Most Amazon bid changes are too large, too fast, and based on too little.

What your bid actually controls

Amazon runs a second-price auction. You set a maximum, the highest-ranking eligible ad wins, and the winner pays marginally more than the runner-up needed. So a bid is not a price, it is an eligibility ceiling. Raising it from 0.80 to 1.00 does not raise your cost by twenty cents on every click. It makes you eligible for auctions you were previously excluded from, and it raises your cost only in the auctions where the competition genuinely sat between the two numbers.

That asymmetry matters because it makes upward moves less dangerous than sellers assume and downward moves less effective. Cutting a bid to reduce spend often reduces impressions far more than it reduces cost per click, so you lose volume without gaining much efficiency. If the goal is spend control, the daily budget is the honest lever. If the goal is efficiency, the bid is the right lever but the placement multiplier is usually the faster one.

Dynamic bidding and the multiplier stack

Amazon offers three bidding rules. Down only lowers your bid in auctions it judges less likely to convert, and never raises it. Up and down does both, and can raise your bid substantially in auctions it likes. Fixed uses your bid unchanged. On top of the rule sit placement multipliers, set per campaign for top of search, rest of search and product pages, which increase the bid for those specific slots.

These compound. A bid with an up-and-down rule and a top-of-search multiplier can clear at a figure well above the number you typed, which is the usual explanation when average CPC looks inexplicable. The practical starting position for a new campaign is down only with no multipliers, so the reported CPC reflects your actual bid and you can read the data cleanly. Add multipliers once the placement report shows you which slot is earning, not before.

Turning a target ACoS into an actual bid

The arithmetic is straightforward and almost nobody does it. Your maximum acceptable cost per click is your target ACoS multiplied by average order value multiplied by conversion rate on that keyword. If a keyword converts at 8%, your average order is 40 dollars, and your target ACoS is 30%, then your maximum CPC is 0.08 times 40 times 0.30, which is 0.96. Bidding 1.60 on that keyword is a decision to lose money, regardless of how strategic it feels.

The reason the calculation gets skipped is that conversion rate at keyword level is noisy, and on low-volume ASINs it is barely measurable. Use the ad group or campaign conversion rate as the input when the keyword's own data is thin, and recompute as the keyword accumulates history. A rough number derived from your own account beats a precise number derived from someone else's benchmark.

How to run an Amazon bid review

1. Set the data threshold first. Decide the minimum clicks or impressions a target needs before it is eligible for a bid change. Everything below the line stays untouched this cycle.
2. Convert your ACoS target into a maximum CPC. Maximum CPC is roughly target ACoS multiplied by average order value multiplied by conversion rate. That gives a bid ceiling grounded in your own numbers rather than a feeling.
3. Check the placement report before touching keyword bids. If top of search is where the loss is, the fix is the placement multiplier, not thirty individual keyword bids.
4. Move in increments, not in halves. Ten to twenty percent per cycle on targets with real data. Halving a bid on eight clicks of evidence is a coin flip you have dressed up as a decision.
5. Change one variable per cycle. Adjust bids or change the dynamic bidding rule or change the multiplier. Do all three at once and you have learned nothing from the result.
6. Wait out the attribution window before reading the outcome. Amazon sales land against clicks from days earlier. Reading a bid change after 48 hours mostly measures reporting lag.

The mistakes that cost the most

• Cutting bids to control spend. The daily budget controls spend. Bid cuts mostly cost you impressions.
• Turning on up-and-down bidding plus a top-of-search multiplier on day one. You have made your realised CPC unreadable before you had any data to read.
• Bidding without ever calculating maximum CPC. Target ACoS times AOV times conversion rate. Two minutes of arithmetic that reprices half the account.
• Large moves on thin data. A 50% bid cut on eight clicks is not a decision, it is a reflex. Set a threshold and respect it.
• Reading the result before the attribution window closes. You will conclude the change failed, revert it, and then watch the delayed conversions arrive.

What you can ask Muze to do

Muze connects Amazon Ads to ChatGPT, Claude or any MCP client with 48 tools, 12 read and 36 write. These are the real tool names behind each request.

• “Raise bids on my top converting keywords by 15% and show me the change first.”
  Reads the performance, proposes the new bid per keyword id, and previews the full list before anything is written. Tools: update_amazon_keyword_bid, list_amazon_campaigns.
• “Set the default bid on this ad group to 0.75.”
  Ad group default bid applies to targets without their own bid. Auto target bids are set separately, per Amazon-created target. Tools: update_amazon_ad_group_bid, update_amazon_auto_target_bid.
• “Adjust the bids on my manual product targets.”
  Sets a bid on an existing manual product target, or batch-creates new ones. New targets are created PAUSED. Tools: update_amazon_product_target_bid, create_amazon_product_targets.
• “Run the bid autopilot once now.”
  Applies bid changes within a guardrail clip range of 0.65x to 1.25x, using the config you set, and writes every action to an audit log you can read back. Tools: run_amazon_bid_autopilot, get_amazon_autopilot_config, list_amazon_autopilot_actions.

Nothing spends without you

• Every write previews first and waits for explicit confirmation.
• New campaigns, ad groups and product ads are created paused. Negative keywords are the exception: they are created enabled, because an exclusion that is switched off is not an exclusion.
• Pause and archive are separate tools. Archiving is irreversible on Amazon, so it is never what happens when you asked to stop spend.
• OAuth only. No Amazon Advertising API application to file, no keys to manage, and access is revocable.
• Never a percentage of ad spend. You pay for the software. Your budget stays yours.

Frequently asked questions

Can AI manage my Amazon bids?

Yes. Muze can update keyword bids, product target bids, ad group default bids and auto target bids over MCP, and can run a bid autopilot that moves bids only within a guardrail clip range of 0.65x to 1.25x. Manual bid changes preview before they run. Autopilot is opt-in and every action it takes is written to an audit log.

How much should I change a bid by?

Ten to twenty percent per cycle on targets with enough data to justify a move, and nothing at all on targets below your click threshold. Large swings on thin data produce a bid history nobody can interpret later.

What is the difference between down only and up and down bidding?

Down only lowers your bid in auctions Amazon judges less likely to convert and never raises it. Up and down does both and can raise the bid substantially. Down only is the cleaner starting point because your realised CPC stays close to the bid you set.

Why is my average CPC above my bid?

Placement multipliers and up-and-down dynamic bidding compound on top of the base bid. Check the placement breakdown and the bidding rule before concluding the bid itself is wrong.

Should I use bids or budgets to control spend?

Budgets. A daily budget caps spend directly. Cutting bids reduces how often you are eligible to appear, which costs you volume and may not improve efficiency at all.

How long before I can judge a bid change?

Long enough for Amazon's attribution window to close and for the target to accumulate clicks at the new bid. Reading the result after two days measures reporting delay rather than the change.

Does Muze charge based on how much I spend on ads?

No. Muze never takes a percentage of ad spend. You pay for the software and your budget stays entirely yours, which also means nothing in the product is incentivised to raise your bids.

Keep reading

The rest of the Amazon cluster