A useful Google Ads audit ends with decisions: which problems need attention first, which claims need more evidence, and which changes should wait. A long checklist is only the starting point. The real deliverable is a short, defensible action plan that connects each recommendation to an observed condition, a business consequence and a way to verify the result.
This guide is for Canadian business owners and paid-media managers reviewing an existing account. It covers goals, access, measurement, campaign structure, policy, spend controls and change risk. Use it to commission an audit, conduct an authorized review or challenge a recommendation before approving it. All worked examples below are fictional. Dollar amounts in those examples are Canadian dollars unless stated otherwise.
Define what the audit must decide
Start with the decision the business faces. An account preparing for a larger budget needs different evidence from one experiencing an unexplained spending increase. An agency handover requires access and ownership checks that a routine campaign review may already have resolved. Put that context at the top of the audit so reviewers do not mistake every available setting for a problem.
Write a brief covering the advertised services or products, serviceable locations, languages supported, meaningful customer actions, commercial constraints and the person who can authorize changes. Include the account currency and reporting time zone. A Canadian business can operate an account in another currency; the business address does not settle how its reports should be interpreted.
Define the boundary between an audit and implementation. The reviewer can inspect authorized evidence and propose actions. Editing budgets, changing conversion goals, removing access and submitting appeals each require an appropriate owner and an agreed approval route. Establish any emergency authority before an incident occurs. A finding marked urgent does not itself grant permission to change an account.
For a fictional commercial cleaning company, the audit question might be: “Can we approve more spend on recurring office contracts while maintaining an agreed spending ceiling?” The reviewer would need the contract objective, current campaign goals, reliable outcome evidence, capacity information and budget controls. A suggestion to add more keywords would be secondary until those foundations were understood.
Set a clear completion standard: every material finding must have evidence, a consequence, an uncertainty statement, a next decision and an accountable owner. Areas that could not be reviewed must remain visible. “No problem observed in the available evidence” is a narrower and more useful conclusion than “the account is healthy.”
Build an evidence pack before making recommendations
Choose a reporting period that answers the question
Record exact start and end dates rather than “last month.” Choose enough history to examine the relevant buying cycle, reporting delay and seasonal conditions. Then identify recent incidents separately. A current broken destination and a quarterly performance question do not need the same observation window.
Compare like with like: the same campaigns, currency, conversion definition, filters and reporting basis. Annotate holidays, business closures, promotions, website releases and significant account changes. If a campaign changed its objective halfway through the period, split the analysis or explain why its combined total cannot support a clean comparison.
Recent conversions may still be arriving. Google’s conversion lag reporting documentation explains how delayed conversions affect interpretation of CPA and ROAS. Record when a report was exported and when it should be refreshed. Do not call an immature period a failed campaign merely because its final outcomes are not yet known.
Collect reproducible evidence, not loose screenshots
Request the smallest evidence pack that can answer the agreed questions. Give each export or capture an evidence ID and keep the original unchanged. A screenshot should show the relevant setting, scope and capture date; a spreadsheet should retain column definitions, filters and totals. Record where an authorized reviewer can retrieve the evidence without putting customer details in the shared action plan.
| Area | Evidence to request | Decision it supports |
|---|---|---|
| Business goals | Approved offer, outcome definitions, capacity and spending authority | Whether the account is pursuing the right result |
| Access | Current users, manager relationships and responsible account administrator | Who can inspect, approve and implement changes |
| Measurement | Conversion-action settings, campaign goal selections and dated diagnostics | Which signals reports and bidding actually use |
| Delivery | Campaign settings, performance exports, available query reports, ads and destinations | Where observable delivery differs from the brief |
| Policy and billing | Exact status messages, affected entities, budget settings and relevant cost reports | Which restrictions or spending concerns require action |
| Change control | Change history, automation settings and pending releases | What changed and what could change again |
We build and run Google Ads campaigns measured on calls, forms and booked jobs, not clicks.
