Coupang Ads built around the Item Winner slot, not the bid

Coupang is a single-page marketplace. Every seller offering the same product sits behind one item page and the Item Winner takes the cart. Advertising is settled by that fact before a bid is set: sponsored placement on an option you do not win sends WOW members to a page that adds a competitor's offer to the basket. Before quoting a Coupang engagement we pull 30 days of winner share by option ID, and if it sits below 60 percent we say so and price a different piece of work. What decides the winner, in the order it usually matters: Delivery promise, then price including shipping, then seller performance score, then stock continuity. Rocket Growth changes the first of those materially. Moving an option into Coupang fulfillment routinely does more for winner share than a fortnight of price moves, and it is the recommendation we make most often to people who arrived asking about bidding. The ad work: Manual campaigns are the base. Smart campaigns go in only where an option already wins consistently and the category carries enough query volume for the automation to have something to learn from. Search-term output harvested weekly into keyword and negative lists. Options in Rocket Growth run in separate campaigns from options you ship yourself, because their conversion behaviour is not comparable and blending them makes the bid decision unreadable. Category rules we check at the start: KC certification for electricals and children's items, cosmetics notification for anything applied to skin, PCCC handling on direct-import routes. We verify these before launch rather than after a takedown, and we will hold a launch over a missing certificate. Measurement: The console reports click-attributed conversions. We reconcile monthly against Wing settlement data, which is the number your finance team will eventually see, and we report the gap rather than smoothing it. Where an option carries enough volume, once a quarter we hold a two-week ad pause on a matched control set to see what the page sells unsupported. Some options sell almost the same. Those come out of the plan, and that recommendation has cost us budget under management more than once. What we do not do: We do not hold your Korean business registration number and we do not act as importer of record. We do not run Korean-language customer service, review responses or CS chat. We do not write Korean creative; we specify what an option page needs and a Korean writer executes it. We do not negotiate Rocket Retail 1P terms with a Coupang category manager. That is a commercial conversation about margin and volume commitments, and you should be in the room for it rather than us. Timezone, plainly: We are in Brazil. Our overlap with Seoul is the first hours of our morning against the end of their business day. Weekly cadence, launch ramps and event preparation work fine on that. Same-hour intervention during a Korean sale day does not. If your plan depends on someone reacting inside the hour during KST business hours, you want a Seoul-based account team and we will tell you that rather than take the retainer. Who this is not for: Brands testing Korea with one option and no fulfillment decision made. Bidding is not what determines whether that test works. Also not for sellers intending to win on price against a cross-border seller in the same option. That contest is settled by landed cost, and there is no bid strategy that survives a structural cost gap. We would rather decline than spend six months proving it with your money.