Restaurant owners on Coupang Eats, one of Korea's largest delivery platforms, say repeat customers were being counted as new customers — leaving them paying an additional advertising fee amounting to 5–10% of the food price.
The dispute only makes sense against the rate structure:
The same order costs twice as much when classified as new. So the classification rule effectively sets the advertising bill.
Some owners reported finding cases where customers who had ordered between 7 and 20 times were being flagged as new. A customer on their twentieth order being labelled new is what pushed owners to question the classification itself.
Owners raised a second concern: following a data breach last year, many customers may not have consented to personal data usage.
If order-history-based classification depends on that consent, customers who declined would appear to have no history — and could therefore be counted as new every time. It is an unverified inference, but it illustrates precisely why classification logic needs to be transparent.
Coupang Eats attributed the issue to incorrect guidance from its call centre, stating the actual criterion is "a customer who has never ordered, or has no order history in the last 60 days."
Under that rule, a customer with twenty prior orders would legitimately be classified as new if the most recent one fell outside 60 days. The criterion is defensible on its own terms.
Owners were not satisfied. They continue to demand transparent disclosure of individual customer order dates and the history of policy changes, pressing for verification.
The crux is not whether the rule is reasonable but whether it can be verified. A 60-day threshold may be perfectly sound, yet without any way for a merchant to confirm how it was applied, trust does not follow.
In performance advertising, the definition is the price. "New customer" looks self-evident, but a single number — 60 days — doubles the fee. That is why checking definitional clauses before running platform advertising matters more than it appears.
Demand a change log for classification rules. The owners' request for policy change history is exactly right. If criteria shift without notice, period-over-period cost comparisons become meaningless.
Keep your own records. Where platform reporting cannot be independently trusted, retaining your own order data provides the basis for cross-verification. That principle extends well beyond delivery apps.
Treat platform policy as operational risk. A structure where a platform's rule change directly changes your costs is not unique to delivery. Platform policies typically take effect quickly after notice.
Structures where merchants cannot verify a platform's own criteria recur elsewhere. On what a verification badge actually checks, see The Gap in Naver's 'I Paid For It Myself' Badge. For the competitive picture among Korean commerce platforms, see Naver Closes the Gap on Coupang in Payment Volume.
Restaurant owners say repeat customers were counted as new customers, adding advertising fees worth 5–10% of the food price. They reported cases of customers with between 7 and 20 prior orders being flagged as new.
New-customer advertising carries a minimum commission rate of 10%, while repeat-order and all-customer advertising is 5% — meaning classification as new doubles the fee on the same order.
It attributed the issue to incorrect guidance from its call centre and stated the actual criterion is a customer who has never ordered or has no order history in the last 60 days.
Transparent disclosure of individual customer order dates and the history of policy changes, so the classification can be independently verified. Verifiability, not the rule itself, is the core issue.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
Meta has begun alerting advertisers that the Placements option is being removed from ad sets, along with platform exclusions. Another notch of control disappears into automation.
NBT signed an ad server deal with Mobwith to expand into display advertising. What advertisers should check when an 11M MAU reward network starts selling standard display inventory.
Naver display ads reach reporting launched August 12. Here is what it measures, why logged-in-only counting and 10,000-unit buckets matter, and which campaigns are excluded.