ACOS Rebuilt from Unit Margin, With Bid Automation

The ACOS you should be working to is calculated from your own cost file, not adopted from a category benchmark. In the accounts we take over it lands between 14 and 22 percent, against a 30 percent target the previous team had been reporting success against. The arithmetic is net receipt per unit: shelf price, less referral fee, less FBA or vendor allowance, less the trailing 90-day return rate on that ASIN. A 9.99 euro paperback netting 1.80 euro breaks even at 18 percent on the incremental unit and loses money at 24, whichever colour the console paints the cell. What we build before touching a bid: A margin table at ASIN level, signed off by your finance contact, carrying a stated ceiling per line and a separate list of items you have decided to lose money on for rank, each with an end date. Anything without a landed cost stays paused until the number arrives. Clients find this part slow. It is also the part that turns every later decision into something arguable rather than a matter of taste. How the bids move: Weekly, through bulk sheets and the Ads API, on 28 days of trailing data, with a floor of 12 clicks before any keyword-level change. In most catalogues we inherit, placement multipliers carry more of the work than base bids, and top-of-search modifiers are set per campaign against measured conversion at that placement rather than applied across the account. We do not run intraday rules that react to a single morning's ACOS. At that resolution the data is noise, and noise moves budget for no reason. Brand terms, and the argument about them: Defensive brand spend gets tested, never assumed. We pause branded exact campaigns on a matched cohort of ASINs for two weeks, keep the rest running, and compare total ordered units rather than ad-attributed ones. In three of the last five accounts the pause cost nothing and released budget into non-brand discovery. In the other two it cost around 6 percent of units and the spend went back. Either way the answer belongs to your catalogue, not to a blog post about someone else's. Where TACOS is allowed to appear: At parent-ASIN or series level, on a rolling eight weeks, read as the share of your own rank you are currently renting. It is not a campaign metric and we will not report it as one. When TACOS falls while ordered units fall faster, that is organic decay, and the response is a retail one: price, stock, review velocity, content. Bidding harder into a page that has stopped converting is how accounts quietly consume a quarter. Not included: Listing copy, A+ modules and image production. Amazon DSP. Vendor terms and allowance negotiation. Marketplaces outside the ones named in the brief. Our fee is a fixed euro figure, so nothing we earn moves when your spend does, and no recommendation we make is worth more to us than to you. Not for you if: You cannot produce landed cost per unit. Without it there is no defensible target, and we will decline rather than invent one. Also not for you if the mandate is top-line growth at any contribution before a funding round. That is a legitimate strategy, badly served by this method, and better bought elsewhere. What you hold at the end: The margin table, the bulk files, the harvest and negative logic written out, and a bid policy short enough that your own team can run it on the Monday after we stop. If the work only functions while we are in the console, we have sold you a dependency rather than a decision.