Amazon bid optimization: a practical guide

ACoS is ad spend divided by ad sales. On its own it says nothing about profit. The number that matters is the ACoS at which an ad sale stops making money: your break-even ACoS.

Break-even ACoS = (price - cost of goods - Amazon fees) ÷ price

Use your own fees, from your settlement reports or Amazon's fee preview: referral and fulfilment fees vary by category and size, so a borrowed number gives you a borrowed target. The break-even ACoS calculator does the arithmetic, and the glossary covers contribution margin and cost of goods.

Your target is then a decision. Below break-even, every ad sale keeps some profit. Above it, you are paying for growth on purpose, for example to build rank during a launch. Both are legitimate; what is not is a target copied from someone else's account.

One more view keeps the target honest. Ad-attributed ACoS ignores organic sales, so watch TACoS too: ad spend divided by total sales, which shows what advertising costs the whole business. The TACoS calculator works it out.

A target ACoS becomes a bid through revenue per click (RPC): what a click on that keyword earns on average.

Revenue per click = ad sales ÷ clicks = conversion rate × average order value
Target CPC        = revenue per click × target ACoS

Price from what you actually paid per click (CPC), not from the number in the bid field. The bid is a setting; the cost per click is a measurement. Because ACoS = CPC ÷ RPC, when revenue per click holds steady you can scale the click price directly:

New CPC = current CPC × (target ACoS ÷ current ACoS)

Keywords with clicks and no sales

With no sales there is no revenue per click to multiply. Rather than cutting the bid at an arbitrary click count, let it fall smoothly as clicks pile up without an order: assume the next click could convert, and price for the clicks already spent.

Target CPC = target ACoS × average order value ÷ (clicks so far + expected clicks per order)

The max CPC calculator runs these numbers for one keyword, and the glossary explains conversion rate and average order value.

A keyword bid is spread across every search term it matches. Broad and phrase keywords, and automatic campaigns, can match many terms that convert at very different rates, so one bid ends up averaging good and bad traffic.

• Read the search term report regularly.
• Harvest terms that convert into their own exact match keywords, where they get their own bid.
• Negate terms that are irrelevant, not terms that are merely expensive. A relevant term with a high ACoS has a bid problem, and a negative removes the term instead of fixing the price. For a single bad term, negative exact is the narrow tool.

The same keyword competes in different places on Amazon. Amazon defines top of search as up to the first five positions in the top row of the first page of results on desktop (up to four on mobile), rest of search as the middle and bottom of results and later pages, and product pages as ads on product detail pages and other non-shopping placements such as the add-to-cart page.^[2]

There is no separate placement bid. A placement adjustment raises your bid when the ad competes for that placement, on top of your bidding strategy. Amazon's own example: a $1.00 bid with a 50% top of search adjustment becomes $1.50, and dynamic bids up and down can take it to $3.00.^[2]

The bidding strategy matters as much as the adjustment. With dynamic bids down only, Amazon may lower your bid for clicks less likely to convert. With dynamic bids up and down, it may raise or lower your bid by up to 100% for all placements. With fixed bids, it uses your exact bid.^[1]

So read each placement's own ACoS and revenue per click. Set the base bid so the weakest placement lands on target, then lift the stronger placements with adjustments in proportion to what their clicks earn.

Conversion rates and click prices can differ by hour. Shaping bids by hour is dayparting, and it belongs on top of a base bid that is already right, not in place of one. Our dayparting guide covers when it is worth it and what Amazon offers natively.

Stock changes one thing: whether a click can turn into a sale. Raising bids on a product with a few days of stock left buys sales you may not be able to ship. Hold raises when the runway is short, and stop spending on ad groups with nothing to sell.

Rules are "if this, then that". They are predictable and easy to audit. They are also blunt. A fixed click threshold ignores how often each product converts, a fixed step ignores how far the keyword is from target, and two rules can push the same bid in opposite directions.

Algorithms compute the bid from the data and the target: the step grows with the distance from target and shrinks when the evidence is thin. The risk is a black box. Insist on a reason beside every change, a preview before it is applied, limits on how far a bid can move, and a way to undo.

