如何超越转化量指标评估 Google Ads Performance Max 效果

A rising conversion count is the first number most advertisers check when reviewing Google Ads Performance Max, and it is also the one most likely to mislead. Volume shows that the system found people who completed an action. It does not show whether those people were profitable, whether the action mattered to the business, or whether the order would have happened without the ad.

Cross-border shopping, which occurs through platforms such as Ubuy, shows how much the conversion differences in the reports oppose the actual metrics because there can be much discrepancy in shipping, cost margins, money rates, and refunds from different markets.(Disclosure: I am a Product Manager at Ubuy, an ecommerce company that delivers worldwide.)

In this article, we analyze how to evaluate the effectiveness of Performance Max campaigns according to the quality of conversions, profitability, data regarding orders, and by conducting incrementality tests and how to determine when to continue, pause, or revise campaign processes.

Why high conversion volume can mislead

Performance Max optimizes toward whatever you tell it to count. When the signal is weak, the system finds more of that weak signal very efficiently.

Four patterns cause most of the trouble:

• Soft primary conversions. Page views, add-to-carts, and checkout starts which are counted as primary actions inflate volume and train bidding on low-intent behavior.
• Duplicate or static values. Double-firing tags and fixed conversion values raise reported return on ad spend (ROAS) without real revenue behind the number.
• Easy demand taking the credit. Branded searches and returning customers usually convert at lower cost, so the system leans on that inventory when a campaign has a return target and no guidance on brand.
• Volume without margin. Orders that carry heavy shipping, duty, or return costs count the same as high-margin orders unless you tell the system otherwise.

A hypothetical ecommerce example: ROAS versus profit

Take, for example, an ecommerce website similar to Ubuy that has a ROAS of 6.0 with an investment of $20,000 every month. This effectively results in a $120,000 conversion value. Keep in mind these numbers are only for example purposes and are not from any live account.

Suppose the search terms report shows that brand queries produced 60% of that value ($72,000) on roughly $4,000 of spend. Removing brand leaves $48,000 of value on $16,000 of spend, a ROAS of 3.0.

With a gross margin of 35%, the break-even return on ad spend is estimated at around 2.9 and can be computed from the formula 1 divided by 0.35. The non-brand part does not make much profit covering its cost; at the same time, the headline figure indicates more than adequate return. The budget should take into account the figure of 3.0 as opposed to 6.0.

Metrics that matter more than the conversion count

A useful review looks at several layers instead of a single number:

• Conversion value and ROAS. It is crucial that these figures are accurate and not duplicated. ROAS should be compared to your break-even point instead of compared to any arbitrary goal.
• Cost per acquisition (CPA) versus cost per profitable order. CPA is a platform metric, while profit per order after fulfilment, shipping, and returns is a business metric.
• Customer lifetime value. A higher CPA can be acceptable for customers who reorder, and a low CPA on one-time, low-margin buyers can still lose money.
• New-customer share. Revenue from first-time buyers is a better indicator of growth than total revenue, which may include customers you already had.
• Qualified leads, where applicable. For businesses that sell through quotes or sales calls, count leads that reach a real sales conversation rather than raw form submissions.
• Profitability. The measures of margins, returns, and fulfillment expenses allow one to determine whether a seemingly good ROAS is, as a matter of fact, profitable or not.

Audit conversion tracking and attribution first

Before judging results, confirm the underlying data is trustworthy:

1. Review primary and secondary conversion actions. Bidding should only be based on actions that lead to real business value. Softer signals, like add-to-cart should become secondary signals to be reported without optimization.
2. Check counting methods and values. Duplicate purchase tags and static values often inflate revenue figures, so compare Google Ads transaction counts with your platform’s order records.
3. Understand attribution limits. Conversions reported by the platform are influenced by modelling and count Google’s own interactions, meaning they cannot measure their impact independently. Importantly, delayed conversions and imported conversions can lead to different reporting results afterwards.
4. Strengthen the signal. Google’s Performance Max guidance stresses that AI is only as good as the data it receives and recommends clear conversion categories and enhanced conversions.

Compare Google Ads data with order and CRM records

Your order management system and customer records are the closest thing to ground truth available.

A sound reconciliation process includes:

• Associate transactions originating from Google Ads with valid orders, since cancellations and returns change the scenario after clicks.
• Comparing reported conversion value to net revenue after refunds.
• Using CRM or customer data to separate new buyers from repeat buyers by campaign.
• Reconciling platform revenue with finance-reported revenue on a regular basis.
• Monitor the gap between what is reported and what is verified. A growing gap is a red flag that optimisation efforts have drifted.

Up-to-date reporting is advantageous as well. All Performance Max campaigns receive channel performance reporting from Google that shows where the ads appear. Analyse this alongside conversion quality, because cost-effective inventory can yield more, but less effective, interactions.

Use incrementality testing to see what Performance Max adds

Attribution decides who receives credit. Incrementality testing asks whether the outcome would have happened anyway, which makes it the strongest test of Performance Max campaign performance.

Common approaches include:

• Google Ads experiments. You can test Performance Max against existing campaign types, and uplift-style experiments measure what the campaign adds alongside other types, as this overview of 2026 updates describes.
• Geo holdouts. Reduce or pause spend in matched regions, then compare total business results rather than platform conversions. This suits multi-country retailers well, as long as the regions are comparable.
• Brand exclusion tests. Remove brand queries from the campaign and track total revenue and orders. If revenue holds while spend falls, that spend was not incremental.

Interpret results in terms of incremental orders, incremental cost, and incremental ROAS. Agree on the test window in advance and run it long enough that normal fluctuation, including seasonality and sale periods, does not distort the result.

When to scale, hold, or adjust

A simple framework prevents reactions to week-to-week swings.

Scale when:

• Verified margin holds as spend increases
• Incrementality tests show genuine lift
• Non-brand ROAS clears break-even with room to spare

Hold when:

• Results look good but data volume is low or recent changes are still settling
• Tracking appears accurate but incrementality has not been tested

Adjust when:

• Brand makes up a large share of conversion value and blended ROAS hides weak non-brand return
• Specific markets or categories fall below break-even despite a healthy blended number
• Reported conversions and order records are diverging
• Tracking errors or duplicate counting are present

Bottom Line

Conversion volume forms the basis for analysis rather than serving as the outcome of the analysis. The use of best optimisation techniques to achieve success with Performance Max starts by clarifying what conversion actually is and matching it with order and finance data, as well as testing for incrementality before allocating any large budget. 

What might look like a good-performing campaign in the dashboard may have actually low performance in reality, while there can be another campaign that looks unpromising in terms of dashboard performance but is actually capable of providing some of the best customers.