The invoice from Google gets bigger every month and the dashboard shows a healthy conversion count. Then the sales team tells you half the enquiries are irrelevant and some never arrive at all. We see this constantly. The account isn't broken; it is measuring the wrong thing and spending in the wrong places.
Here is how we audit a Google Ads account. The order matters: measurement first, then structure, then traffic quality, and only then bidding and automation. At the end there is a one-page checklist.
Why does an audit start with measurement?
Smart Bidding tries to produce more of whatever you tell it counts as a conversion. Count everyone who stays on a page for two minutes, and it will find you more people who stay for two minutes. So before we look at a single keyword, we open Goals, then Conversions, then Summary.
Primary and secondary conversion actions
Every conversion action is marked either primary or secondary. Primary actions are used for bid optimisation and appear in the Conversions column. Secondary actions are for observation only: they show up under All conversions and don't influence bidding. Account-default goals apply to every campaign unless a campaign has its own goal selection.
The mistakes we find most often in audits:
- Weak signals such as page views, scroll depth or visits to the contact page set as primary.
- The same form submission counted twice as primary, once by the Google Ads tag and once through a key event imported from GA4.
- Tags that fire again whenever the thank-you page is refreshed or revisited.
- A click on the phone number weighted exactly like an actual phone call.
One business outcome, one primary action. Everything else stays secondary and serves reporting.
Counting: Every or One?
The Count setting offers two options. Every counts all conversions after an ad interaction, which suits an online shop. One counts a single conversion per click, which is right for service businesses where the same person might fill in a form three times. If a lead generation account is set to Every, your cost per conversion looks better than it really is. Changing the setting doesn't rewrite history; only future conversions follow the new rule.
Enhanced conversions and CRM data
Enhanced conversions take first-party data entered on your site, such as an email address or phone number, hash it with SHA256 and send it to Google, which improves matching when cookies are missing. The web version measures purchases and forms on the site; the leads version measures enquiries that later turn into sales in your CRM. For service businesses the second matters more, because the system learns what a customer looks like, not just a form filler.
Check two things: does the diagnostics view show data arriving, and does your privacy notice cover this transfer?
Consent Mode v2 and cookie consent
If you show ads to users in the EEA, the UK or Switzerland, Google expects their consent choices to be passed through Consent Mode. There are four signals. ad_storage and analytics_storage govern cookie storage, ad_user_data governs whether user data may be sent to Google for advertising, and ad_personalization covers personalised ads and remarketing. In a basic implementation, tags wait for consent before loading. In an advanced one, they load straight away, send cookieless pings when consent is refused, and Google models conversions from them.
In the UK, PECR sits alongside UK GDPR and requires consent for non-essential cookies; across the EU the ePrivacy rules do the same job. A banner that only offers Accept, or one whose Reject button changes nothing in the tags, is a problem twice over: legally and in your data.
Does the account structure follow your business or the setup wizard?
In a healthy account, campaigns follow business lines: margins, budget needs, regions. Take a heating engineer in Manchester. Boiler installation and emergency repairs belong in separate campaigns, because one searcher is comparing quotes and the other needs someone today. Put both in one campaign and the budget drifts towards whichever side has cheaper clicks.
The opposite mistake is just as common: a campaign for every town, product and match type spreads the data so thin that Smart Bidding never sees enough conversions. Google's own guidance is to consolidate where you can.
Keep brand searches separate. Searches for your own name convert at high rates and, mixed into non-brand campaigns, make the whole account look healthier than it is, so one of our first audit steps is to compare brand and non-brand spend.
Search terms: what are people actually typing and clicking?
A keyword is what you chose. A search term is what the user actually typed. You'll find the report under Campaigns, then Insights and reports, then Search terms. Pick the last 30 or 90 days, sort by cost and read the top fifty non-converting rows one by one. Tedious, yes, and usually the half hour that saves the most money.
The report doesn't show everything. For privacy reasons Google leaves out terms with too little volume, so the visible terms add up to less than total spend. A large gap means your matching is too loose.
Adding negative keywords at the right level
Almost every business has words that are almost never worth paying for: jobs, vacancies, apprenticeship, how to, second hand, PDF, course notes. Put those in an account-level negative list. Campaign-specific ones, like excluding repair from the installation campaign, go at campaign level. Negatives don't cover close variants: excluding boiler won't block boilers.
