Master Google Ads Conversion Tracking Audit 2026

If Google Ads reports conversions that are more than 15% off from your CRM or backend for the same period, I treat that as a tracking failure, not a bidding problem, and I stop optimization work until it's fixed (Black Propeller).

I'm Chase McGowan. I run Google Ads as an independent operator, not an agency with layers. When an account is spending real money, usually $10k to $100k+ per month, the fastest way to waste budget is to trust broken conversion data and then start changing bids, match types, budgets, or Performance Max settings as if the numbers are real. They often aren't.

I've seen too many accounts where the platform looked fine on the surface, but the tracking underneath was skewing every decision. That's why my first move in a Google Ads conversion tracking audit is simple: verify measurement before touching optimization. If the inputs are wrong, Smart Bidding learns the wrong thing. If the reported lead count is inflated, CPA looks better than reality. If purchase values are wrong, ROAS becomes fiction.

Table of Contents

• Why Audit Tracking Before Bidding
• Confirm Account Level Settings
  • Start inside Google Ads before you open GTM
  • The settings I check first
• Verify Tag Implementation with GTM and Gtag
  • What I inspect on the site itself
  • What has to fire and how I test it
• Check Analytics Alignment in GA4 and Universal Analytics
  • Why GA4 and Google Ads rarely match cleanly
  • What I compare when numbers drift
• Set Up Advanced Conversions and Offline Imports
  • Why client-side tracking leaves gaps
  • How I handle offline imports
• Validate Attribution and Deduplication Settings
  • Count settings break lead gen constantly
  • Attribution and values need to match the real outcome
• QA Testing Pitfalls and Remediation Reporting
  • The QA tests that catch what setup reviews miss
  • How I document fixes so someone can actually implement them

Why Audit Tracking Before Bidding

I don't optimize bids on top of broken data. That's guesswork dressed up as account management.

A lot of Google Ads advice starts with campaign structure, keyword cleanup, negative lists, tROAS targets, or Performance Max segmentation. Some of that matters. None of it matters first if the conversion layer is lying. A rigorous audit source found that 87% of high-spend accounts fail the initial tag presence verification across critical pages, largely because Conversion Linker is missing or misconfigured (Negator). That tells you how common the problem is in accounts that are already spending serious money.

Practical rule: if measurement is unstable, every optimization decision downstream gets weaker.

Often, businesses get stuck with agency reporting. The dashboard looks active. There are meetings. There are bid changes. But nobody has proven that the primary conversion is counted once, valued correctly, and tied back to something the business values.

For businesses trying to improve lead quality and understand what customers do before they convert, broader customer analysis also matters. That's why I like resources on strategies to improve client experience. They help frame the behavior side of the problem. But in Google Ads, behavior analysis only becomes useful after the conversion plumbing is trustworthy.

My bias is simple. Fix tracking first, then optimize. That's how I work in the 30-day Google Ads Sprint. Hire me for 30 days before you hire me forever.

That approach has held up in real accounts. In one premium ecommerce account, platform-reported revenue went up $6.65M on $119K less spend, with ROAS moving from 13x to 33x over three years. That kind of outcome doesn't come from random bid tweaks. It comes from feeding Google Ads cleaner signals, tighter conversion definitions, and better value data before asking the platform to scale.

Confirm Account Level Settings

Bad account settings can make clean tags look fine and still poison bidding.

I check Google Ads itself before I touch GTM, because a lot of expensive tracking problems start in conversion settings, attribution choices, or account configuration. In $10k to $100k-plus monthly accounts, that matters more than people want to admit. Agency layers often blur this part. The dashboard shows conversions, the tags fire, and nobody stops to ask whether Google Ads is optimizing toward the right action in the first place.

Start inside Google Ads before you open GTM

The first question is simple. Can this account's conversion reporting be trusted enough to guide spend?

If Google Ads and the CRM tell very different stories, I stop there and investigate the setup before I change bids, budgets, or campaign structure. I do not care how polished the reporting deck looks. I care whether booked revenue, qualified leads, or actual sales line up closely enough to make optimization decisions without guessing.

The settings I check first

I review these in a fixed order because each one changes how I interpret the rest.

1. Time zone alignment. Google Ads, GA4, the CRM, and the client's internal reporting need to use a consistent time basis, or at least a clearly understood one. If they do not, day-level and week-level comparisons become noisy, and people start reacting to reporting artifacts instead of real changes.

