A ChatGPT campaign can report a lower cost per purchase while the business pays more to acquire each new customer. That does not automatically indicate broken reporting. The ratios may count different people, costs and periods. Before moving budget, establish what question each number can answer.
In this guide, CPA means media spending divided by reported purchase conversions. Blended CAC means the business’s defined total acquisition cost divided by unique new customers in the reporting period. It is an internal operating ratio. It describes the whole acquisition system without assigning credit between advertisements on ChatGPT and other contacts with the business.
Start with the customer definition
Decide what qualifies someone as a new customer. For a shop, an internal definition might be the first valid paid order in the customer record. Document how test orders, cancellations, guest checkouts and merged accounts are handled. Where the business can reliably and appropriately establish identity, a different email address should not make an existing customer new again.
Keep customers with an unknown marketing source in the business total. Removing them shrinks the denominator precisely when channel tracking is weak, making blended CAC partly a measure of tracking coverage. Report the unknown share separately. Likewise, do not remove someone simply because a later order is returned; apply the customer qualification rule consistently.
OpenAI describes conversions as configured events connected to campaigns in Conversion Tracking. A purchase event and a unique first-time customer are different analytical units. Assess customer history in the business’s own records rather than assuming that a conversion total supplies it.
Set the spending boundary once
A practical starting point is all expenditure consistently classified as acquisition: media across channels, relevant production, agency work and any included internal time. If the calculation contains media alone, label it blended media CAC. A broader measure should state which additional costs it includes.
Decide how shared production and activity serving existing customers will be treated before comparing periods. Some costs may need a documented allocation. Do not change that allocation because a different rule makes the ChatGPT campaign appear stronger. The guide to fixed and variable campaign costs helps preserve the distinction between launch work and recurring operations.
Use one reporting currency and a consistent treatment of recoverable taxes. Product costs normally belong in customer contribution rather than CAC. Including them in both places would charge them twice when contribution is compared with acquisition spending. Repeat orders must not increase the new-customer denominator.
Read three different ratios from one example
Consider a hypothetical month denominated in SEK. ChatGPT media costs 12,000 and the selected report shows 80 attributed purchases. Reported CPA is SEK 150. That alone does not reveal how many purchases came from existing customers or whether one customer placed several orders.
Separately, the shop’s documented internal channel rule classifies 45 unique new customers to ChatGPT. Dividing media spending by that group produces approximately SEK 267. This is another ratio, using a different population, rather than a correction to platform CPA. Changing event records or customer histories to force agreement would conceal the distinction.
Across the business, total acquisition cost is SEK 80,000 and there are 320 unique new customers. Blended CAC is SEK 250. In this example, the cost comprises 12,000 of ChatGPT media, 54,000 of other media and 14,000 of production and work. No individual channel can claim all 320 customers as its proven acquisitions.
The three figures can sit beside one another, provided each carries its numerator, denominator and counting rule. A short table with those definitions is more useful than a dashboard that displays three similar-looking “customer cost” cards with no explanation.
Three ratios, different denominators
- Reported CPA
SEK 12,000 ChatGPT media / 80 attributed purchases = SEK 150 per purchase.
- Internal channel view
12,000 / 45 new customers under the shop’s channel rule ≈ SEK 267. Different population and rule.
- Blended CAC
SEK 80,000 total acquisition cost / 320 unique new customers = SEK 250 per customer.
Why the previous month is not a control group
Suppose a comparable earlier month had SEK 60,000 of media, 12,000 of production and work, and 300 new customers. Blended CAC was SEK 240. In the following month, customer count rose to 320 while spending increased from 72,000 to 80,000. The ratio therefore rose to SEK 250.
The spending difference of 8,000 divided by 20 additional customers equals SEK 400. That is a descriptive difference ratio, not a proven incremental CAC for ChatGPT. Seasonality, offers, changes in demand and other advertising could affect both months. A causal conclusion requires a design that distinguishes those explanations.
Timing also matters. This month’s advertising can contribute to purchases next month. Keep the fixed calendar-period ratio, but supplement it with equally aged customer groups where the records permit that analysis. A longer observation period for one group should not become an invisible advantage in the comparison.
Match the investigation to the disagreement
If CPA falls while blended CAC rises, inspect the share of returning buyers, spending outside media and changes to customer qualification. If blended CAC falls while ChatGPT CPA rises, other channels or organic demand may have improved. Neither pattern alone determines which budget should increase.
OpenAI’s Reporting distinguishes reporting time basis from attribution windows. Preserve those settings when tracking CPA over time. For the internal CAC row, document the customer’s qualifying acquisition date and the cost period instead. Those clocks may differ, but the difference must be visible to the reader.
Complete the review with customer economics. A lower CAC can accompany customers who buy a discounted item once and never return. Compare acquisition cost with repeat-purchase contribution for mature, comparable groups. The resulting decision should state what evidence is missing and what investigation comes next, rather than naming a winning channel from one blended ratio.
Sources and scope
Interpret reported channel cost per conversion alongside total cost per unique new customer and choose an appropriate investigation before changing budgets.
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.