Real Client Google Ads Audit Part 3: The 5 Issues I Found in a Live Ecommerce Account

This is Part 3 of an ongoing series where I open a real client account on camera and audit it live. No scripted scenarios, no sanitized examples. Just what I actually find when I look at a real ecommerce Google Ads account that has been running for over a year.

If you missed the earlier parts, start with Part 1 here for context on how I structure an account audit from the top down. The checklist format from Part 2 also applies to everything I cover here.

In this video I found five issues in a Shopify ecommerce account. Some are common. Some are subtle. All of them were costing this client money in ways they could not see without digging into the right places.

Before getting into the issues, here is the setup. Ecommerce brand on Shopify. Google Ads account with three active campaigns: a non-brand Search campaign, a Performance Max campaign, and a brand Search campaign. Monthly spend sitting between $2,000 and $3,000. The account had been running for over a year, managed previously by a different agency.

Performance had been declining for four consecutive months. The client reported dropping revenue from Google Ads despite roughly stable spend. When I looked at the account at a high level, it appeared to be running. Campaigns enabled. Ads approved (mostly). Budget not capping. On the surface, nothing screamed broken.

That is exactly the scenario where an audit matters most. The damage happens quietly, in settings nobody checks after the initial setup.

The first thing I always do before looking at performance data is verify conversion tracking. There is no point analyzing campaign metrics if the conversion signal feeding Smart Bidding is corrupted. That check uncovered Issue 1 immediately.

The conversion setup was tracking Begin Checkout as the primary conversion action. Not purchases. Begin Checkout fires when a user reaches the checkout page. Purchase fires only when payment completes. These are not interchangeable signals.

When Smart Bidding optimizes toward Begin Checkout, it targets users who are likely to initiate a checkout. That sounds reasonable until you realize checkout initiation and purchase completion are very different behaviors. Someone who adds to cart and starts checkout but abandons is counted as a conversion. Someone who actually pays and buys is also counted as a conversion. From the algorithm's perspective they look identical.

The account was essentially paying Google to find more cart abandoners, not more buyers. The volume looked fine in the conversion column. The actual revenue from Google Ads was detached from what was being reported.

The fix for this account was straightforward in concept: demote Begin Checkout to a secondary conversion action (not included in the conversion column), promote Purchase to primary, and wait. The difficult part is the waiting. Smart Bidding needs two to three weeks of clean purchase data before making reliable bidding decisions. During that window, resist the urge to cut budgets or restructure campaigns based on short-term fluctuations.

One important note: do not delete the old Begin Checkout conversion action. Demote it. Deleting conversion actions wipes the historical data, which affects Quality Score signals and audience building. Just remove it from the primary column and mark it as secondary.

Two ads were disapproved. Both had been in that state for more than six weeks. The client had received the notification emails. The previous agency had received them too. Nobody acted.

This is more common than it should be. When an account has multiple active ad variations, a disapproved ad simply stops serving while the other variants continue. The campaign does not pause. Budget keeps spending. Impression share takes a quiet hit that does not show up as a dramatic alert anywhere in the interface.

In this case, both disapprovals were in the Search campaign. One ad triggered Google's sensitive categories policy due to specific phrasing in the headline that Google flagged as potentially misleading, even though the actual product had nothing to do with the flagged category. This is a known issue with broad automated policy enforcement. The fix is to rephrase, not to appeal. Appeals take weeks and often fail even on legitimate cases.

The second disapproval was a capitalization violation. A headline used ALL CAPS formatting, which Google's policy prohibits. Fixed in under two minutes.

Both ads were back in rotation within 24 hours of resubmission. Small issue in isolation. But the six weeks of reduced impression share in an already-declining account compounded the overall performance drop the client was experiencing.

The Performance Max campaign was capturing branded queries. I confirmed this by pulling the search terms report from the brand Search campaign and cross-referencing with the asset group and product data coming through PMAX. The overlap was significant.