Separate direct observations from explanations. “The selected goal includes a page-view action” is an observation. “This setting caused poor sales” is a hypothesis requiring further evidence. The first statement may justify a goal review without proving the second. Keep that distinction in every finding, especially when presenting conclusions to a business owner.
Use four evidence states: observed, corroborated, hypothesis and unavailable. An observed setting can be certain while its financial effect remains unknown. Corroborated means an independent source supports the same conclusion, such as an approved business brief and the settings export showing a mismatch. It does not mean the change’s future benefit has been proven.
Where reports disagree, document the units and timing before choosing a winner. Google’s guidance on reporting discrepancies identifies conversion-time columns as one way to make certain comparisons more aligned. A platform conversion, an enquiry and a sale are different records; matching totals alone does not establish that the same events were measured.
Check access, ownership and the ability to act
Confirm the correct account and the relationship between the business, its agency and any manager accounts. Ask who maintains the account, who controls payment administration, who can approve additional spend and who can restore access during a handover. These responsibilities may belong to different people.
Use individual invitations with access appropriate to the task. Google’s account access-level documentation describes permissions, while its account security guidance covers protective measures and reviewing manager relationships. Start an inspection with read-only access where it supplies the necessary evidence. Have an authorized administrator provide restricted information the reviewer cannot see.
Review current users, pending invitations and manager connections with the business owner. An unfamiliar name is a question to resolve, not proof of a breach. Confirm the person’s role and any operational dependencies before recommending removal. Document a supported administrator route before removing access used for billing, recovery or campaign operations.
For temporary reviewers, consider Google Ads access expiration and record the end date in the account time zone. Plan a handover for open findings before that access ends. Access expiry does not resolve outstanding work; it simply limits how long the temporary permission remains available.
If evidence indicates unauthorized changes, move the issue out of the normal optimization queue. Preserve relevant timestamps and affected entities, involve the responsible administrator and follow Google’s compromised-account reporting guidance. Keep sensitive account evidence in a restricted location. The audit register should identify the incident owner and next action without containing credentials or recovery secrets.
Audit the goals behind the performance numbers
Map each campaign to its intended outcome
For every active campaign, record its commercial purpose, selected conversion goals and the actions those goals contain. Then compare that configuration with the approved brief. A campaign can be operating as configured while pursuing an outcome the business no longer wants.
Google distinguishes primary and secondary conversion actions. Primary actions normally appear in the Conversions column and are used for bidding when their standard goal is selected. Secondary actions normally serve observation, but inclusion in a custom goal is an important exception: those actions can be used for bidding. Read Google’s primary, secondary and custom-goal rules before concluding that a label alone explains campaign behaviour.
Capture the effective campaign selection, including account-default goals, campaign-specific choices and custom goals where present. Do not stop at the account’s conversion list. Two campaigns can use the same action differently because their selected goals differ.
For each material action, record its source, business definition, counting setting, value basis, currency where relevant, attribution settings and recent diagnostic status. Ask whether the event describes a useful outcome or only a step towards one. A completed order, an enquiry and a brochure interaction should not become interchangeable simply because each is available for reporting.
Separate signal integrity from lead quality
Inspect whether the evidence supports the event definition. Look for unexplained discontinuities after releases, multiple actions apparently representing the same milestone, unexpected values and missing documentation. Treat these as investigation leads. Two similar conversion names do not prove duplicate counting, and a quiet diagnostic screen does not prove the entire customer journey works.
A fictional audit might find an action called “Quote request” and another called “Quote confirmation.” The reviewer should establish their triggers, sources and campaign use before recommending that either be removed. They could represent the same event, different milestones or an obsolete configuration. The immediate action is to resolve that ambiguity, not to delete the less familiar name.
Where the business reports unsuitable enquiries, create a separate diagnosis task using the Google Ads lead-quality review process. Record the dependency in the audit. The account-wide review needs to know whether its measurement and decisions are trustworthy; it does not need to reproduce a full enquiry-classification exercise.
Similarly, if the desired outcome exists only in a CRM, specify the missing measurement capability and assign an implementation owner. Keep import design, consent handling, identifiers and reconciliation in a dedicated implementation plan. This audit should state the decision blocked by missing evidence and what would unblock it, rather than quietly treating an unbuilt integration as available.
