A Smart Bidding campaign behaves exactly as well as the conversion data you feed it. That sounds obvious. But in practice, the gap between what advertisers believe they are tracking and what Google Ads is actually receiving is wide enough to send a campaign in entirely the wrong direction. When performance slips, the usual response is to question the bidding strategy - raise the target CPA, switch to Maximise Conversions, request a review. The conversion setup rarely gets opened.
It should. The most damaging errors in a paid search account are not bid strategy miscalculations or poor keyword selection. They are silent tracking failures that report confidently, fire consistently, and still produce garbage data. Smart Bidding optimises towards whatever signal you give it. If that signal is wrong, the algorithm does exactly what it is designed to do - and takes your budget with it.
Why Tracking Errors Are Hard to Spot
The insidious quality of conversion setup errors is that most of them do not produce obvious symptoms. A broken tag fires zero conversions and triggers an alert. A misconfigured tag fires plenty of them - just not the right ones. Conversion volume looks healthy. Google Ads reports a cost per conversion that seems plausible. Campaigns get through automated reviews without issue. The problem only surfaces when you compare reported conversions to actual CRM entries or revenue, and the numbers do not reconcile.
This matters more now than it did three or four years ago. Smart Bidding and Performance Max are far more dependent on clean conversion input than manual or enhanced CPC campaigns ever were. The algorithm adjusts bids at auction time based on predicted conversion probability, built from your historical conversion data. Feed it duplicated conversions and it overestimates how often a click converts. Feed it the wrong conversion event and it optimises towards an action that has nothing to do with revenue. The output looks like a bidding problem. The root cause is measurement.
Duplicate Conversions: The Most Common Distortion
Duplicate conversion tracking is probably the single most widespread setup error in accounts that have been through multiple configuration changes. The typical pattern: GA4-imported conversions are running alongside a separately created Google Ads conversion action for the same event. Both are set to include in 'Conversions'. The result is that a single lead is counted twice, reported conversion volume doubles, and Smart Bidding calibrates itself against a fiction.
The fix requires checking which conversion actions are marked as primary versus secondary in Google Ads, and auditing whether any GA4 goals have been imported on top of native Google Ads tags tracking the same event. One source of truth per conversion type - either GA4-imported or a Google Ads native tag - is the principle. Running both simultaneously without one being set to 'Observation only' is a configuration error, even if it happened by accident during a platform migration.
Attribution Windows and the Micro-Conversion Problem
Attribution window settings are another category of error that tends to accumulate quietly over time. A 90-day click attribution window on a conversion action that a genuine customer typically completes within 24 hours will pull in assisted and view-through touches that look like direct conversions. Bidding inflates as a result, because the algorithm believes it is performing better than it is. Tightening the window to match your actual customer decision-making timeline is not just a reporting preference - it directly shapes what the bidding model thinks is working.
Micro-conversions in the 'Conversions' column create a different but related distortion. If a button click, page scroll, or PDF download is marked as a primary conversion action, Smart Bidding will optimise for it as though it were a lead or a sale. The campaign may generate impressive conversion volume at a low CPA while producing almost no actual business outcomes. The correct treatment for micro-conversions is to track them as secondary actions - useful for analysis, excluded from bidding signals.
Google Tag Manager Configurations That Create Silent Errors
Google Tag Manager introduces its own category of tracking failures, particularly on sites that have been through redesigns or CMS changes. Triggers built on CSS selectors or element IDs break when a developer renames a class or restructures a form. The tag continues to exist in GTM, passes validation, and appears active - it simply never fires because the element it references no longer exists in the page DOM. Conversion reporting shows zero or near-zero for that event, but without a clear 'tag inactive' warning in Google Ads.
A less obvious GTM problem is tag firing order and duplicate dataLayer pushes. If a conversion tag fires on both a form submission event and a page view trigger for the thank-you URL, you get two fires per conversion. GTM's built-in preview and debug mode will show this, but it requires someone to actually walk through the conversion journey in debug mode rather than relying on the Ads interface to flag it. Regular GTM audits - particularly after any site update - are not optional if you are running Smart Bidding campaigns at meaningful spend.
Consent Mode Gaps and What They Do to Bidding Data
Consent Mode is now a requirement under Google's EU User Consent Policy, and its interaction with conversion tracking is more complex than many accounts reflect. When a user declines analytics cookies, Consent Mode v2 allows Google to model conversions using aggregated, cookieless signals rather than dropping the data entirely. But this only works correctly if the implementation is properly configured - both the consent signals and the conversion tags need to respect the consent state.
Poorly implemented Consent Mode can create a scenario where conversion tags fire regardless of consent state, generating inflated data for non-consenting users - which then enters the Smart Bidding model. Or the opposite: a misconfigured setup that suppresses all conversion data from non-consenting sessions without enabling modelling, creating an artificial drop in reported conversions that causes the algorithm to underbid. Neither scenario is flagged as an error. Both corrupt the signals that Performance Max and Smart Bidding depend on.
Auditing Conversion Setup Before You Touch Bidding Strategy
Before making any change to a Smart Bidding strategy in a struggling campaign, a structured conversion audit should come first. That means checking the Conversions column in Google Ads for every primary action currently included in bidding, cross-referencing reported volume against CRM or backend data for the same period, and walking through GTM in debug mode for each conversion type. It also means reviewing attribution windows against your actual sales cycle, and confirming that GA4-imported goals are not doubling up with native Google Ads tags.
The audit does not need to be complex, but it does need to be systematic. Most accounts accumulate conversion tracking debt over time - tags added during tests, imported goals from an old GA3 setup, attribution windows set at platform defaults rather than chosen deliberately. Smart Bidding has no way to distinguish between a clean conversion signal and a corrupted one. It treats both with equal confidence. That responsibility sits with the account manager, not the algorithm.