Equally, don't throw every non-converting term into the negatives. A low-volume term that genuinely fits what you sell deserves time to gather data.
Match types, broad match and AI Max
The line between exact, phrase and broad match isn't as sharp as it used to be; even exact match covers close variants with the same meaning. Google recommends pairing broad match with Smart Bidding. That advice holds when measurement is sound. In an account with faulty tracking, broad match is the quickest way to spread a budget across irrelevant searches.
If AI Max for Search campaigns is switched on, look at three features separately: search term matching, text customisation and final URL expansion. The last two are linked; turning off text customisation turns off final URL expansion too. With URL expansion on, Google can send visitors to another page it considers more relevant, and pinned headlines are then ignored. Clicks landing on blog posts or the careers page mean it's time to check URL exclusions.
The settings that quietly burn money
Location options: presence or interest?
The default under location options in campaign settings is Presence or interest. It shows ads to people who are in, or regularly in, your targeted area, plus people who have shown interest in it. A heating engineer targeting Manchester can end up in front of someone in Dubai researching a flat their parents own there. For a hotel, that's fine. For businesses serving customers on site, Presence is the right choice.
For exclusions, only Presence now applies. The geographic reports show where clicks really come from; spend from cities you don't serve means the setting is wrong.
Search partners and Display expansion
When you create a Search campaign, two network boxes usually arrive ticked: Include Google search partners and Include Google Display Network. The first adds search results on partner sites; the second lets the campaign spend leftover budget on display placements.
We recommend switching off Display expansion in Search campaigns. Display advertising deserves its own campaign, creative and budget. Judge search partners on their own numbers by segmenting the campaign table by network.
Auto-applied recommendations
The auto-apply settings on the Recommendations page let Google apply certain recommendations without your approval: new keywords, match type changes, removal of keywords it deems redundant, bid strategy switches. In an audit we go through its Manage and History tabs. Handing those decisions to automation on purpose is fine; the problem is when nobody knows it has happened.
What do ads, assets and Quality Score tell you?
With responsive search ads, filling every headline slot matters less than making the headlines genuinely different from each other. An ad that says the same sentence five ways gives the system nothing to test. Pin headlines only when you must; every pin reduces the number of combinations. Ad strength is a drafting check, not a performance metric.
Assets make the ad larger and influence click-through rate: sitelinks, callouts, structured snippets, call assets, images, business name and logo. Also check the assets Google has created automatically at account level; stale lines pulled from your site sometimes still show.
Quality Score is a diagnostic tool
Quality Score is given per keyword on a scale of 1 to 10 and is built from three components: expected click-through rate, ad relevance and landing page experience. Each is rated Above average, Average or Below average against other advertisers on the same keyword. The columns are hidden by default, so add them to the keyword table.
Google doesn't use this score directly in the auction; it's there to diagnose the account, and it still points the way. Low ad relevance usually means a muddled ad group. Low landing page experience means the problem is on the page.
The landing page is part of the audit
The ad account only buys the click. Everything after that is the page's job. We check how quickly the page loads on mobile, whether the promise in the ad is repeated above the fold, whether the phone number can be tapped, and how many fields the form asks for. The privacy notice should be linked near the form, and a shop needs delivery, returns and cancellation information where consumer law expects it. Page speed is covered in detail in our Core Web Vitals article.
Bid strategy, the learning period and budgets
Smart Bidding strategies (Maximise conversions, Target CPA, Maximise conversion value, Target ROAS) are only as good as the data they are fed. The first audit question is whether the target matches what the last 30 days actually delivered. Setting a target CPA at half your real cost per conversion doesn't speed anything up. It just chokes impressions.
What people get wrong about the learning period
According to Google's help page, a bid strategy enters Learning in three situations: when it is newly created or reactivated, when campaigns, ad groups or keywords are added to or removed from it, and when the conversion goals it optimises towards change significantly. The same page says budget and target changes don't trigger this status, and that Learning never lasts more than 7 days.
That doesn't mean you can fiddle as much as you like. Google notes that performance can take one to two conversion cycles to settle on a new target and advises against several target changes within a single cycle. A conversion cycle is the time from click to conversion: hours for an emergency plumber, weeks for a B2B software sale.