Choose verification appropriate to the claim. Inspecting settings verifies configuration. An authorized controlled event can test a particular path. Observing reporting after processing checks whether the expected record appears. None alone establishes long-term data completeness or commercial improvement. Record which layer has been checked and leave the others open.
Review structure, targeting and the promises people see
Judge account structure by the decisions it enables. Can the manager control spend for materially different services, markets or objectives? Can reports distinguish those purposes? Can someone identify the owner and destination of a campaign without guessing? Naming consistency helps inspection, but a naming problem is usually less consequential than an unmanageable budget or goal mismatch.
Create a compact campaign inventory: purpose, campaign type, budget source, bidding approach, selected goals, geography, language, schedule, primary destinations and responsible owner. Note shared dependencies. One landing page, conversion action or budget can affect several campaigns, so the unit of repair may be broader than the row where the symptom first appeared.
Do not prescribe a universal number of campaigns or ad groups. Splitting an account may improve control while increasing maintenance and reducing the evidence available within each segment. Consolidation may simplify operations while combining goals that should remain distinct. Explain the specific control or reporting problem before proposing either.
Use query evidence within its actual coverage
Review the available search-term evidence for material mismatches with the offer. State the report scope, period, filters and campaign types. Google explains that some queries are omitted for privacy reasons in its search terms report documentation. The visible rows are not a complete list of every search that contributed to account activity.
Keep the audit recommendation narrow: identify the supported mismatch, affected scope, approximate observed cost and the next decision. Do not extrapolate a small visible sample into an account-wide waste percentage. When campaign and report totals are comparable, record the unrepresented cost separately; otherwise state why a coverage calculation would be misleading.
A fictional report contains CAD 2,400 of visible query cost within a comparable campaign total of CAD 3,000. The visible share is 80%, leaving CAD 600 outside those rows. If CAD 180 of the visible cost is confirmed unrelated to the approved service, that is 7.5% of visible cost and 6% of the campaign total. Neither calculation establishes the relevance of the remaining CAD 600 or guarantees CAD 180 of future savings.
For implementation detail, use the existing guide to reviewing high-intent search queries alongside current platform documentation. The audit itself should identify the evidence and decision owner. It should not become another keyword-rebuild tutorial.
Use the reports available for each campaign type. Google now documents a Performance Max search terms report, including its own data and exclusion limitations. Avoid a blanket claim that these campaigns provide no query visibility. Equally, a search report does not describe every channel or placement within a multi-channel campaign.
Compare reach settings with service delivery
Check included and excluded locations, advanced location options, language settings and schedules against the brief. Google’s advanced location guidance distinguishes presence from presence or interest and says location detection is not perfectly accurate. A targeted city name therefore does not prove that every click represents a serviceable customer.
Recommend a setting change only after establishing the commercial requirement. A fictional local repair firm serving properties within a defined area may need a different reach policy from a fictional relocation service accepting enquiries from people planning to move there. Both need a documented rationale. Neither should inherit a location rule solely because another advertiser uses it.
For Canadian campaigns, have the business confirm municipalities, travel boundaries, remote-service availability and the languages the team can support through fulfilment. Verify opening hours and response coverage with operations. Record any mismatch between advertised availability and actual capacity as a finding with an operations owner, not merely an ad-copy task.
Inspect ads, assets and destinations together
Review the promise people see and the page or contact route that follows it. Check current offers, service areas, availability, phone numbers, destination URLs and the next step requested. Include relevant asset associations and any automated text or destination features enabled in the account. Record what is actually configured rather than assuming a feature is active.
Open destinations directly and document redirects, visible errors and the relevant desktop or mobile experience. Do not generate paid clicks to inspect your own landing pages. If a controlled submission or call is needed, arrange it with the receiving team, label it as a test and record the exact scope. A successful browser visit does not demonstrate successful form delivery.
