How to Audit Google Ads Conversion Tracking You Didn't Set Up

Inherited an account? Before you touch bids, prove the conversion numbers are real by tracing one action from click to report.

October 15, 2026 · QueryCut

To audit conversion tracking you didn't set up, pick one conversion action, fire it yourself, and confirm it shows up exactly once with the right value in Tools > Conversions. Then check the three settings that silently inflate or deflate your numbers: counting mode (Every vs One), attribution window, and whether a single user action is being recorded by two overlapping tags. Everything else in the account is downstream of whether these numbers are true, so this is the first thing you do, not the last.

Start with one action, end to end, before you trust anything

Do not start by reading the conversions column. Start by generating a conversion yourself. Open the site in an incognito window, click through from an ad or a test URL, complete the action (submit the lead form, hit the thank-you page, trigger the call), and then watch what lands. This is the only way to know what the previous owner actually wired up versus what they intended to.

Go to Tools > Conversions and open the specific action. Check the 'Status' column first. 'Recording conversions' is good. 'No recent conversions,' 'Inactive,' or 'Tag inactive' on an action that is supposed to be your primary goal is the single most common inherited failure: the campaign is optimizing toward a signal that stopped arriving months ago. An action with zero recent data is worse than no action, because Smart Bidding keeps bidding as if the absent conversions were just bad luck.

The reason you fire it live rather than trusting the UI is that Google's conversion pipeline has lag and dedup logic you cannot see from the settings page. A tag can be present in the page source, pass the Google Tag Assistant check, and still not reach the account because of a consent-mode block, a GA4 import that broke, or a container that was republished without the tag. Watching your own test click resolve is the only end-to-end proof.

The three settings that quietly change the number

Once you know an action fires, audit how it counts. Three fields in the conversion action do most of the damage in inherited accounts, and all three are easy to set wrong without the number ever looking obviously broken.

Counting mode ('Every' vs 'One') is the big one. 'Every' counts every conversion after a click, which is correct for ecommerce purchases where a user can legitimately buy twice. 'One' counts a single conversion per click, which is correct for leads where the same person submitting a form three times is not three customers. An inherited lead-gen account set to 'Every' will show inflated conversions and a flattering CPA, and your Target CPA bidding will chase that phantom volume. A store set to 'One' undercounts revenue and starves the bidding of real signal.

Attribution and the conversion window matter next. Check the click-through conversion window (default 30 days, can be set 1 to 90) and the attribution model. A long window on a short sales cycle just delays and smears the data; a model change the previous owner made can make two periods non-comparable even though the underlying performance is identical. Finally, confirm the value: is a static value assigned, is it dynamic from the page, or is it zero? A primary action reporting value of zero will break any Target ROAS or Maximize Conversion Value strategy entirely, because the optimizer has no value to maximize.

Find the double-count: the trap that makes everything look fine

The failure mode that survives a casual audit is the duplicate. One user action gets recorded by two mechanisms at once: the native Google Ads tag plus a GA4 conversion imported into Ads, or a Google Tag Manager trigger that fires on both page-load and a button-click event. The conversions column looks healthy, CPA looks reasonable, nobody questions it, and the whole account is optimizing on numbers that are roughly double reality.

To catch it, look at your conversion actions list for two entries that clearly track the same event (for example 'Lead Form' as a website action and 'GA4 Lead' as an imported action) that are both marked as primary. If both feed the 'Conversions' column, you are double-counting. The fix is to keep one as the source of truth and set the other to 'Secondary,' which keeps it visible in reporting but removes it from the optimization column and the headline total.

The reason this is the dangerous one: inflated-but-stable numbers do not trip any alarm. A broken tag shows a cliff in the data and someone notices. A clean 2x overcount shows a smooth, believable line, and it will quietly justify spending more on worse traffic for as long as it runs. Trace your single test action and count how many times it appears across all actions. If it shows up twice, you found it.

Decision rules for what to fix now versus leave alone

Not every imperfection is worth touching on day one, and changing conversion settings resets the data history that bidding depends on. Here is how to triage.

Fix immediately, regardless of consequences: any primary action with status 'Inactive' or 'No recent conversions,' any confirmed double-count feeding the primary column, and any value-based strategy where the primary action reports a value of zero. These are not optimizations, they are the account lying to itself, and leaving them costs money every day.

Change with caution and document the date: counting mode and attribution window. Switching 'Every' to 'One' on a live lead-gen account will make conversions drop the next day, and if you did not write down why, you or the client will panic and revert it. Make the change, annotate it, and warn stakeholders that the number is about to fall and that the new number is the real one.

Leave alone unless you have a reason: enhanced conversions setup, offline conversion imports, and historical attribution model choices on campaigns that are performing. If it is counting once, firing reliably, and carrying a sane value, resist the urge to re-architect. Every reset of the conversion configuration costs the bidding algorithm its learning, and a slightly imperfect signal that is stable beats a perfect signal the optimizer has never seen.

What you still cannot know, and how to hedge it

An audit tells you whether the tracking is internally consistent. It does not tell you whether the conversions are true in the sense that matters to the business. A form fill that fires perfectly can still be a bot, a job applicant, or a wrong-number inquiry, and Google Ads will count all of them identically. The only way to close that gap is to reconcile the Ads conversion count against the CRM or back-end system for the same date range. If Ads says 120 leads and the CRM shows 70 that were real humans, your tracking is fine and your definition of 'conversion' is the problem.

You also cannot fully see consent-mode behavior from inside the Ads UI. If the site uses a consent banner, some conversions are being modeled rather than observed, and the proportion is not exposed to you at the action level. Treat heavily-modeled accounts with extra caution and lean harder on the CRM reconciliation, because the raw count is partly an estimate.

When the picture is genuinely ambiguous, say so out loud to whoever owns the account. 'The tracking fires correctly but I cannot confirm these leads are qualified without CRM access' is a precise, honest statement. It is far more useful than a confident green checkmark on a number you have not actually tied back to revenue.

Questions people also ask