Here is why this matters. When PMAX and brand Search compete for the same branded queries, Google's routing logic often sends those impressions to PMAX. The result is that PMAX gets credit for converting users who searched the brand name directly. These are the easiest, cheapest, highest-intent users in the entire funnel. They were going to buy anyway. Attributing their conversions to PMAX makes the PMAX campaign look more efficient than it actually is.

Simultaneously, the brand Search campaign sees fewer impressions and fewer conversions. Its performance data starts to look worse. If you make budget or structure decisions based on this distorted picture, you end up starving the campaign that should be running cheaply while rewarding a campaign for work it did not actually do.

The fix: navigate to the PMAX campaign settings, open the brand exclusions section, and add your branded terms. This forces PMAX to compete exclusively on non-brand queries. After applying exclusions in this account, expect PMAX ROAS to decrease. That decrease is not a performance problem. It is data accuracy. You are now reading what PMAX is actually doing for acquisition rather than what it is taking credit for.

If you want to understand what PMAX is actually spending budget on across Search, Shopping, Display, and YouTube, the PMAX channel performance breakdown shows how to pull that data and what to do with it.

On the surface this looks like an easy decision: kill Search, scale PMAX. The $158 CPA gap is massive. But that conclusion would be wrong, and acting on it would have made this account worse.

The $47 PMAX CPA was partly driven by the branded traffic captured before we applied exclusions (Issue 3). Once brand exclusions were in place, PMAX's CPA moved up materially. Not to $205, but the gap narrowed considerably. The comparison had been distorted from the start.

The Search campaign had a structural problem separate from bidding. It was running all non-brand queries through a single ad group: one set of ads, one bid strategy, covering everything from high-commercial-intent product queries to broad informational queries that rarely convert. High-intent keywords were sharing budget and bidding headroom with informational queries that were attracting clicks but generating almost no purchases.

When you mix these query types into a single ad group, the bidding algorithm averages performance across all of them. The CPA looks high because it includes all the low-converting informational clicks alongside the high-converting purchase intent clicks. The fix is segmentation: separate ad groups for commercial intent, research intent, and informational intent, each with appropriate bids and budget allocation.

After restructuring, the Search campaign CPA came down significantly. Not to PMAX levels, but to a range where it was contributing profitable new customer acquisition. Before this audit the client was one reporting period away from shutting down a campaign that just needed better structure.

Before cutting any campaign based on CPA alone, check ecommerce benchmarks by category and verify that the comparison point (in this case PMAX) is using clean, attribution-accurate data.

Enhanced conversions were not enabled. This is one of the most consistently under-implemented features in ecommerce Google Ads accounts, and it has a measurable impact on reported conversion volume.

Here is how it works. When a user converts on your site, standard conversion tracking fires a tag and reads a cookie to identify who that Google user is. Enhanced conversions add a second layer: at the moment of conversion, hashed first-party data (email address, phone number) is sent alongside the conversion event. Google uses this hashed data to match conversions back to Google accounts even when the cookie was blocked, expired, or unavailable due to cross-device behavior.

In practice this means accounts without enhanced conversions are systematically undercounting conversions in a post-cookie environment. The actual sales happen. Google just cannot attribute them. Your reported CPA looks worse than reality, which can trigger bidding adjustments that reduce spend on campaigns that are actually performing well.

For this account, enabling enhanced conversions through GTM using the purchase event brought in roughly 12% additional attributed conversions within two weeks. The actual purchase volume had not changed. The reporting just became more accurate. That 12% shift was enough to move the Smart Bidding algorithm's baseline assumption and improve bid efficiency across both campaigns.

This is permanent infrastructure. Set it up once and it runs. There is no good reason to be without it on any account that has been live for more than a few weeks.

Five issues. The temptation is to fix them all at once. That is a mistake. Simultaneous changes make it impossible to diagnose what worked when performance shifts. Here is the order I used and the reasoning behind it.