Paid Search Audit: 12 Places Marketing Budgets Leak

Paid search accounts rarely fail in one obvious place. Spend usually leaks through a series of small, defensible-looking decisions: a conversion action that counts the wrong thing, a match type that drifted wider than intended, a bidding strategy trained on low-value signals, a landing page that quietly stopped working on mobile. Each leak looks minor in isolation. Together they decide whether the account buys revenue or buys activity.

The thesis of this article is simple: a paid search audit should start with measurement, not with keywords. If the account is optimizing toward the wrong signal, every downstream improvement makes the wrong outcome more efficient. Only once you trust what is being counted does it make sense to review queries, bids and budgets.

The 12 checks below are grouped in the order we recommend working through them. They apply in principle to Google Ads and Microsoft Advertising; interface names and defaults change over time, so treat platform-specific wording as a pointer to verify in the current account, not as a fixed rule.

How to use this audit

Work through each check and record three things: what you found, the evidence (screenshot, export, setting, date range), and a severity rating. Use transparent definitions so that anyone reviewing the audit can challenge the rating rather than accept it on trust.

SeverityDefinition
CriticalThe account is optimizing toward, or reporting, a signal that does not reflect business value. Decisions based on current reports are unreliable.
HighA material share of spend is exposed to traffic or settings that are unlikely to produce qualified outcomes, and the cause is identifiable.
MediumEfficiency is reduced, but measurement is sound and the issue is contained to specific campaigns or segments.
LowHygiene or governance issue with limited current cost, but a risk if the account scales.

“Material share” should be defined with the budget owner before the audit begins, as a proportion of spend that matters to that business. We deliberately avoid universal thresholds: what counts as material for a small lead-generation account differs from a large retail catalog.

Measurement leaks

1. Conversion actions that count activity instead of value

What to check: list every conversion action marked as primary (or otherwise included in the bidding goal). For each, confirm what triggers it: a form submission, a thank-you page view, a click on a phone number, a scroll, a page visit.

Why it matters: automated bidding optimizes toward whatever is counted. If a button click or a page view is treated as a conversion, the system will find people who click buttons, which is not the same as finding buyers. This is the most common root cause behind accounts where reported conversions rise while sales do not.

2. Duplicate or inflated conversion counts

What to check: compare conversion counts in the ad platform with the source system (CRM, order database, form backend) for the same period. Look for double-firing tags, the same action imported from both an analytics property and a native tag, and “every” counting applied to lead forms where “one” would be appropriate.

Why it matters: inflated counts make cost per conversion look better than it is and feed bidding algorithms with false positives. A reconciliation does not need to match perfectly, but the gap should be explainable.

3. Missing downstream outcomes and consent gaps

What to check: whether qualified-lead, opportunity or revenue stages are fed back into the platform (for example through offline conversion imports), and how consent choices affect what is measured. Review whether consent mode or an equivalent approach is implemented as intended in the regions where it is required, and whether modeled conversions are being interpreted as observed ones.

Why it matters: without downstream signals, the platform cannot tell a lead that became a customer from one that never answered the phone. Without a clear view of consent coverage, teams can misread a measurement change as a performance change.

Targeting and query leaks

4. Irrelevant search terms

What to check: export the search terms report for a meaningful period and classify terms by intent: relevant, adjacent, informational, job-seeking, competitor, unrelated. Sum spend and outcomes by class. On large accounts an AI model can do the first pass of this classification quickly; treat its labels as a draft, spot-check a sample against your products, and never add negative keywords from model output alone.

Why it matters: this is the classic leak, and it is still worth checking. The useful output is not a long negative keyword list, but an understanding of why irrelevant terms are matching, which points to structural fixes in checks 5 and 6.

5. Match types and automation without guardrails

What to check: which keywords use broad match, how much spend they carry, and whether they run alongside a bidding strategy and conversion signal that can actually guide them. Check negative keyword lists for coverage and whether shared lists are applied to the campaigns that need them.

Why it matters: broad matching can work when the conversion signal is strong and the account has enough data. When the signal is weak (see check 1), broad matching plus automated bidding tends to expand into traffic the business would never have chosen deliberately.

6. Brand and non-brand mixed together

What to check: whether branded queries run in separate campaigns from non-brand queries, and whether non-brand campaigns are picking up brand terms. Review whether brand spend is reported separately in performance summaries.

Why it matters: brand searches typically convert well because the person already intends to find you. Mixing them into non-brand reporting makes prospecting look more efficient than it is, and can hide how much of the account is paying for demand that might have arrived anyway. How much brand spend is incremental is a separate question that requires testing, not assumption.

7. Location, network and audience settings

What to check: location targeting options (people physically in a location versus people showing interest in it), inclusion of search partner networks or display expansion, language settings, ad schedule and device adjustments. Segment performance by each where the data allows.

Why it matters: default or legacy settings often widen reach beyond the intended market. These leaks are easy to fix once seen, and easy to miss because they sit in campaign settings rather than in reports people look at daily.

Bidding and budget leaks

8. Bidding strategies trained on the wrong goal

What to check: the bidding strategy for each campaign, its target (CPA, ROAS or none), the conversion actions it optimizes toward, and when targets were last changed. Look for targets that were set once and never revisited, and frequent target changes that keep the system in a learning state.