How far can daily spend exceed the budget?
The figure you enter is an average daily budget. For most campaign types, Google can spend up to twice that amount on a single day, but what you're charged in a month won't exceed 30.4 times the average daily budget. Raising and lowering the budget mid-month muddles that calculation. If a campaign is limited by budget, check cost per conversion first; more budget for a poor campaign just enlarges the problem.
Performance Max doesn't have to be a black box
Performance Max reaches Search, Shopping, YouTube, Display, Discover, Gmail and Maps from a single campaign. A pattern we see often: the campaign reports a high ROAS, but much of the spend goes on brand searches from people who were already looking for you. A success on paper, and in reality you've paid for your own customers.
The checks we run:
- Brand exclusions. Add your own brand, and resellers where relevant, to a brand list and exclude it. In retail campaigns with a product feed, the exclusion can apply to text ads only, so branded Shopping traffic carries on.
- Campaign-level negative keywords. Google began rolling them out to all advertisers in early 2025. They apply only to Search and Shopping inventory; for Display and Video you need excluded content keywords in the content suitability settings.
- Search themes. Themes in asset groups give the system direction; search term insights show whether a term came from keywordless targeting or from your themes.
- The search terms report. On the Search terms screen, the Performance Max search terms and landing pages view shows which term went to which page and ad format.
- The channel performance report. It shows how spend and conversions split across channels. A quiet channel isn't a fault in itself, but if most conversions come from one channel, ask why.
- Final URL expansion and URL exclusions. Limit which pages on your site the campaign can send traffic to.
Switching Performance Max off is rarely the answer. With accurate measurement, brand limits and a clean product feed, it can reach demand that Search campaigns miss, particularly in ecommerce.
Competitors and changes: why did clicks get more expensive?
The Auction insights report shows other advertisers entering the same auctions and how you compare. Impression share is how often you appeared in the auctions you were eligible for. Overlap rate is how often a competitor showed at the same time as you. Position above rate is how often they ranked above you when you both showed. Outranking share is how often you beat them.
This report earns its keep when someone asks why clicks cost more than last month. Before raising bids, check your margin; you needn't win every auction. Read the report alongside Change history. Sometimes the culprit isn't a rival but a change made in your own account a fortnight earlier.
Where to start: the checklist and the order
The table condenses what we check, in order. It needn't be done in one day, but keep the sequence.
| Check | Where to look | What you want to see |
|---|---|---|
| Primary conversions | Goals > Conversions > Summary | One primary action per business outcome, weak signals secondary |
| Counting | Conversion action > Edit settings | One for lead actions, Every for sales |
| Double counting | Conversion sources and GA4 imports | Each form or sale counted as primary by a single tag |
| Enhanced conversions and consent | Conversion diagnostics, banner test | Data arriving, correct consent signal for users who decline |
| Search terms | Insights and reports > Search terms | Most spend on terms relevant to your business |
| Negative lists | Shared library and campaign negatives | Up-to-date lists at account and campaign level |
| Location option | Campaign settings > Locations | Presence for on-site services, clicks from your service area |
| Networks | Campaign settings > Networks | Display expansion off, search partners judged on their own numbers |
| Auto-apply | Recommendations > Auto-apply | Only recommendation types switched on deliberately |
| Quality Score | Extra columns in the keyword table | An action plan for every below-average component |
| Bid strategy | Campaign bid strategy status | Realistic targets, no needless relearning |
| Performance Max | Brand exclusions, channel performance | Brand traffic limited, channel split understood |
| Change history | Campaigns > Change history | An owner and a reason for every significant change |
- Spend the first day on measurement alone: clean up primary actions, check counting, and verify every tag with a test conversion.
- Then read the last 90 days of search terms and build your account-level negative list.
- Go through location, network and auto-apply settings campaign by campaign.
- Separate brand and non-brand spend, and add brand exclusions to Performance Max.
- Bring bid targets back to what recent results really show, then leave targets alone for a full conversion cycle.
- Log every change with its date, and come back to the same table two weeks later.
At AYTE Digital we audit accounts in this order, measurement first, and turn the findings into a to-do list your own team can act on. The account stays in your name, the data stays with you, and every decision comes with written reasoning.
