How to Audit a Google Ads Agency’s Work

Most owners try to audit a Google Ads agency by reading the report they were sent. That is understandable and almost useless. A monthly report is a curated document, written by the party being assessed, showing the metrics that party chose. It can be entirely accurate and still say nothing about whether anyone opened the account in week three.

The account itself is a different kind of document. Google records what was changed, when, and by which email address, without being asked. It records the searches that triggered your ads and how each conversion action was built. None of it is written for your benefit, which is exactly why it is worth reading.

So stop reading the summary and start reading the trail. This piece sets out how, in the order a consultant at IZI Digital Marketing would run it. We sell campaign management ourselves, so weigh the conclusion accordingly.

PART 1 · DIAGNOSE

What Does Auditing a Google Ads Agency Actually Mean?

IN BRIEFAn audit checks three things in order: whether the work happened, whether the measurement is sound, and only then whether the results are good. Skipping to results is how owners fire competent agencies and keep negligent ones. The order is what makes Google Search Ads management assessable at all.

Three questions sit inside every honest audit, and they are not equally weighted:

• Did the work happen? Evidence of activity on a schedule, not a promise of it. The cheapest question to answer, and the one owners almost never ask.
• Is the measurement trustworthy? If conversion tracking is broken, every number in every report is decoration. Nothing downstream survives a failure here.
• Are the results good? Worth asking, but meaningless until the first two clear. Bad results with sound measurement are a strategy problem; bad results with broken measurement are not yet a fact.

Results-first auditing punishes the wrong agency. A careful team in a hard market shows soft numbers, while a careless team riding one strong branded campaign shows comfortable ones. Only the trail separates them, which is why an audit differs from the vetting you do before signing an agency. Vetting predicts. Auditing verifies.

PART 2 · DIAGNOSE

Get Admin Access Before You Judge Anything

IN BRIEFYou cannot audit an account you cannot open. Google offers five access levels, and only Admin lets you see users, billing and settings in full. If your agency will not grant it on the account you pay for, you have your first finding before reading a single metric.

Google’s published access levels run from Admin down through Standard, Read only, Email only and Billing. Admin sees everything, including who else holds access. Read only shows you performance but not the account’s governance, which is often where the interesting findings sit.

Request Admin access in writing and note how long the reply takes. A settled agency grants it the same day, because nothing in the account embarrasses them. Hesitation is data.

The related question is whose name the account sits in at all. If the agency created it under their own manager account and billing, your audit becomes a negotiation rather than a right. Settle that first, using who should own your Google Ads account, alongside what a marketing report should contain.

BENCHMARK BRIEFING 1 OF 4

What Evidence Does Google Ads Already Keep?

IN BRIEFFive trails exist in every account by default, and none of them is written by your agency. Together they answer whether work happened, whether waste is being cleaned, whether the numbers are real, and whether the constraint is budget or strategy. An audit is largely the act of reading them in order.

The table below maps each trail to the audit question it settles and where to open it.

Read the middle column as the point. None of this evidence depends on anyone choosing to disclose it, which is what makes it the right place to start.

PART 3 · DESIGN

Read the Change History Before the Monthly Report

IN BRIEFChange history is the closest thing to a timesheet Google will ever give you. Filter the last six months, look at the shape of the activity rather than the count, and check whether the pattern matches what you are paying for. It settles the question the DIY-or-agency decision only guesses at.

Google keeps this record for the past two years and shows the email address of whoever made each change through the interface. Automated changes appear under names such as “Google Ads system”, and edits pushed through the API by a third-party tool appear under the tool’s name. That distinction matters: an account whose only movement comes from automation is an unmanaged account with a fee attached.

Four patterns tell you most of what you need:

• Rhythm over volume. Steady activity every month beats a burst in month one and silence afterwards. Volume is easy to manufacture; rhythm is not.
• Type of change. Negative keywords, bid targets, ad copy and budget edits are management. Pausing and unpausing the same campaign is motion.
• Named humans. Real email addresses, ideally more than one, appearing across the period rather than in a single week.
• Timing around invoices. Activity clustered in the two days before each report is a reporting habit, not a management habit.

Then check whether the changes make sense together. Adding hundreds of broad keywords with no matching negative list is a recognisable pattern once you understand how match types actually behave.

PART 4 · DESIGN

Does the Conversion Data Mean What the Report Claims?

IN BRIEFOpen the conversion actions list before you read a single conversion figure. Check what each action counts, whether it is marked primary, and whether the same enquiry is being counted twice. Most disputes about performance are really disputes about how conversions were tracked.

Three failures account for most inflated reports, and all three are visible in the conversion actions list rather than in any chart:

• Soft actions counted as conversions. Page views, scroll depth, clicks on a phone number that never connected. Countable, and not a customer.
• The same enquiry counted twice. A form tracked through both a direct tag and an imported analytics event doubles the total without doubling the business.
• Tags broken by a site change. A theme update, a new form plugin or a rebuilt thank-you page can sever tracking while the account keeps reporting whatever still fires.

The test is not technical. Submit a real enquiry through your own website, note the time, and confirm it appears as one conversion of the expected type. If it appears twice, or never, the reporting argument is over before it starts. Where the deeper problem is measurement across the whole funnel, that belongs in analytics and conversion rate work.

BENCHMARK BRIEFING 2 OF 4

Which Audit Findings Cost the Most?

IN BRIEFAudit findings are not equal, and working through them as a flat checklist throws away the advantage. One finding invalidates every other number; the rest each cost a share of monthly spend. Fix in descending order of exposure, not in the order you happened to find them.