2. Primary and secondary conversion actions. This setting drives bidding behavior. If a micro-conversion such as a page view, a scroll event, or an early form step is marked as primary, Smart Bidding will happily buy more of it. That usually looks good in-platform and looks bad in the sales pipeline.

3. Goal inclusion by campaign. I check account-default goals and any campaign-level overrides. A common mess in mature accounts is old goals lingering in the background, still included in “Conversions,” still feeding bidding, and no one remembers why.

4. Conversion windows. The click-through and engaged-view windows need to fit the actual buying cycle. Short windows can suppress legitimate conversions in higher-consideration accounts. Long windows can over-credit campaigns that had only a light touch.

5. Count settings. For ecommerce, “Every” is often right. For lead gen, “One” is often safer for the core lead action. I still verify the business model before deciding. A repeat purchase and a duplicate lead are not the same thing, and Google Ads should not treat them the same way.

6. Conversion source and status. I want to know whether the account is using website tags, imported GA4 events, offline imports, phone call tracking, or some mix that grew over time without a cleanup. I also check whether old actions are inactive, unverified, or still sitting in the account as clutter that confuses reporting.

7. Call conversion rules. In call-heavy accounts, I check minimum call length, call reporting source, and whether the account counts any answered call as success. For many local service and healthcare advertisers, that inflates reported performance fast.

One wrong primary action is enough to send bidding in the wrong direction for weeks.

I also compare the conversion names and values in Google Ads to the language the business uses internally. If sales talks about qualified consults and closed deals, but Google Ads is optimized to “Submit Lead Form,” there is already a translation problem. Those mismatches are common in inherited accounts. They happen after agency handoffs, rushed migrations, or years of incremental edits with no owner keeping the conversion framework clean.

For lead gen, I usually map each conversion action to a business stage. Inquiry. Qualified lead. Appointment set. Sale. If Google Ads only sees the first stage, I note that gap immediately. That does not mean the account is unusable. It means the bidding model is operating on shallow feedback, and any performance read has to be treated with more caution until better data comes in. I write about that kind of tracking-first cleanup in the Google Ads audit and strategy articles on my blog.

If account settings are off, a tag can fire perfectly and the account can still optimize toward the wrong outcome.

I also look for resilience gaps at this stage. If the account depends only on client-side website conversions, I flag that before the technical review starts. Browser limits, consent behavior, and messy handoffs between forms and CRMs can all weaken what Google Ads receives. That trade-off might be acceptable for a simple account. It is rarely acceptable once spend is high enough that small measurement errors change bidding decisions.

Verify Tag Implementation with GTM and Gtag

A lot of wasted spend starts here. The account can be structured well, bids can be sensible, and reports can still be wrong because the tag fires at the wrong moment, fires twice, or never carries the click data through the path that matters.

For advertisers spending $10k to $100k+ a month, I treat this as a hands-on inspection, not a checkbox review. Agency layers often leave behind partial GTM setups, legacy hardcoded tags, and conversion actions nobody has validated in months. A single-operator audit has one advantage here. It strips the setup back to what fires, what gets counted, and whether that count deserves to guide bidding.

What I inspect on the site itself

I start with the pages and states that can create or break attribution. Homepage, high-intent landing pages, product or service pages, cart or lead flow, and the final confirmation state. If the site uses direct gtag installs, I inspect page source and network requests. If it uses GTM, I use Preview Mode and trace the event path from trigger to tag to request.

I also check whether Google Ads tags and GA4 tags are split across different methods. That mixed setup is common in inherited accounts. It can work, but it often creates drift because one event is managed in GTM while another sits hardcoded on the site with no version control.

The misconfigurations that show up most often are predictable:

• Wrong trigger logic. The conversion fires on button click, form start, or field interaction instead of a confirmed success state.
• Duplicate implementations. The same Google Ads conversion exists in GTM and in hardcoded site scripts.
• Wrong conversion ID or label. The request fires, but it maps to the wrong Google Ads action.
• Static values. Every purchase or lead is sent with the same number, which makes value-based bidding less useful.
• Currency problems. Value passes, but the currency code is missing or wrong.
• Broken click ID persistence. Redirects, third-party forms, or cross-domain steps drop the Google click identifier before conversion.

Conversion Linker gets its own check every time. If it is missing or misfiring, click identifiers are more likely to break across redirects, browsers, and longer paths. That problem is easy to miss because the tag interface can still look clean while attribution quality degrades underneath.