Prioritize broken promises that affect active traffic. Cosmetic inconsistency can remain in the maintenance queue unless it prevents understanding or completion. A fictional expired offer on an active destination merits a specific correction and review; it does not justify replacing every campaign’s creative at once.
Treat policy and eligibility findings as separate work
Record the exact status, affected ad or asset, cited policy, platform decision date, first observed date and whether other eligible routes remain available. A disapproved ad, an ad under review and a limited-eligibility status call for different investigations. Do not collapse them into “Google is blocking the account.”
Google’s disapproval and appeal guidance explains how to inspect policy details and distinguishes correcting content from disputing a decision. Link the relevant policy in the finding and assign someone who can assess the actual offer, wording and destination. Check current appeal eligibility and submission limits before selecting a response. Record any required specialist review without claiming that an audit certifies legal or regulatory compliance.
Keep advertiser-verification requests, payment issues and policy restrictions distinguishable in the register. Capture the actual notice and deadline where one exists. Assign the person with the authority and documents to respond; the campaign manager may not be that person.
The verification criterion should describe the platform status that must be observed after the authorized response. Submission is not approval, and approval is not a guarantee of delivery or results. Leave the finding awaiting verification while a decision is pending. Do not invent a completion date for a process outside the team’s control.
Audit spending authority and budget exposure
Start with the business’s approved spending amount and its scope. Is it media spend only, or does it include management, creative, software and taxes? Which period and currency apply? The Google Ads cost-planning guide covers those broader distinctions. Here, the task is to check whether the live configuration and monitoring process respect the documented authority.
Inventory unique budget objects, their campaign associations, current amounts, budget types and scheduled changes. Google describes a shared budget as one average daily budget used by multiple campaigns. Count that budget once when assessing configured exposure. Separately inspect total-budget campaigns and any relevant account-level controls. Do not assume the sum of numbers visible in campaign rows represents a single enforceable account cap.
For most campaigns using an average daily budget, Google documents a daily spending limit of twice that amount and a monthly limit of 30.4 times that amount. These are qualified platform rules, not a statement that spending is identical every day. Check Google’s spending-limit documentation for exceptions and distinguish served cost from the amount ultimately billed.
A fictional account has two independent, unchanged average daily budgets of CAD 80 and CAD 120, active for a full month and subject to those standard limits. Their combined monthly reference is (80 + 120) × 30.4 = CAD 6,080. If the approved media envelope is CAD 5,500, the configuration exceeds that reference envelope by CAD 580. This establishes a control mismatch to resolve; it does not predict that the account will spend CAD 6,080.
When budgets change during a month, the simple full-month calculation is insufficient. Google’s budget-change rules explain the effect on spending limits, including use of the highest average daily budget selected that day for the day’s limit. Reducing a budget later the same day does not necessarily reduce that day’s maximum to twice the new amount.
Also identify campaigns using total budgets. Google describes these as a budget for a defined campaign duration, with different pacing from average daily budgets. Its campaign total-budget documentation states that these do not have the same daily spending limit. Do not apply the 30.4 calculation to them.
Assign a budget owner, review frequency, escalation threshold and authorized response. Set thresholds from the business’s actual tolerance and the relevant controls, not an industry benchmark. A monitoring alert is useful only if someone receives it in time and knows what they are permitted to do.
Understand change risk before choosing the sequence
Review what changed before interpreting a performance movement. Google’s change history documentation describes a two-year history and records changes made through several routes, including automated rules, the API and Ads Editor. Use timestamps to identify plausible explanations, then test those explanations against the affected campaigns and periods.
Inspect enabled rules, scripts, third-party management connections and opted-in recommendations. Record their owner, scope and intended behaviour. Google’s auto-apply management guidance explains how to review subscriptions and their history. Automation should have a documented purpose and an accountable reviewer. Its existence alone is neither a defect nor a reason to remove it.
A rollback plan needs more than “undo the change.” Save the previous configuration, identify who can restore it and establish the trigger for doing so. Google says most change types from the previous 30 days can be undone, with exceptions. Confirm what is possible for the specific change instead of relying on a universal undo button.
