.
The five numbers, and how to work them out yourself
You do not need a tool for any of these. Every one is a division you can do on the spend and the results the account already shows you.
| Number | How to work it out | What it is actually for |
|---|---|---|
| Click through rate | Clicks divided by impressions | Judging the ad, not the account |
| Cost per click | Cost divided by clicks | Judging how competitive the auction is |
| Conversion rate | Conversions divided by clicks | Judging the page and the offer |
| Cost per acquisition | Cost divided by conversions | Judging whether the campaign pays |
| Return on ad spend | Revenue divided by cost | Judging ecommerce, on revenue and not profit |
Work each one out at three levels: the account, the campaign, and the single keyword. The account number tells you whether you have a problem. The keyword number tells you where it is.
One of these five is not really a paid search number at all. Conversion rate is decided by the page, not the auction, which is why it is the metric that improves fastest once someone looks at it properly. That is the job of conversion rate optimization, and it usually costs less than raising bids.
Now the part most guides get wrong. People go looking for the average click through rate or the average cost per click in their industry. They find a set of published Google Ads benchmarks and measure themselves against it. Those figures blend thousands of businesses with different margins, different offers and different countries. A cost per click that is ruinous for one business is cheap for another.
Your benchmark is your own account. Take the last 90 days, work out those five numbers, and write them down. That is the line every change from now on gets measured against. An outside average cannot tell you whether you are improving. Your own last quarter can.
Where the money leaks: match types and search terms
This is where most wasted ad spend actually lives, and it is the fastest thing to fix.
Keywords are not instructions, they are suggestions. Match types control how loosely Google is allowed to interpret them:
| Match type | What it does | Where it costs you |
|---|---|---|
| Broad match | Shows for anything Google considers related | Widest reach, and the biggest source of irrelevant spend |
| Phrase match | Shows for searches containing your meaning | Reasonable middle ground for most accounts |
| Exact match | Shows for that search and close variants | Tightest control, smallest volume |
Then open the search terms report. This is the single most useful screen in the platform, because it shows the searches people actually typed rather than the keywords you bought. Read the last 30 days and sort by cost.
You are looking for three patterns. Searches for things you do not sell. Searches with words like free, cheap, jobs, salary or DIY. And searches for competitors by name. Every one of those becomes a negative keyword, which tells the platform never to show you for that term again.
One thing worth knowing: Google added campaign level negative keywords, so the old habit of one shared list for the whole account is no longer the best structure. Exclusions can now sit where the intent problem actually is.
Location settings: the leak nobody checks
This one takes two minutes, and it is missed in almost every account we open.
When you target a location, Google gives you two options. Presence means your ads show to people who are actually in that place. Presence or interest means they also show to people anywhere in the world who have shown interest in that place. Google's guidance on geographic targeting sets out the difference, and presence or interest is the wider of the two.
Now read that again with the UAE in mind. A very large share of searches about Dubai come from people who are not in Dubai. Tourists planning a trip, expats who have not moved yet, buyers comparing from another country. If what you sell needs the customer to be here, presence or interest means you are paying for clicks from people who cannot buy. You will never spot it in the keyword reports either, because the keyword is fine. The location is the problem.
Three checks:
- Which setting is selected. It sits in the campaign's location settings, under the advanced options. If you serve a physical area, presence is usually the right answer.
- The geographic report. It shows where clicks actually came from. Anything from a country you do not serve should be excluded outright.
- What you have excluded. Ruling out places you cannot deliver to is as useful as adding negative keywords, and almost nobody does it.
If most of your customers have to be physically near you, fix this on both channels at once. The same searches feed your local SEO results too.
Campaign types, and the Performance Max question
Most accounts contain a mix of campaign types, and each one hides waste differently.
| Type | Best for | What to check in an audit |
|---|---|---|
| Search | Capturing existing demand | Match types, search terms, ad relevance |
| Performance Max | Broad reach across every Google surface | Whether it is cannibalising your brand searches |
| Google Shopping | Product catalogues | Feed quality, because a bad feed will undo a good bid |
| Display Network | Cheap awareness | Placement reports, and app traffic in particular |
| Remarketing | People who already know you | Audience recency and frequency caps |
If your account runs several of these at once, that spread is usually the reason nobody can say which one works. Sorting it out is most of what Google Ads management actually involves.