What has to fire and how I test it

I run real tests. Screenshots from a prior setup or a green check in GTM are not enough.

For a lead gen account, I submit the primary form myself and watch each step in Preview Mode and in the browser network panel. For ecommerce, I test the purchase path as far as the client will allow. For calls, I check the call reporting setup and confirm what event is supposed to count, on-site call click, forwarding number call, or imported qualified call outcome.

The standard is simple. The primary conversion should fire once, on the final success state, with the right conversion ID, label, value, and currency. If one form submission creates two conversion requests, reported CPA is inflated and smart bidding starts learning from noise.

I also compare what I see in the browser with what appears in Google Ads diagnostics after testing. A tag can fire in GTM preview and still fail in production because of consent settings, redirect behavior, script conflicts, or filters that only affect live traffic. That is why I test both the trigger logic and the actual request leaving the browser.

One practical check catches a surprising number of bad setups. Submit the main form once, then refresh the thank-you page and watch what happens. If the conversion fires again, the trigger is tied to page load with no guardrail. If the thank-you page is indexable or easy to revisit, that error can sit in the account for months.

I keep notes at the level that matters for remediation: page tested, trigger used, tag name, conversion ID and label, whether value populated, whether the request fired, and whether the count matched one real action. That gives a developer or in-house marketer something they can fix without another layer of interpretation. I share more of that tracking-first audit approach in the Google Ads audit articles on the Come Together Media blog.

The rule for this section is blunt. Count the actual outcome once, and only once.

Check Analytics Alignment in GA4 and Universal Analytics

Even with tags firing correctly, Google Ads and GA4 can still disagree.

That isn't automatically a problem. Different platforms see different parts of the journey and use different attribution logic. What matters is whether the mismatch is explainable and stable. What I don't accept is random drift that nobody can account for.

Why GA4 and Google Ads rarely match cleanly

GA4 is event-based analytics. Google Ads is an ad platform that reports conversions according to the conversion actions and attribution settings inside the account. If one side is counting a purchase event and the other is counting a different imported action, the numbers won't line up.

I compare the setup before I compare the totals:

A lot of mismatches come from soft conversions being imported into Google Ads because they were easy to set up, not because they were the right signals for bidding.

What I compare when numbers drift

When I audit alignment, I segment by device, browser, landing page, and conversion action. I don't just compare account totals. Problems usually cluster. If mobile form submissions look weak in Google Ads but normal elsewhere, I focus there. If GA4 shows events but Google Ads misses them, I check tag routing and conversion settings. If both platforms look incomplete, I look at the form process itself.

Stable disagreement is manageable. Unexplained disagreement isn't.

I still run this section with Universal Analytics references in mind because older setups often leave behind historical imports, legacy event naming, or old assumptions that still affect reporting behavior. The point isn't to preserve UA. The point is to find inherited conversion logic that still shapes the account.

I also check consent mode handling here. If consent state isn't passed cleanly, some conversions can disappear from observed data and confuse the comparison. That doesn't mean the ads failed. It means the measurement setup needs to be interpreted carefully and, if possible, tightened.

For business owners, this is the practical takeaway: don't ask whether GA4 and Google Ads match perfectly. Ask whether both systems are measuring the same action, from the same trigger, with the same value logic. If they aren't, the gap isn't a mystery. It's configuration.

Set Up Advanced Conversions and Offline Imports

For advertisers spending $10,000 to $100,000 or more per month, browser-only conversion tracking usually leaves enough missing data to skew bidding. I treat this as a data integrity issue first, especially in accounts that have already been passed through agency layers and patched together over time.

Why client-side tracking leaves gaps

A browser tag can record the thank-you page and still miss part of the conversion path. Safari restrictions, consent behavior, ad blockers, form handoff issues, and cross-device journeys all reduce match quality. In an audit, I do not assume a firing tag means clean measurement.

One LinkedIn post on server-side tracking notes that advertisers can lose a meaningful share of observable conversions without Enhanced Conversions, and argues that pairing Enhanced Conversions with server-side GTM is where best practice is heading (LinkedIn post on enhanced conversions and server-side tracking). That lines up with what I see in larger lead gen accounts. If the account is bidding on incomplete browser data, Smart Bidding still optimizes. It just optimizes against a thinner version of reality.

