“They will come back” is an easy way to justify an expensive first purchase. For advertising on ChatGPT, that claim needs a usable evidence base: which customers return, when they return and how much contribution remains after the associated costs. Only then should repeat purchases influence the amount the business is prepared to spend on acquisition.
Begin with a bounded measure, such as observed customer contribution over 180 days. Do not describe it as complete lifetime value when later years have not been observed. This is an internal economic calculation, not a claim that ChatGPT supplies a finished LTV report or that a new campaign will reproduce historical customer behaviour.
Establish a cohort that can be followed
Group customers by the date of their first qualifying purchase. Freeze membership, then follow their orders, refunds and costs. Use a consistent new-customer definition, ideally the same definition used in blended CAC. Someone should not move into another acquisition cohort because a later purchase arrives through a different channel.
Describing a cohort as acquired through ChatGPT requires an explicit internal channel rule and a statement of its limitations. A campaign label in an analytics tool is not proof that the advertisement caused acquisition. Customers with an unknown source should remain a visible group instead of being assigned to whichever channel needs more customers.
OpenAI’s Conversion Tracking describes purchases as events. Assemble the history across purchases in the business’s own records. A reported purchase amount does not establish either the cost of supplying the order or the customer’s later contribution.
Build value from contribution, not revenue
For each order, start with realized revenue after discounts and refunds, using the business’s chosen tax basis. Deduct product costs and relevant variable costs for payment processing, delivery, fulfilment and service. Treat recovered inventory and return expenses consistently. Do not use a fully return-adjusted margin and then deduct another allowance for those same returns.
Keep initial acquisition expenditure outside this contribution calculation because the resulting value will be compared with CAC. Costs required to obtain repeat purchases, such as a specific reactivation campaign, should reduce repeat-purchase value unless already deducted elsewhere. Every cost needs one identifiable place in the model.
Show shared fixed costs separately. Contribution after order expenses and retention can fund acquisition and fixed costs; it is not automatically net business profit. The guide to return-rate sensitivity helps when refunds are still developing and the final order economics are uncertain.
Divide by the whole original cohort
Consider a hypothetical cohort of 100 new customers, all observed for 180 days. Amounts below are in SEK excluding recoverable VAT. First purchases produce total contribution of 22,000 after order costs and return adjustments, or SEK 220 per original customer.
During follow-up there are 40 second orders, each contributing SEK 180, for 7,200. There are also 10 later orders contributing SEK 150 each, for another 1,500. Specific retention activity for the cohort costs 1,700. Total observed contribution is 22,000 plus 7,200 plus 1,500 minus 1,700, giving SEK 29,000.
Divide that amount by the original 100 customers. The result is SEK 290 per customer over 180 days. Do not divide repeat contribution only by the people who returned and apply that inflated figure to all newly acquired customers. Customers who never buy again are part of the cohort’s economics, not missing observations.
At a hypothetical CAC of SEK 260, the cohort leaves SEK 30 per original customer before shared fixed costs. The first purchase alone falls SEK 40 short of CAC. This explains why both short and longer horizons matter, but does not yet establish when the cash becomes available.
Build customer value from contribution
- First purchases
100 × SEK 220 = SEK 22,000 order contribution after order costs and return adjustments.
- Repeat purchases
40 × 180 + 10 × 150 = SEK 8,700. Count orders rather than assumed revenue.
- Retention and denominator
(22,000 + 8,700 − 1,700) / 100 = SEK 290 per original customer.
- After acquisition
290 − SEK 260 CAC = SEK 30 left per customer before shared fixed costs.
Test dependence on the repeat orders
Create a separate conservative scenario for the next acquisition cohort. Suppose there are 25 second orders instead of 40 and five later orders instead of ten, with the same per-order contribution and retention expense. Total contribution becomes SEK 25,550, or 255.50 per original customer. That is SEK 4.50 below the assumed CAC of 260.
The scenario is not a forecast. It exposes a decision that reverses when repeat purchasing weakens. Decide how much uncertain future contribution the budget may depend on. A new campaign might initially use a first-purchase spending boundary, with a longer horizon considered only after relevant customer history develops.
Choose comparison cohorts with similar first products, discounts, seasons and opportunities to reorder. A customer purchasing a replenishable item has a different natural cycle from someone buying a durable product once. Averaging those groups together can hide a shift in product mix created by the ChatGPT campaign’s offer.
Also inspect concentration. A small number of unusually valuable customers can account for much of the cohort’s later contribution. Show their influence as a sensitivity analysis without deleting them from the actual result. The planning question is whether the next cohort can reasonably depend on that concentration recurring.
Lock the observation period before presenting results
Do not compare a cohort aged 30 days with one that has had 180 days to purchase. Show cohort age, original customer count and the data cut-off beside contribution. Mark younger groups as immature; filling their unobserved months with zero creates a misleading conclusion about customer quality.
Preserve the offer and campaign IDs associated with the cohort. Campaign Management describes IDs as durable integration identifiers. They also help internal histories survive campaign renaming.
Once the cohort has matured, update the acquisition assumption using observed contribution and an explicit allowance for uncertainty. Then check payback timing. A customer value that appears sufficient after six months can still require more funding than the campaign can support during its first weeks.
Sources and scope
Decide whether observed repeat-purchase contribution supports using a longer customer-value horizon when planning acquisition through advertisements on ChatGPT.
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.