Performance Max deserves its own warning. It reaches everywhere and reports very little. So it can look like your best performer while quietly taking credit for people who searched your brand name and would have found you anyway. In an audit, always check whether brand searches are being counted as its wins.
Bidding: when automation helps and when it hides things
Automated bidding, which Google calls smart bidding, is usually better than manual, with one condition: it needs conversion data to learn from. Fed correctly it will beat a human at adjusting bids. Fed badly, it optimises confidently toward the wrong thing.
So the audit question is not manual against automated. It is whether the strategy matches the data available. A target CPA strategy on an account with four conversions a month has nothing to learn from, and the same goes for target ROAS. A maximise clicks strategy on an account that needs sales is aimed at the wrong outcome entirely.
Check what the strategy is optimising for, then check whether that is what the business actually wants. Those two answers disagree more often than you would expect.
Brand and non-brand: the split that changes every other number
If you take one structural finding from this article, take this one. It is the most consistent thing an audit turns up.
People searching your company name already know you. They convert far more often than people searching for what you sell. Both kinds of traffic are worth having, but they are different jobs, and when they share a campaign the account tells you a story that is not true.
Two things go wrong at the same time. Your reported numbers look healthy, because the brand conversions carry the average. And automated bidding learns from that same blended average, so it bids on cold searches as though they convert like brand searches. It is being confidently wrong with your money.
The audit question is short. What share of your conversions came from searches containing your own name? If nobody can answer that, the split does not exist yet.
| What you are looking at | Brand searches | Non-brand searches |
|---|---|---|
| Who they are | People who already knew you | People who do not know you yet |
| What they measure | How well you capture existing demand | Whether the channel creates demand |
| Fair expectation | Cheap, and high converting | More expensive, and slower |
| Common mistake | Counting them as new business | Judging them against brand numbers |
The fix is three steps and none of them are difficult. Put your brand keywords in their own campaign. Add your brand name as a negative keyword in the non-brand campaigns, so they stop absorbing traffic that was already yours. Then check Performance Max is not quietly collecting brand searches. That is what its brand exclusion setting is for, and it is usually switched off.
Give it four to six weeks after separating them. The blended number will look worse and the account will be healthier, because you are finally seeing the two halves for what they are.
This is the point where paid search stops being a channel you switch on and starts being performance marketing. Most of that discipline is just measuring the two halves separately and spending accordingly.
Quality Score, honestly
Quality Score is Google's rating of how relevant your keyword, ad and landing page are to each other, scored one to ten. A higher score lowers what you pay for the same position.
You cannot game it. You can only make the three things agree: the keyword someone searched, the ad they saw, and the page they landed on. When a keyword scores badly it is almost always because the ad is generic or the page is a homepage rather than an answer.
That last point is where PPC meets the website. Landing page experience is part of the score, and page speed is part of that, which is measured by Core Web Vitals. A slow page raises your cost per result twice: once by losing visitors, and again by lowering the score that sets your price. If the fix turns out to be the page rather than the account, that is a web design job, not a bidding one.
Account and campaign structure: how mess hides waste
Structure is not housekeeping. It decides what you can see.
Hundreds of keywords in one ad group means one ad serving all of them, so relevance drops and you cannot tell which keyword earned what. Duplicate keywords across campaigns make them bid against each other. Everything in a single campaign means one budget for products with very different value.
The audit test is simple: can you answer which keyword produced your last ten leads, in under a minute? If not, the structure is hiding the answer, and no bid change will reveal it.
Remarketing: the cheapest audience you already own
Almost every audit finds remarketing, or retargeting as it is often called, either switched off or set up once and forgotten. It is the cheapest audience in the account, because these people already visited.
Check three things. Recency, because someone who visited four months ago is a different prospect from yesterday's visitor. Frequency caps, because without them you pay to irritate people. And exclusions, because continuing to advertise to existing customers is spending to reach people who already bought.
Budget and pacing
Two questions, and they are simpler than they look.
Is the budget spread so thin that nothing gathers enough data to optimise? Small budgets across many campaigns is the most common structural mistake in small accounts.
Is anything limited by budget that is also profitable? That is the only situation where spending more is obviously correct, and it is the one people miss while raising budgets on campaigns that lose money.