Amazon automates too: dynamic bidding adjusts bids auction by auction, and rule-based bidding adjusts base bids to hold an eligible Sponsored Products campaign to a ROAS guardrail you set.^[1] A sound set-up lets an algorithm set base bids, uses rules for business exceptions such as a promotion, and lets a person override both.

Change a bid, then let it buy enough clicks to show what the new price does. A second cut before the first has been measured stacks two guesses, and the steps compound.

• Treat recent days as incomplete. Amazon reports Sponsored Products conversions against the hour of the click they are attributed to, and conversion data can be revised for up to 60 days after the click, so the last few days usually look worse than they will.^[3]
• Bound each step. A cap on how far a bid can move per run, up and down, means one odd week cannot wreck a campaign. Big moves still happen, across runs.
• Pick the window on purpose. Long enough for the keyword to have orders, recent enough to describe today's price, stock and season. Start it after the last big change, such as a price change, a stock-out or a promotion.
• Base bids on settled data, hours on hourly data. Base bids are best judged on daily results that have settled. Hourly shaping is a separate job.

• Bidding to an industry ACoS instead of your own break-even.
• Calculating from the bid setting instead of the cost per click you actually paid.
• Reading a keyword with one order as a winner or a loser. One order is mostly luck.
• Negating converting search terms because their ACoS is high.
• Cutting so far that the ad stops being shown. A bid far below the market is not cheap, it is dark.
• Raising keywords inside a campaign that is over target as a whole.
• Re-pricing a dayparted campaign from the current hour's bid instead of its base.

Shurq's bid optimizer follows the steps above. This is what it does, as the product explains each change to the seller:

• Your target is the instruction. Break-even is measured from your own profit and loss: what each product keeps after Amazon's fees and the cost of goods you entered. Presets set the target from it: Profit at 50% of break-even, Balanced at 75%, Growth at 110% (paying slightly past break-even to buy volume). Custom uses exactly the number you type, and break-even is shown beside it as information, never as a hidden cap.
• Four groups. Every keyword and product target lands in High ACoS (cut toward target), Low ACoS (raised), High Spend No Sales (judged on whether its next click can pay for itself) or Low Visibility (nudged up). Rows on target are left alone.
• Priced on the next click. A row's own clicks and orders are blended with how its product family converts. The blend never reaches into a different product line or an account average; a product with no record holds rather than borrowing one.
• Placements on their own evidence. When a losing placement has to pay, only that placement gets cheaper, and placements that were earning keep the price they were paying.
• Guardrails. Your own maximum increase and decrease per run; a delivery floor, so no bid is cut below the point where the ad stops showing; raises capped by what a click earns back; a cool-off before repeating a move that has no evidence yet; no raises inside a campaign over its target; an inventory guard; and rank defense for tracked exact-match keywords. A dayparted campaign is priced from its schedule's base.
• A reason on every row, and nothing hidden. Each row names its group and the single rule that set its number. Nothing reaches Amazon until you press Apply, and every applied batch is logged in Optimization History and can be rolled back in one click.

Bidding lives in the AI Ad Manager, alongside rules, dayparting, harvesting and negatives. The underlying methodology is in our documentation, and pricing is a flat monthly subscription with no percentage of ad spend.

• Amazon dayparting: shaping bids by hour of day.
• Agency vs software vs in-house: who should run your bids.
• Bid optimization and bid adjustments in the glossary.
• Compare Shurq with other Amazon PPC tools.

1. Amazon Ads Support Center, Bidding strategies for Sponsored Products. https://advertising.amazon.com/help/GCU2BUWJH2W3A8Z7 (updated 28 September 2026). Checked 3 October 2026.
2. Amazon Ads Support Center, Adjust Sponsored Products bids. https://advertising.amazon.com/help/GYYZVM7LGSRYGWV5 (updated 28 September 2026). Checked 3 October 2026.
3. Amazon Ads API, Sponsored Products performance datasets (Amazon Marketing Stream). https://advertising.amazon.com/API/docs/en-us/guides/amazon-marketing-stream/datasets/sp-performance. Checked 3 October 2026.