The model below ranks common findings by the share of monthly ad spend each one typically puts at risk.

Note the shape. The top row is not a bigger version of the others; it is a different category, because it removes your ability to judge any of the rest.

PART 5 · DESIGN

Audit It Yourself, or Pay Someone to Do It?

IN BRIEFOwners can run the evidence half themselves in an afternoon. Judging whether the strategy was right needs someone who reads accounts for a living. Do the first half yourself before paying for the second, because the findings often make the paid review unnecessary.

Three routes exist, and they suit different situations rather than different budgets.

Free audits from agencies pitching for the work are not worthless, but read them as proposals: the findings will be true and the framing will favour a switch. A paid review with no management attached removes that incentive. Some findings are only tool-visible, which is where the Google Ads tools worth paying for earn their keep, and any quoted fee should be read against management fee versus ad spend.

BENCHMARK BRIEFING 3 OF 4

What Should a Managed Account’s Activity Look Like?

IN BRIEFBoth a managed and a coasting account look busy in month one. The difference appears from month three onwards, when the build is finished and only maintenance remains. Reading change history over twelve months, not one, is what makes that difference visible.

The illustrative curve below shows logged changes per month in two accounts of similar size.

The absolute numbers will differ in your account. The shape should not: a plateau, not a cliff.

PART 6 · DEPLOY

The Seven-Day Audit You Can Run This Week

IN BRIEFSeven days is enough because most of the work is reading, not analysing. Run it in a fixed order so each step’s answer changes what the next step means, and write findings down as you go rather than reconstructing them later.

Do it in this sequence:

1. Day 1 — request Admin access in writing. Record the date you asked and the date it arrived. Both go in the findings.
2. Day 2 — test the conversion tracking. Submit a real enquiry, note the time, and confirm it appears once, as the right action type.
3. Day 3 — read six months of change history. Note rhythm, change types, and which email addresses appear.
4. Day 4 — read 90 days of search terms. Count the queries that could never buy from you, and check whether any were blocked.
5. Day 5 — check structure and impression share. Confirm brand and non-brand are separated, then see whether budget or competition is the ceiling.
6. Day 6 — compare cost per enquiry to your own economics. Not to a published benchmark, but to what a customer is worth to you.
7. Day 7 — write one page. Findings, severity, and the single question you want answered.

Day 6 is where owners most often go wrong, reaching for an industry average instead of their own margin. Published return-on-ad-spend benchmarks are context, not a verdict on your account.

BENCHMARK BRIEFING 4 OF 4

What Do Common Findings Cost by Ad Spend Tier?

IN BRIEFPercentages feel abstract until they are ringgit. The same two uncorrected findings cost a small advertiser a modest amount and a larger one a staff salary. That gap is the real argument for auditing annually rather than when something already feels wrong.

The model applies two of the earlier findings across four Malaysian spend tiers.

Compare the right-hand column with what an audit costs. For most accounts above RM 8,000 a month, the arithmetic answers itself.

PART 7 · DRIVE

What to Do With What the Audit Finds

IN BRIEFMost audits end in a correction, not a firing. Send the findings as questions with dates attached, give thirty days, and re-read the same trails afterwards. Only a second failed reading justifies moving, because switching costs a learning period you have already paid for once.

Sort your findings into three piles before you send anything:

• Fixable this month. Negative keywords, campaign separation, a soft conversion action removed. Ask for a date, not a discussion.
• Structural. Account ownership, reporting cadence, who actually works on the account. These belong in a contract conversation.
• Disqualifying. Access refused, tracking knowingly inflated, or a change history showing a year of nothing.

Then re-audit in thirty days using the same steps. A team that fixes six findings in a month is worth keeping, and the second reading is where you find out which sort you have.

FAQ

Common Questions About Auditing a Google Ads Agency

1. How often should I audit my Google Ads agency?

Once a year as a rule, plus any time results shift without an explanation you accept. It depends on spend, since a larger account carries more ringgit behind each uncorrected finding. Above roughly RM 8,000 a month, add a light quarterly reading of change history and search terms.

2. Can I audit the account without damaging the relationship?

Yes, if you frame it as routine rather than suspicious. It depends on how you open, because “we audit every supplier annually” lands very differently from “I think something is wrong”. Good agencies expect it, since a clean audit ends the doubt.

3. What is the first thing to check when I audit a Google Ads agency?

Conversion tracking, before anything else. It depends on nothing, because inflated or broken tracking makes every other figure unreadable. Submit one real enquiry through your own site and confirm it appears once, as the correct action type, at the expected time.

4. Should I pay another agency to audit my current one?

Only after you have read the trail yourself. It depends on what your own reading found, because an independent review adds most where the evidence looks clean but results stay poor. Prefer a paid review with no management attached, and apply the scepticism you would bring to any agency shortlist.

5. What does a genuinely failed audit look like?

Admin access refused, a change history that stops after month two, and conversion counts nobody can explain. It depends on how the agency responds when shown the evidence, since a candid explanation and a repair schedule can still rescue things. Silence is the answer.

THE VERDICT

Judge the Trail, Not the Slide Deck

An agency controls its report completely and its account record not at all. That asymmetry is the whole opportunity. Did the work happen, is the measurement sound, is waste being cleaned. Every question that matters already has an answer sitting in the account, written by the platform rather than by the party you are assessing.

So do the reading in order: access, tracking, change history, search terms, structure, then results. If the first four hold up and results are still soft, you have a strategy problem worth solving with the agency you already have. If they do not hold up, no amount of dashboard polish changes what you found.