Some effects cannot be reversed by restoring settings. Money already spent, missed enquiries and elapsed review time remain part of the result. A conversion-goal change may also alter how later performance should be interpreted. Keep a dated annotation and preserve the old baseline so the team does not compare unlike periods.
Distinguish the risk of waiting from the risk of changing. A confirmed broken contact route may need prompt containment. A proposed restructuring based on incomplete data may deserve further investigation. Where several campaigns share a dependency, review the complete affected scope before changing the shared component.
Prioritize with explicit judgement, not a score out of 100
Use severity to describe the consequence of the finding, evidence confidence to describe what is known, and change risk to describe what implementation could disrupt. These are separate dimensions. A potentially serious access issue can have limited evidence. A minor naming defect can be fully confirmed. Neither becomes more important through arithmetic alone.
The following priority lanes are a proposed working method, not a validated industry scale. Agree their meaning with the business before applying them. Do not add them into an account-health percentage, and do not equate a large count of findings with a poor account.
| Lane | Use when | Required next decision |
|---|---|---|
| Contain | Credible evidence indicates an active security, spending-authority or critical delivery problem | Authorised owner decides immediate containment and preserves evidence |
| Repair | A consequential mismatch is supported and a bounded correction is understood | Approve the specific change, dependencies and verification plan |
| Investigate | The consequence could matter but evidence or cause remains incomplete | Approve a defined evidence-gathering task with a decision deadline |
| Improve | The current arrangement works but a supported alternative may help | Choose a controlled evaluation with an agreed success criterion |
| Monitor or accept | The risk is tolerable, the change is not worthwhile, or current evidence supports no action | Record the rationale, owner and condition for reopening |
Within a lane, resolve dependencies first, then consider affected scope, ongoing exposure, reversibility and the effort required. An evidence task that enables a budget decision may outrank several easy cosmetic fixes. Estimate effort in a way the team can defend, and mark missing cost estimates as unknown. Do not invent savings to make a recommendation look attractive.
Keep Google’s optimization score in context. Google describes it as a settings-and-performance estimate, and applying or dismissing recommendations changes that score. Treat recommendations as items to assess against the brief. A higher platform score does not by itself demonstrate profitable customers, secure ownership or compliance with the business’s approval process.
If stakeholders disagree, record the disputed assumption and the evidence that could settle it. A decision to defer should identify the cost of waiting where it can be supported, the review date and the person accepting the uncertainty. “We will revisit this” without an owner or trigger is an unfinished action.
Use the audit scorecard and evidence register
The companion contains a blank scorecard, a separate evidence register, a field dictionary, instructions and fictional worked records. It is designed for CSV import into a spreadsheet. It does not connect to Google Ads, retrieve account data or make changes. Keep a clean copy of the templates and use a separate working copy for each account and audit date.
Use one scorecard row per decision, not one row per symptom. Several screenshots may support one goal-selection finding. Conversely, an access issue and a budget issue need separate rows even if they were discovered together. Link evidence by ID so the same export can support several findings without being duplicated.
- Complete the audit scope in the instructions: account alias, dates, currency, time zone, business objective and authority.
- Add evidence records with their capture dates, reporting periods, filters, storage references and limitations. Keep sensitive material outside the scorecard.
- Write each observation and consequence separately. Choose severity, confidence and priority, then state the unresolved question.
- Assign a named owner, dependency, proposed action and decision deadline. Record approval before implementation.
- Save the baseline, rollback trigger and verification criterion. After action, attach verification evidence and record the result or remaining uncertainty.
Blank evidence does not mean no issue was found. Mark an unreviewed area as unavailable or not reviewed and assign the next evidence request. Use “not applicable” only with a reason. For accepted or deferred findings, complete acceptance_rationale, reopen_condition and next_review_due, and retain the accountable owner. These fields keep an agreed risk visible after the initial review.
The audit-calculation-checks.csv file includes explicit inputs, units and expected outputs for the two fictional scenarios in this guide. Its formulas are written as plain-text instructions, and its result fields are static values. Recalculate every dependent result after changing inputs; the file is not a live calculator. A blank input is missing, not zero. Use a percentage only when its denominator is positive and the costs have comparable scope. The accompanying checks validate local synthetic data, not a Google Ads account or integration.
