Margin-Led ACoS and TACoS Control With Hourly Bidding

A target ACoS should fall out of your contribution margin, not arrive from a planning meeting. On a pet consumable holding 34 percent margin after FBA fees and returns, where 40 percent of first orders convert to 定期おトク便, break-even on the first order and break-even on the customer are two numbers about a factor of two apart. One account-wide target is wrong for both. The first fortnight of this engagement is spent building the margin table per ASIN with your finance team. Bidding changes after that, not before. Hourly instead of daily: Amazon Marketing Stream pushes Sponsored Products, Brands and Display metrics hourly through AWS Firehose, and that is the data the dayparting is built on. Japanese demand has a commute window and a late-evening window that do not convert at the same rate, and a daily average hides the difference entirely. Bids move at most once an hour and by at most 20 percent in a step, because rule engines that move faster oscillate and then get blamed on the market. Placement before keyword: The top-of-search modifier moves more money in a week than a month of keyword tinkering. We set it per ASIN against the margin table rather than per campaign, and we treat it as a change with a fixed read window instead of a setting somebody nudges. What TACoS is for: TACoS is a diagnostic, not a target. Read it against organic share from Search Query Performance. Rising TACoS with rising organic share means you are buying position you will keep. Rising TACoS with flat organic share means you are renting, and we will recommend cutting spend, which is the recommendation a percentage-of-spend fee makes expensive to give. That is why we do not charge one. The branded-terms test: Every account we take over is defending its own brand terms, and most cannot say what that spend earns. We pause branded defence on half the ASINs for 14 days and count total branded orders across both halves. We have shut down branded lines running into six figures a year, and we have also confirmed that a client facing an aggressive competitor genuinely needed the defence. Both answers are worth what the test costs. Rakuten and Yahoo!: RPP gives you item-level CPC and very little of the negation surface Amazon has, so the lever is the catalogue and the item bid, not the keyword. Rakuten also runs お買い物マラソン or スーパーSALE in nearly every month, which resets the baseline twelve times a year: we compare event to comparable event and never week over week, and we plan point spend against inventory rather than chasing the ranking board. Yahoo! アイテムマッチ is priced and managed separately and gets its own budget ceiling. What is not included: Detail page and A+ work, which is a separate engagement and often should come first. Amazon DSP buying. AMC query work. Creative production. Reselling somebody else's black-box bidder under our name. Who this is not for: Accounts under roughly 20,000 dollars a month in retail media, where hourly segments are too thin to act on and a weekly hand pass beats anything automated. Brands that will not share COGS, fulfilment fees and return rates, because everything here is derived from them. Anyone looking for a promised ACoS number in the proposal; that number depends on your margin and your pages, and we quote a method rather than an outcome. Handover: The rules live in your own Ads API account and are documented in plain language, with the reasoning behind each threshold. End the engagement and you keep a working system rather than a dependency.