Enhanced Conversions improve match rates by sending hashed first-party data such as email address or phone number after the conversion. They do not fix broken triggers, bad forms, or poor action design. They improve recovery and matching once the base setup is working.

My audit questions are simple:

• Is Enhanced Conversions enabled on the primary conversion action used for bidding?
• Is the site collecting the customer data needed to support matching?
• Is the account still relying on browser-only tracking for high-value decisions?
• Would server-side GTM solve an actual measurement problem here, or just add complexity the business will not maintain?

That last point matters. I do not recommend server-side tagging by default. For some accounts, Enhanced Conversions plus clean on-site tagging is enough. For others, especially high-CPC lead gen with long sales cycles, browser-only tracking is too weak to trust.

How I handle offline imports

For lead generation, the form fill is usually the first recorded event, not the result the business wants. Qualified leads, booked calls, attended appointments, approved applications, and closed revenue are better signals if the CRM can support them.

Offline conversion imports fix that gap by sending back the Google click ID with later-stage outcomes. Done properly, this changes bidding from "find more people who submit forms" to "find more people who become real pipeline or revenue."

I check four things during the audit:

1. GCLID capture: The click ID needs to be stored in the form flow or CRM record at the time of lead creation.
2. Status mapping: Sales stages need clear rules so only meaningful milestones get imported.
3. Import timing: Uploads need to happen often enough to be useful for optimization.
4. Value logic: Imported values need to reflect actual pipeline or revenue logic, not placeholder numbers.

A common failure looks like this: the account records every form as a primary conversion, but the business only closes 10 to 15 percent of those leads. Google Ads keeps chasing cheap submissions because no one ever fed back the qualified lead or sale. The tracking is technically "working," but the optimization target is wrong.

Digitelia makes a similar point in its Google Ads audit checklist. Low-volume, high-ticket accounts often get better results from feeding back lead quality than from forcing more top-of-funnel conversion volume. I see that often in legal, medical, home services, and premium B2B.

If the build path is still unclear, the practical options are internal implementation, developer support, or a specialist who can handle tracking and account cleanup together. One factual option is Google Ads tracking and measurement support through Come Together Media services, which includes GTM, GA4, call tracking, and server-side tagging work where it fits.

Start with one check inside Google Ads. Open the primary conversion action you use for bidding and confirm whether Enhanced Conversions are enabled. Then trace whether the account can connect that click to a qualified lead or closed sale later in the CRM. If the answer is no, the audit should treat offline imports as a priority fix, not a nice-to-have.

Validate Attribution and Deduplication Settings

Plenty of accounts look profitable until you check how conversions are counted. In lead gen, I often find the platform reporting more "wins" than the business created because duplicate submissions, page reloads, or mismatched imports were never cleaned up.

Count settings break lead gen constantly

Analysts at Verde Media found that many lead-generation accounts had the Count field set to Every instead of One for form submissions, which inflated reported conversions and distorted revenue reporting in ecommerce setups using fixed values. I see the same pattern in accounts spending $10,000 to $100,000 a month, especially after multiple agencies, freelancers, or in-house teams have touched tracking without one owner checking how Google Ads will learn from the signal.

For lead generation, One is usually the right setting for form fills, calls, and booked appointments. A page refresh, duplicate webhook, or CRM retry should not create a second primary conversion.

For ecommerce, Every is usually correct for purchases. But that only works if each transaction has its own order ID and the same purchase is not being sent from both the website tag and an imported source without a deduplication rule.

I audit count settings with the business model in front of me, not from a generic checklist:

• Lead forms: usually One
• Phone calls counted as leads: usually One
• Purchases: usually Every
• Newsletter signups, page views, scroll depth, PDF downloads: usually secondary, not bidding targets

The ugly version of this mistake is common. Google Ads shows a stable CPA, the agency report looks fine, and sales teams are still complaining about lead quality or volume. The account is optimizing to duplicate actions, not actual demand.

A low CPA built on duplicate lead counts is a reporting error, not efficiency.

Attribution and values need to match the real outcome

Value settings matter just as much as count settings. If every purchase sends the same number, or every lead gets assigned an arbitrary amount no one uses in forecasting, bidding cannot separate high-value outcomes from cheap noise.

In ecommerce, I check whether Google Ads receives the accurate transaction value, the right currency, and one unique order reference. In lead gen, I check whether values reflect something the business can defend, such as appointment type, location, product line, or qualified pipeline stage. Placeholder values create fake ROAS. Fixed values can still be useful, but only if they are deliberate and tied to actual close-rate math.