Why it matters: a target CPA or target ROAS is only as good as the value behind it. If lead values are flat placeholders, or revenue values ignore margin and returns, the algorithm will chase volume that looks efficient on paper.

9. Budget allocation and pacing

What to check: which campaigns are limited by budget and which have spare budget, and whether that allocation matches business priorities. Compare spend distribution with the distribution of qualified outcomes, not platform conversions alone.

Why it matters: budgets often reflect history rather than current evidence. A campaign that is limited by budget but produces qualified pipeline may be starved while a well-funded campaign produces low-quality volume.

10. Overlapping and opaque campaign types

What to check: whether automated campaign types (such as Performance Max) overlap with standard search or shopping campaigns on the same products, audiences or brand terms, and what reporting is available on where their spend goes. Check brand exclusions and asset-level reporting where offered.

Why it matters: overlapping campaigns can compete for the same users and make attribution between them unreliable. Limited transparency is not automatically a problem, but it should be a known, documented risk rather than an unexamined default.

Landing page and offer leaks

11. Message mismatch between ad and landing page

What to check: for the highest-spend ad groups, open the landing page and compare it with the query and ad text. Does the page answer the specific intent, show the same offer, and make the next step obvious? Check that final URLs are not pointing to generic pages by default.

Why it matters: paying for a click and then sending the visitor to a page that does not continue the conversation is one of the most expensive forms of waste, because the targeting work was done correctly and then lost at the last step.

12. Broken forms, slow pages and lead routing

What to check: submit test leads on mobile and desktop, confirm they arrive in the CRM with source data intact, and confirm someone follows up. Check page load behavior on a mid-range mobile connection and whether tracking parameters survive redirects.

Why it matters: a technically broken conversion path wastes all spend directed at it. Lead routing failures are particularly costly because the platform may still record a conversion, so nothing looks wrong in the ad account.

Turning findings into priorities

A list of 12 issues is not a plan. Rank findings using two questions: how much spend does this affect, and how confident are we that fixing it changes outcomes? Measurement problems (checks 1 to 3) usually go first even when the direct spend affected is unclear, because every other fix depends on them.

PriorityRule
Fix firstCritical severity, or High severity affecting a material share of spend with strong evidence
Fix nextHigh severity with partial evidence, or Medium severity affecting a material share of spend
TestPlausible improvement where the effect is uncertain (for example brand spend incrementality, bidding target changes)
MonitorLow severity, or issues dependent on a measurement fix that has not yet landed

Ranking paid search findings

Findings sorted by how much spend they affect and how confident we are that fixing them changes outcomes: high spend with strong evidence is fixed first, high spend with uncertain effect is tested, smaller issues with strong evidence are fixed next, and the rest are monitored.

Spend affected →

Test — large spend, uncertain effect

  • Brand spend incrementality
  • Bidding target changes

Fix first — large spend, strong evidence

  • Conversion actions counting activity
  • Duplicate conversion counts

Monitor — small spend, uncertain effect

  • Issues waiting on a measurement fix
  • Low-severity settings

Fix next — small spend, strong evidence

  • Irrelevant search terms in one campaign
  • Broken form on a secondary landing page

Confidence the fix changes outcomes →

Example placements; measurement problems (checks 1–3) usually go first because other fixes depend on them.

The “Test” category matters. Some apparent leaks, such as brand spend or broad match expansion, may or may not be waste depending on incrementality. Those deserve a structured experiment with a defined hypothesis and decision rule, not a unilateral change.

Caveats

An audit identifies where spend is exposed to risk. It does not, on its own, prove how much money a fix will save or earn. Treat any estimated impact as a hypothesis to validate.

Platform interfaces, defaults and campaign types change frequently. Verify every setting in the live account rather than relying on documentation or past experience.

Reconciliation between ad platforms and CRM data will rarely match exactly because of attribution windows, consent, time zones and deduplication rules. The goal is an explainable gap, not a perfect match.

Some changes, such as tightening match types or adding negatives, can reduce volume before efficiency improves. Agree success criteria and a review period before making them.

Action checklist

  • List every primary conversion action and confirm what triggers it
  • Reconcile platform conversions with CRM or order data for the same period
  • Confirm whether qualified-lead or revenue outcomes are imported back into the platform
  • Classify search terms by intent and sum spend by class
  • Review broad match usage against the strength of the conversion signal
  • Separate brand and non-brand campaigns and reporting
  • Audit location, network, language and schedule settings
  • Check each bidding target against the value it is optimizing toward
  • Compare budget allocation with qualified outcomes, not platform conversions alone
  • Document overlap between automated and standard campaigns
  • Compare top ad groups with their landing pages for message match
  • Submit test leads and confirm they reach the CRM and a person

If you want a second opinion on where your account is leaking, our paid traffic audit follows this structure and separates confirmed issues from hypotheses that need testing. You can request an independent review with as much or as little account detail as you are comfortable sharing at first.

Want an independent view of your own setup?

We review teams, spending and channel performance against your commercial goals, and deliver a prioritised, evidence-backed plan.