5 min read
Demand Gen Platform-Comparable Conversions: What Changes and What Doesn't
Understand the attribution change before treating a lower reported CPA as better performance.
Demand Gen's platform-comparable columns offer another way to attribute conversions when comparing it with advertising outside Google. They are not an extra set of customers to add to your sales total.
The practical question is what changes when you switch columns. If reported CPA falls while spend and orders stay the same, the campaign has not suddenly become more efficient. You are looking at the same activity under a different set of counting rules.
Understand what the column changes
Google's platform-comparable reference describes a Demand Gen-only attribution view that includes view-through conversions and removes other Google campaign types from its attribution scope. It is reporting, not a bidding setting, and Google advises against using it to compare Demand Gen with Search or Performance Max.
The documented restrictions matter: the adjusted method supports biddable Google Ads website-tag conversions; GA4 and Floodlight actions show the primary-column result instead. Historical coverage begins with the feature's launch in the account. The column also counts every conversion rather than honoring the one-per-click option.
Those restrictions can explain why a number is missing or looks unchanged. Neither outcome, on its own, tells you whether the advertising worked. Check the conversion source and available dates before making a performance diagnosis.
A lower reported CPA can be a different calculation
Consider a hypothetical retailer that spends $4,800 on Demand Gen. Its usual reporting view assigns 80 purchases to the campaign, giving a CPA of $60. Another attribution view assigns 120, giving a CPA of $40.
The $20 difference is real arithmetic, not evidence that a campaign edit saved $20 on each sale. Spend did not change. The business did not necessarily receive 40 additional orders. The report assigned credit differently.
If another advertising platform also claims some of those purchases, adding its attributed total to Google's can count the same order more than once. Keep the business's order records alongside the platform reports when reconciling totals.
This is also why a small gap between columns is not proof of a particular customer journey. Several configurations can produce similar totals. An aggregate count cannot tell you that every buyer took the same path.
| Hypothetical view | Attributed purchases | Spend ÷ attributed purchases |
|---|---|---|
| Original report | 80 | $4,800 ÷ 80 = $60 |
| Alternative attribution view | 120 | $4,800 ÷ 120 = $40 |
Don't confuse reporting with view-through bidding
There is a separate view-through conversion optimization beta. When enabled, eligible view-through conversions can enter the main Conversions column and influence bidding. Google currently supports image and video on YouTube and video on Discover Feed. The setting is off by default; find Include view-through conversions under the campaign's Conversions Optimization settings.
Adding a platform-comparable column does not enable that setting. Equally, the old blanket statement that view-through conversions never affect bidding is no longer reliable. Check the campaign configuration, not just the column label.
If a conversion total jumps after a settings change, examine whether measurement changed before calling it an improvement. Document the change date and compare like-for-like definitions where possible. A presentation that silently replaces yesterday's metric with today's broader one can make flat business performance look like growth.
Keep a concise definition beside the chart: the outcome counted, attribution window, reporting dates and any relevant configuration change. The purpose is to let someone interpret the number without having to reconstruct your analysis.
Use the comparison to ask a better budget question
A useful comparison identifies a decision: whether to test a new creative direction, investigate poor lead quality, or try a bounded change in spend. A table of CPAs does not make that choice by itself.
Suppose one platform reports cheaper leads but the sales team rejects most of them. A second reports fewer, more expensive leads that become customers. Compare the cost of the outcome the business values, using the available CRM evidence, rather than choosing whichever column makes the acquisition cost smallest.
Where the data cannot support an exact comparison, say what you can learn. You may have enough evidence to test a hypothesis without enough evidence to declare a winner. Set the test up to answer the uncertain part, and decide how much the answer is worth.
Report-based attribution and incrementality measurement answer different questions. The latter investigates outcomes caused by advertising relative to a comparison group. Reassigning credit among existing interactions is not a substitute.
Questions and answers
Should platform reports add up to the order total?
Not necessarily. Different platforms can attribute the same order to their own advertising. Use transaction-level business records when reconciling unique orders, and explain what each platform's total represents.
Does a better-looking CPA justify more spend?
Not by itself. Determine whether performance improved or only the attribution view changed, then consider business quality, profitability and the uncertainty a further test would resolve.