Attribution comes after that. First clean up what counts as a conversion. Then make sure that conversion represents something worth bidding toward.

I am less interested in finding the perfect attribution model than in finding preventable inconsistency. If the account changed from one model to another and nobody documented the date, period-over-period comparisons are compromised. If Google Ads uses one logic while CRM reporting uses another, the team starts arguing about whose numbers are right instead of fixing the setup.

A practical audit here includes four checks:

1. Primary conversion actions are limited to outcomes worth bidding on
2. Duplicate submissions and imported records are deduplicated with a stable key, usually order ID or click-linked CRM record
3. Values reflect real revenue or a defensible lead value model
4. Attribution changes are documented so performance shifts are not misread

For advertisers who are tired of agency layers, this is the part that usually explains why scaling stalled. Bidding strategies cannot recover from bad conversion logic. If you want a reference point for what cleaned-up tracking looks like in practice, review these Google Ads and measurement case studies.

QA Testing Pitfalls and Remediation Reporting

A tracking setup can look clean in GTM, fire in preview mode, and still fail the moment real users hit the site. That gap is where bad optimization decisions start. For advertisers spending $10,000 to $100,000 or more each month, I treat QA as the point where an audit either becomes useful or turns into documentation theater.

The QA tests that catch what setup reviews miss

I do not sign off on conversion tracking until it passes three real-world paths: a new desktop visitor, a real mobile user, and a returning user. That framework lines up with common failure points identified by Clixtell, but the value is practical. Each path breaks in different ways, and polished implementation docs rarely reflect that.

Desktop often passes first. Mobile is where broken form listeners, sticky headers, slow page loads, and consent prompts interfere with events. Returning-user tests expose attribution loss, especially when a site uses redirects, third-party forms, or session resets between landing and submission.

My baseline checklist is simple:

1. New desktop user
   Click through from ad to landing page to thank-you state. Confirm the intended primary conversion fires once, with the expected parameters and value.

2. Mobile user
   Run the same journey on an actual phone. Mobile browsers, autofill behavior, tap interactions, and consent flows create tracking failures that desktop preview mode will not catch.

3. Returning user
   Visit, leave, return later, and convert. Common issues observed in this journey include gclid loss, overwritten cookies, cross-domain breaks, and duplicate event logic.

I also test failure scenarios on purpose. Submit the form twice. Refresh the thank-you page. Use a blocked-cookie browser. Decline consent, then accept it on a second visit. If tracking only works in the clean path, it is not ready for bidding.

The patterns I find most often are predictable:

• Double-counting. A thank-you page fires on load and a separate event fires on form submit.
• Phantom conversions. Page views, button clicks, or partial form completions are marked primary even though no business outcome happened.
• Consent-related gaps. The site records conversions only for users who accept immediately, but nobody has documented what that does to reported volume.
• Identifier loss. Auto-tagging is on, but redirects, embedded forms, or payment steps strip the click identifier before conversion.

A useful account-level sense check is repeat behavior in the primary conversion action. If one lead routinely produces more than one recorded conversion, I assume duplication until proven otherwise. I do not need a platform screenshot to tell me that repeated inflation will distort Smart Bidding. In live audits, that issue is common.

How I document fixes so someone can actually implement them

Remediation reporting should help an owner, marketer, developer, or in-house operator act on the findings without sitting through three status calls. Agency-style audit decks often fail here. They describe the problem, then leave the implementation path vague. I prefer a working document that assigns responsibility and defines the retest.

I write fixes in a priority table:

That order matters. I do not spend time tuning bids while the account is still arguing with itself about what a conversion is.

Each remediation item should answer four questions:

• What is broken
• Why it affects spend or reporting
• Who owns the fix
• How the fix will be retested

I also keep a change log with dates for GTM publishes, site releases, form updates, CRM import changes, and conversion-setting edits. A lot of accounts that look unstable are not unstable at all. Someone changed the site, swapped a form tool, added a redirect, or flipped a secondary action to primary without recording it.

If you want to see how cleaned-up measurement shows up in live account outcomes, these Google Ads and measurement case studies are a useful reference point. The common thread is not flashy bidding tactics. It is getting the tracking layer into a state where reported performance deserves to be trusted.

That concludes the audit. Once QA passes and the fixes are documented, CPA and ROAS become usable operating metrics. Before that, they are just numbers in a dashboard.