How Amazon Advertising Bidding Actually Works

What bidding actually controls on Amazon

A bid on Amazon is the most you're willing to pay for one click on a keyword or product target. It is not what you pay. What you actually pay — your actual CPC — comes out of an auction against other advertisers targeting the same keyword, adjusted by whichever bid strategy you've chosen and any placement multipliers layered on top. Most sellers set a bid once and rarely check how far it drifts from that number in practice.

Three things combine to produce the price you pay in any single auction: your base bid, your bid strategy setting (down only, up and down, or fixed), and your placement bid adjustments — separate percentages for top of search, rest of search, and product pages. Change any one of the three and the effective bid on every future auction changes with it, which is why "raising the bid" rarely means what people think it means.

The three bidding strategies, and what each is actually doing

Amazon gives you three settings, and each behaves differently under pressure:

• Dynamic bids – down only: Amazon can lower your bid below what you set when a click looks unlikely to convert. It never raises it. This is the safest default and the one Amazon applies automatically if you don't choose otherwise.
• Dynamic bids – up and down: Amazon can raise your bid up to 100% for top-of-search placements and up to 50% for other placements when it predicts a high chance of conversion, and lower it otherwise. This is where most volume — and most ACOS volatility — happens.
• Fixed bids: No conversion-based adjustment at all. Only your placement percentages apply. You carry the full timing risk yourself.

A worked example: setting a bid from your margin, not a suggestion box

Say you sell a $40 item. After referral fee and fulfillment cost, it leaves $15 of contribution margin per unit — 37.5% of the sale price. If your conversion rate on a given keyword runs at 10%, your break-even cost per click is: 10% × $40 × 37.5% = $1.50. Bid above that on a sustained basis and you're buying the sale at a loss before you've covered anything else the business owes.

Most sellers don't run that calculation. They set the bid to whatever the suggested-bid range shows, or to whatever got a top-of-search impression last week, and never tie it back to their own margin. The suggested-bid range is Amazon's estimate of what wins auctions. It has no idea what your margin is.

Placement adjustments stack on top of the strategy — and they multiply

Here's the part most explainers skip: placement adjustments and dynamic bidding stack multiplicatively, not additively. A $0.80 base bid, with dynamic up-and-down at its 100% cap and a +50% placement adjustment for top of search, can produce an actual bid of $0.80 × 2 × 1.5 = $2.40 in a single auction — three times the intended number, and well past the $1.50 break-even from the example above.

That gap is where most avoidable ACOS damage happens. It isn't one bad decision. It's three reasonable-looking settings compounding on top of each other in a placement nobody checked separately.

The common mistake — and one we've made ourselves

The mistake we see most often, including on our own early accounts, is reacting to a rising ACOS by cutting the base bid across every keyword in a campaign instead of pulling the placement report first. Cutting the base bid punishes the keywords that were converting fine and barely touches the placement or search term actually causing the overspend. The ACOS number moves, but the account gets worse at finding the customers who were buying.

The discipline that avoids this: every proposed change should carry the evidence behind it, a measurement plan, and a rollback trigger, agreed before the change runs — not decided after the ACOS report comes in three weeks later. That's the standard we hold changes to on managed accounts, and it works just as well as a checklist for anyone bidding manually.

When the bid change doesn't work

If a bid change doesn't work — ACOS stays high, or a lowered bid kills volume you needed — check three things before touching the bid again. First, sample size: a keyword with fewer than roughly 15–20 clicks since the change hasn't told you anything statistically yet. Second, the placement report broken out separately, because a campaign-level ACOS can hide one placement subsidizing another. Third, whether the setting you think you changed actually took — dynamic bidding and placement percentages are stored separately, and it's easy to change one and assume the other moved with it.

If the number is genuinely wrong — a bid that's simply too high for that product's margin — the fix isn't a bigger cut. Go back to the break-even calculation for that specific ASIN, set the bid there, and give it the same waiting period before judging it again.

Where this fits if you'd rather not run the math by hand

Bidding by hand works fine on a handful of SKUs with stable margins. It gets harder to hold together across a full catalog, where every ASIN has a different break-even bid and placement mix shifts weekly. Dr. PPC reads the account, runs that break-even math per product against the brand's real economics, and proposes each bid change with the evidence, the measurement plan, and the rollback trigger attached — the client sets how much of it runs automatically. It's built and run by Full Circle, which has managed more than $500M in revenue across 100+ brands, priced at $300 a month plus 3% of ad spend, capped, with the first 30 days free. Whether or not that's the right fit, the break-even bid math above is worth doing regardless of who turns the dial.