How to Measure the Effectiveness of a Google Ads Agency: Myths, Facts, and a Post-Hire Audit Checklist
A practical framework for U.S. founders and marketing directors to judge an agency they already pay - using account-level evidence, not the agency's own summary deck.

Image via 123RF
In This Article
Audit the Account, Not the Deck
Benchmarks Are Starting Points
Incrementality Beats Correlation
Why Judging a Google Ads Agency Is Harder Than Judging Google Ads
Most guidance on Google Ads measurement assumes you are running the account yourself. But if you are 18 months into a retainer, the question is different: not "is my Google Ads working?" but "is my agency earning its fee?" Those are not the same question, and they require different evidence. An agency can post a healthy blended ROAS while your bank balance stays flat - and the reported numbers can be technically accurate the whole time.
The 18-Month Retainer Trap: When Reported ROAS and Bank Balance Disagree
Picture a U.S. DTC brand spending the equivalent of ZAR 600,000 to ZAR 1,200,000 per month. The monthly dashboard shows a stable 4.2x blended ROAS. Gross margin is 30%. Break-even ROAS at that margin is roughly 3.3x - so on paper the account is profitable, yet the owner sees flat cash. Why? Because blended ROAS includes brand-term searches, remarketing to existing customers, and retargeting audiences that would have converted anyway. The agency is harvesting demand, not creating it, and the blended figure hides that.
Myth 1 - 'The Monthly Report Tells Me Everything I Need to Know'
Fact: Reports Are Summaries; the Account Is the Evidence
A monthly PDF is a curated summary. The account itself - its change history, search terms, audience settings, and conversion definitions - is the primary evidence. Google Ads change history shows every edit, who made it, and when. If your agency claims a month of "heavy optimisation" but change history shows three edits, that is a concrete, verifiable discrepancy you can resolve in a single screen share.
A single account review with change history open is worth more than six months of summary decks. Ask for a shared screen, not a PDF.
Myth 2 - 'A Rising Blended ROAS Proves the Agency Is Adding Value'
Fact: Blended ROAS Can Be Manufactured by Brand-Term Bidding
If you sell a branded product, brand searches convert at enormous rates for almost no cost. An agency can quietly add brand match types and a brand campaign, and blended ROAS will rise without creating a single new customer. The fix is simple: demand a permanent split between brand and non-brand in every report. Non-brand ROAS is the one that reflects genuine demand generation. Any agency that resists that split is telling you something.
Myth 3 - 'More Conversions Means the Agency Is Doing Better'
Fact: Conversion Definitions Drift, and Cheap Conversions Hide Poor Lead Quality
Conversion counts are not comparable across time unless the definitions are identical. An agency can change what counts as a conversion - adding a newsletter signup, a low-value micro-conversion, or a broad "engaged session" - and the count rises while lead quality falls. For lead-gen and high-ticket accounts, offline conversion imports and call tracking are the only way to connect ad clicks to closed revenue. Without them, you are optimising to a proxy.
Myth 4 - 'Optimization Score Is a Performance Grade'
Fact: Optimization Score Is a Google Product Metric, Not an Outcome Metric
Optimization Score measures how closely an account follows Google's own recommendations. It is not a grade of agency performance, and applying every recommendation can actively harm accounts - for example, broad match expansion or auto-applied recommendations that spend into irrelevant queries. A 100% score is a talking point, not a KPI. Treat it as input to a discussion, never as evidence of effectiveness.
The KPI Dashboard to Demand From Your Agency
Below is a practical set of metrics that reveal agency behaviour rather than just account outcome. Ranges are starting points, not targets: your own historical baseline beats any published average.
| Metric | What it proves about the agency | Reasonable starting range (U.S., varies by sector) |
|---|---|---|
| Non-brand ROAS | Shows whether the agency is creating new demand or harvesting existing demand | Depends on gross margin; above break-even (1 ÷ margin) is the minimum |
| CPA / CPL by campaign type | Reveals whether the agency understands your unit economics or just chases volume | Compare to your own 12-month baseline, not an industry average |
| Conversion rate (site and landing page) | Tests whether the agency is fixing post-click experience or blaming traffic quality | 2-4% for eCommerce; 5-10% for high-intent lead gen - wide variance |
| CTR (search and shopping) | Indicates ad relevance and whether the agency is maintaining creative and keyword hygiene | 3%+ on search is a common heuristic; shopping often 0.5-1.5% |
| Impression share lost to budget vs rank | Tells you whether the bottleneck is budget or bidding/quality - very different agency actions | No universal target; the split matters more than the absolute figure |
| Quality Score (weighted) | A rough signal of ad relevance and landing page experience, not a performance grade on its own | 7+ on core commercial terms is a workable floor |
| Search term waste % | Shows whether the agency is running a disciplined negative keyword routine | Waste should trend down; double-digit % of spend on irrelevant terms is a red flag |
| Change history volume | Confirms whether optimisation is happening inside the account at all | Expect meaningful edits most weeks, not one bulk change per month |
Leading vs Lagging Indicators: What You Can Actually Steer Mid-Quarter
Lagging indicators - ROAS, revenue, profit - tell you what already happened. Leading indicators - impression share, CTR, conversion rate, search term quality, ad strength - tell you what is likely to happen next. If your agency reports only lagging indicators, you cannot intervene until the quarter is over. A credible agency shows both, and uses leading indicators to explain the lagging ones.
Break-even ROAS formula - a 30% margin business needs 3.3x just to cover ad cost, before fixed expenses.