Click fraud and irrelevant clicks, without the panic
Invalid clicks exist, and Google filters most of them automatically and credits them back. It is worth checking rather than worrying about.
Look at the placement report on Display campaigns for app names and odd sites, and exclude them. Turn off Search Partners for a month and compare, since that network is rarely audited and often converts worse than Google itself. Look for click patterns from locations you do not serve. Beyond that, spending heavily on click fraud tools before fixing your match types is fixing the small problem and leaving the big one alone.
What is different about PPC in the UAE
The mechanics are the same everywhere. Three local factors change what an audit finds here.
Seasonality is severe. Costs climb through Ramadan, the shopping festivals and the National Day period because every advertiser is bidding at once. The summer months go the other way as residents travel. Comparing this month to last month without accounting for that will make you draw the wrong conclusion.
Arabic keywords are usually cheaper. Far fewer advertisers write Arabic ads, so the same intent often costs less to reach. Most accounts we audit have never tested it. One warning if you do. Translated Arabic reads as foreign here, and copy written by someone who thinks in Arabic performs differently from copy put through a translator.
The click often becomes a WhatsApp message. This is worth understanding properly, because it is a measurement problem and not a preference. When someone submits a form on your site, auto-tagging has already attached a click identifier to that visit. Google can then match the lead back to the search that produced it. When someone taps a WhatsApp button instead, the conversation moves into an app and that identifier does not travel with it. You paid for the click and lost the receipt.
In most markets that is a small gap. Here, where WhatsApp is where business conversation actually happens, it can hide most of your results and make a working campaign look like a failing one.
Automating the audit with the Google Ads API and Claude
The tedious half of an audit is pulling and comparing data, and that part can be automated.
The Google Ads API returns your account data programmatically: spend, conversions, cost per result, search terms, placements, and breakdowns by campaign and device. Connect that to an AI assistant such as Claude and you can ask questions of it in plain language rather than exporting spreadsheets.
What that is genuinely good for:
- Search terms triage. Ask for last month's search terms sorted by cost with zero conversions. That is your negative keyword list, written for you.
- A weekly summary. Spend, conversions and cost per result by campaign, with the three biggest changes explained.
- Budget pacing checks. Which campaigns are limited by budget, and which are underspending.
- Structural checks. Duplicate keywords, ad groups with only one ad, campaigns with no negatives.
- Anomaly spotting. What changed against last week, so a broken tag gets noticed in days rather than months.
Two honest limits. It reads and reports, it does not decide: automated bid changes based on an AI summary is how budgets disappear. And it inherits your tracking, so if the conversion setup is wrong it will report wrong numbers faster than before. We wrote a separate walkthrough on getting API access to a marketing platform for Meta, and the Google process follows the same shape.
The 30 minute version
If you only have half an hour, treat this as your PPC checklist and do these five in order:
- Check a conversion actually records. Submit your own form and watch it appear.
- Read the search terms report for the last 30 days, sorted by cost.
- Add negative keywords for everything in there you would never want to pay for.
- Check what your bid strategy is optimising for, and whether that is the business goal.
- Open your top landing page on a phone and count the seconds before it is usable.
That is most of the value of a full audit, in one sitting. If you have ten minutes more, check your location setting and work out what share of your conversions came from people searching your own name. Those two answers change how you read everything else.
What to do with the findings
Fix in order of certainty, not size:
- Tracking, because it decides whether you can trust anything else.
- Negative keywords, because the saving starts the same day.
- The location setting, because it takes two minutes and it is usually wrong.
- The brand split, because until it exists every other number is an average of two different things.
- Structure, then bidding, then the pages themselves.
The first three are cheap and near certain. The last three take longer and need the first three to be right before they mean anything.
And judge the results on the right clock. Paid search reacts faster than organic, but a changed bidding strategy still needs one to two weeks to settle before its numbers mean anything. If you also run organic, our guide to how long SEO takes sets out that slower timeline. The SEO versus Google Ads comparison covers which channel deserves your next budget increase. For paid social, the same thinking applied to Instagram ads is worth a read.
If you would rather have someone else run it, that is what our PPC management work is: find the leaks first, then decide what to build. We will tell you honestly whether paid search is even the right channel for your business, and the campaign work we have published shows how we report it.
Running paid ads without auditing them is not advertising. It is a subscription.
