A bid is a forecast
Every bid is a prediction: how much is one click on this keyword worth to me? Bid too high and you overpay for traffic that was never going to pay it back. Bid too low and you quietly disappear from the placements that would have converted.
Almost all bad bidding comes from making that forecast with a feeling instead of a number. The good news is the number is simple, and you already have everything you need to calculate it.
Halve the conversion rate and the justified bid halves with it. That is why listing work makes every campaign cheaper — it raises the bid you can afford on every keyword at once.
What you are actually bidding into
Your bid is not what you pay. It is the most you are willing to pay, entered into an auction against everyone else who wants that impression — and what you actually pay is set by the competition, not by your number.
The industry consensus is that Amazon runs a second-price auction, where the winner pays just above the runner-up rather than their own bid. Amazon does not publish the mechanics, so treat that as a working model rather than a fact: it is consistent with what advertisers observe, and it has never been confirmed.
What Amazon does state is that relevance participates. The highest bid does not automatically win, because an ad unlikely to be clicked or to convert is worth less to Amazon than a cheaper one that will be.
This assumes a second-price auction, which is the industry consensus but is not documented by Amazon. Treat it as a working model. Either way, relevance participates — the highest bid does not automatically win, because an ad nobody clicks is worth less than a cheaper one that converts.
Revenue per click is the whole foundation
A click is worth what it earns. If a hundred clicks produce ten orders at $40, those clicks earned $400 between them — so a click on that term is worth $4. That is revenue per click, and it is just conversion rate multiplied by order value.
Everything else in this guide is a modification of that one number. Get RPC per keyword and bidding stops being a matter of opinion.
Break-even ACoS is not your target ACoS
Two numbers get confused constantly, and the difference is your entire profit.
Break-even ACoS is your profit margin. If you keep 30% of the sale price after all costs, then at 30% ACoS you are handing every cent of that margin to advertising — you have bought a sale and earned nothing. Target ACoS is the number you actually aim at, and it sits below break-even by however much profit you intend to keep.
This is why “what is a good ACoS?” has no general answer. A 35% ACoS is excellent on a 60% margin product and ruinous on a 25% margin one.
Equal to your margin. Here you have bought a sale and earned nothing.
Set below break-even by however much profit you intend to keep.
This is why “what is a good ACoS?” has no general answer. 35% is excellent on a 60% margin product and ruinous on a 25% margin one.
The maximum bid formula
Put those together and the bid falls out. Your maximum CPC is what a click earns multiplied by the share of it you are willing to spend:
- RPC = conversion rate × average order value.
- Max CPC = RPC × target ACoS.
- Worked: a term converting at 10% on a $40 product earns $4.00 per click. At a 25% target ACoS, that justifies a $1.00 bid.
What to bid when you have no data at all
The formula needs a conversion rate, and a brand-new keyword has none. Rather than guessing, borrow: use the conversion rate of a similar term on the same product, or the product’s overall rate, and treat the resulting bid as a hypothesis to be corrected rather than an answer.
Start from Amazon’s suggested bid as a sanity check on the auction. If your calculated bid is far below it, you are unlikely to get enough impressions to learn anything; if it is far above, be careful — you are about to pay a lot to find out.
Give every keyword a job
Before touching a bid, decide what you want the keyword to do. The right move depends entirely on which situation you are in:
Spend with no sales is not a bid problem
The most commonly mis-handled case deserves saying plainly: a keyword that has spent real money and produced no orders does not need a lower bid. Lowering the bid just makes it waste money more slowly.
If the term is irrelevant, negate it. If it is relevant but the page cannot convert it, that is a listing problem. Either way the bid is not the lever, and treating it as one is how accounts accumulate dozens of low-bid keywords quietly bleeding budget for months.
The three bidding strategies
Amazon can adjust your bid in real time based on how likely it thinks a click is to convert, and which strategy you pick changes what your bid actually means.
Testing, or when you want the number to mean what it says
Unproven terms — Amazon lowers when a sale looks unlikely
Proven converters you want to win more often
Placement multipliers apply on top of this, so the ceiling is higher again. Set the base bid knowing it can move, rather than bidding aggressively and letting Amazon raise it further.
Placement multipliers stack on top
Placement adjustments are applied on top of your bid, for top of search, rest of search and product pages, and each can go up to 900%.