Benchmarks in Context: The 3-3-3 Rule and the 'Is ROAS 4 Good?' Question
Benchmarks are useful only as a starting point for a conversation. The right question is not "is 4x good?" but "is 4x good for this brand, at this margin, at this stage?" A 70% gross margin brand profits at a 1.5x ROAS; a 30% margin brand barely breaks even at 3.3x. The same number means opposite things.
The 3-3-3 Rule Is an Industry Heuristic, Not a Google Standard
The "3-3-3 rule" - 3% CTR, 3% conversion rate, 3x ROAS - circulates widely in PPC communities. It has no primary-source definition from Google and should never be presented as an official standard. Use it as a folk heuristic to sanity-check a new account, not as a KPI. If an agency uses the 3-3-3 rule to defend mediocre performance, ask for the non-brand numbers, the margin-adjusted break-even, and the trend over 12 months.
Break-Even ROAS Formula and a Worked ZAR Example
Break-even ROAS is simple arithmetic. Divide 1 by your gross margin:
break_even_roas = 1 / gross_margin
# Example: 35% gross margin brand
# break_even_roas = 1 / 0.35 = 2.86x
# If the agency reports blended ROAS of 4.2x
# but 60% of that revenue comes from brand terms,
# non-brand ROAS may be closer to 1.8x - below break-even.
In a U.S. context where the brand spends ZAR 800,000 per month, a 1.8x non-brand ROAS on R560,000 of that spend is a loss-making acquisition programme dressed up by brand harvesting.
Agency-Specific Signals That Live Inside the Account
Access Audit: Who Owns Google Ads, GA4, GTM, and the Conversion Setup
You should own every asset. The agency should have user-level access, not admin ownership of the Google Ads account, GA4 property, Google Tag Manager container, or the domain's DNS for server-side tagging. If your contract does not explicitly state that all accounts and data remain your property, that is a material risk. Request a written asset inventory at every quarterly review.
Change History as the Most Underused Agency Audit Tool
Google Ads change history records edits, users, and timestamps. Run a 90-day export and count changes by week. A common pattern: heavy activity in the first 60 days, then near silence. That is maintenance mode. Maintenance mode is not automatically bad, but it should be priced and disclosed accordingly - not sold as active optimisation.
If the agency cannot show a meaningful change log, ask what they have been doing. It is a fair question, not an accusation.
Reporting Cadence, KPI Definition Sheets, and Fee Transparency
Demand a one-page KPI definition sheet: what counts as a conversion, how revenue is attributed, whether brand is separated, and which date range is used. Without it, every monthly report is unfalsifiable. On fees, percentage-of-spend models reward higher spend, while flat retainers reward scope clarity - neither is inherently better, but the incentive should be stated openly.
Attribution, Credit-Stealing, and Incrementality
Brand vs Non-Brand Splits: Separating Demand Creation From Demand Harvesting
Brand and non-brand campaigns should be reported separately, permanently. Brand searches often convert at 8-15% or higher because the customer already knows you. Non-brand is where an agency demonstrates real skill. If the agency refuses to split the two, red flag.
Geo Holdouts and Pause Tests: The Only Honest Incrementality Check
Correlation is not causation. The only reliable way to know whether the agency is adding incremental revenue is a controlled test: pause campaigns in a matched set of geographies for 2-4 weeks and compare total revenue against a held-out control group. This is uncomfortable for agencies because it can reveal that some "performance" was happening anyway. It is also the most honest conversation you can have about value.
Consent Mode, CCPA/CPRA, and Why Totals Won't Match Exactly
Under CCPA/CPRA and consent mode, modelled conversions and analytics-platform totals will not match Google Ads exactly. A 5-15% variance is normal and is not proof of misreporting. What matters is that the agency can explain the variance, not hide it.
Red Flags and the Questions to Ask at Your Next Quarterly Review
Checklist: Eight Warning Signs in an Agency's Reporting and Account Behaviour
- Blended ROAS reported without a brand/non-brand split.
- No change history activity for weeks at a time.
- Conversion definitions changed mid-quarter without disclosure.
- Optimization Score presented as a performance KPI.
- Impression share reported as a single number, with no budget-vs-rank split.
- Offline conversions and call tracking missing in a lead-gen account.
- Agency owns the Google Ads, GA4, or GTM accounts.
- No written KPI definition sheet or fee structure.
The Post-Hire Verdict Checklist
A 90-Day Audit Sequence You Can Run Without Pausing Spend
- Weeks 1-2: Export change history and conversion definitions. Compare to what the agency claims was done.
- Weeks 3-4: Build a brand vs non-brand report for the last 12 months. Calculate non-brand ROAS against your break-even.
- Weeks 5-6: Request the asset inventory and confirm ownership of Google Ads, GA4, GTM, and offline conversion workflows.
- Weeks 7-8: Design a geo holdout or pause test on one campaign group. Do not pause all spend - run a controlled test.
- Weeks 9-10: Review search term waste and negative keyword hygiene over the past quarter.
- Weeks 11-12: Compare leading indicators (impression share, CTR, conversion rate) against the lagging indicators. Present findings at the quarterly review and agree on corrective actions.
The typical point at which a retainer quietly drifts into maintenance mode - the reason this audit exists.
Judging an agency is not about distrust. It is about running a structured review that any high-performing in-house team would run on itself. When you have clean access, a written KPI definition, a brand split, and one incrementality test, the conversation shifts from opinion to evidence - and both sides benefit.