Worked fictional review: three findings, three different decisions
Consider a fictional Canadian service business reviewing two campaigns before increasing spend. The audit brief approves CAD 5,500 in monthly media spend. Its outcome is an accepted service enquiry. No customer account was accessed to create this scenario, and the findings below do not represent Canada Create client results.
Finding one: configured budgets exceed the approved reference
The fictional settings export shows independent daily budgets of CAD 80 and CAD 120. Under the unchanged, full-month assumptions described earlier, the combined reference is CAD 6,080. The approved brief documents CAD 5,500. The evidence supports a CAD 580 configuration gap; actual final spend remains unknown.
Assign Repair with high confidence in the mismatch, and require the budget owner’s decision before any increase proceeds. The proposed action is to reconcile the configured budgets with the authorized envelope, including other spending dependencies. Preserve existing values, document the approved replacement and verify the saved configuration and subsequent cost reports. The owner should separately decide whether monitoring requires immediate escalation.
Finding two: a supporting action is included in a custom goal
The fictional evidence shows a brochure interaction included in the custom goal used by both campaigns. The business brief identifies an accepted service enquiry as the desired outcome. Even if the brochure action carries a secondary label, its custom-goal use matters. The setting is observed; its effect on sales has not been established.
Assign Repair, with the measurement owner responsible for checking the complete goal map before proposing a change. Approval should specify the intended goal composition and affected campaigns. Verification includes the resulting settings and reporting interpretation, followed by a later outcome review. Do not promise a lower acquisition cost simply because the goal definition becomes more appropriate.
Finding three: a performance decline follows several changes
The fictional account’s enquiry count fell after a destination release, a targeting edit and a staff-coverage change. Their timing makes each a plausible contributor, but it does not identify the cause. Mark confidence as limited and the priority as Investigate. Avoid presenting a broad account rebuild as the established remedy.
Assign an owner to assemble the relevant change records and compare affected routes, reporting periods and operational coverage. Set the next decision as whether evidence supports a specific repair or a controlled evaluation. A further inspection may reveal a functional failure requiring immediate action; until then, preserve the uncertainty rather than converting it into a sales forecast.
Close findings with evidence and maintain the plan
Finish the audit with a decision meeting, not a slide count. Show the material findings, areas not reviewed, proposed sequence, dependencies and approvals needed. Keep supporting detail available for inspection. The business owner should be able to explain why the first action comes first and what evidence would justify changing the order.
For each approved action, separate implementation completion from verification. “Saved” confirms an edit. A fresh settings capture confirms the intended state. An authorized functional check can confirm a specific path. A later, sufficiently mature performance review addresses outcomes. Use the status “awaiting verification” until the promised check is complete.
Define success before making the change. A budget-control repair succeeds when the approved configuration is in place and the monitoring responsibility is functioning. An access repair needs confirmation that required administrators can still perform their work. A performance experiment needs its own outcome measure, comparison basis and stopping rules. These are different acceptance criteria.
Keep unsuccessful or inconclusive work in the record. If a change introduces a problem, follow the approved rollback trigger and document what was restored. If results remain ambiguous, retain the finding as unresolved or accept the uncertainty explicitly. A completed task list is not evidence that every underlying risk disappeared.
Choose a follow-up cadence based on exposure and change frequency. Review active incidents as their evidence develops; review consequential repairs after implementation and again when outcome data is usable. Reopen the broader audit after a major handover, goal change, spending expansion or website release. Reuse the same evidence definitions so the next review can distinguish new problems from old assumptions.
Platform documentation was checked on 8 October 2026. Interfaces and available features can change; verify the current setting and official instructions before implementation. The method and priority lanes in this guide are practical editorial recommendations. The fictional examples illustrate decision-making and arithmetic, not measured campaign results.
Bring your completed scorecard to Canada Create’s Google Ads team to discuss an authorized account review.