Multipliers stack with dynamic bidding rather than replacing it, which is how effective CPCs end up several times the number in the bid column. If you are running up-and-down with a top-of-search modifier, your $1.00 bid is capable of clearing well above $2.00, and the bid column will keep telling you $1.00.
Top of search is a different product
Top of search usually converts substantially better than the rest of the page and costs substantially more. Both halves of that are real, so the placement is worth paying for on exactly the terms where you can convert the traffic — and worth avoiding elsewhere.
The clean pattern is to earn it rather than buy it blindly: find terms that already convert well and are losing top-of-search impressions, and push the modifier there. Applying a modifier across a whole account is just a general price increase.
Illustrative shape, not measured data — check your own placement report. Both bars move together, which is the point: top of search is not a free upgrade, it is a different price for different traffic.
Your bid is relative, not absolute
A bid you set in March is not the same bid in November. You did not change it, but everyone around you did — and a static bid in a rising auction is a quietly shrinking share of the traffic.
This is why accounts seem to decay without anyone touching them. Impressions drift down, the ACoS still looks fine because the remaining traffic is your best traffic, and by the time it is obvious you have lost months of volume.
Nobody touched this campaign. ACoS probably still looks fine, because the traffic you kept is your best traffic — which is exactly why the decline goes unnoticed until the volume is gone.
How much data before you move a bid?
This is the same discipline as keyword research, and it is broken just as often. Two clicks is not a signal. Neither is one good day.
Judge a bid change on clicks rather than days, using the threshold from the Green Belt — roughly one divided by your conversion rate is the clicks needed to expect a single sale, and a few multiples of that before the number means anything. Amazon suggests reviewing bids around every two weeks, which is deliberately unexciting advice for the same reason.
Move in steps, not leaps
Bid changes should be proportional to your confidence. On a term with a hundred clicks of history, a 15% adjustment is a reasonable correction. On a term with twelve clicks, any adjustment is a guess, and a 50% one is an expensive guess.
Small consistent corrections compound and stay legible. Large swings produce a sawtooth: overbid, panic, underbid, lose the placement, overbid again — with a fresh learning period each time and no useful data at the end of it.
The swinging account is never at the right bid, and each move restarts the learning period — so it never collects the clean data that would have told it where to settle.
Bids and budgets are different levers
A bid decides whether you enter the auction and where you land. A budget decides when you stop for the day. Confusing them produces two opposite mistakes, and both are common.
Raising a bid on a campaign that already hits its daily cap does not buy more sales — it buys the same number of clicks at a higher price, and runs out sooner. Raising a budget on a campaign that never hits its cap does nothing at all. Check which constraint is actually binding before you touch either.
Controls price per click and placement. Raise it to be seen more often or higher up.
Controls total daily volume. Raise it only if you are actually hitting the cap.
Bidding to rank is a different calculation
Everything so far assumes you are bidding for profit on this sale. Sometimes you are not — during a launch, or when defending a position, the sale is a means to an end and the RPC formula will tell you to stop short of the bid you actually want.
When you bid to rank, the return is not the margin on the order, it is the organic position the velocity buys you. That is real but slower and harder to measure, so it needs an explicit budget and an explicit deadline rather than a quietly relaxed ACoS target.
- Bidding for profit — the bid is capped by RPC × target ACoS. The sale pays for itself.
- Bidding for rank — the bid is capped by what you decided to invest, and judged on rank movement and organic share.
- Bidding to defend — the bid is capped by what losing the position would cost you, which is usually more than the order is worth.
Bid to the goal, not to one number
Not every keyword should be bid to the same target. A defensive term on your own brand, a head term buying visibility during a launch, and a long-tail term printing margin are three different jobs, and holding all three to one ACoS number will make you turn off the first two.
The account-level number worth watching is TACoS — ad spend against total sales — because it tells you whether your ad spend is buying organic growth or just renting sales. A rising ACoS with a falling TACoS is usually a business getting healthier, not sicker.
Bidding pitfalls to avoid
- ✓Calculate RPC per keyword, not per account
- ✓Set target ACoS below break-even deliberately
- ✓Check which constraint is binding first
- ✓Move in steps sized to your confidence
- ✓Change one thing at a time
- ✓Push top-of-search where you already convert
- –Treating margin as a target
- –Lowering bids on zero-sale terms
- –Reacting to two clicks
- –Bid, budget and placement changes in one week
- –One ACoS target across every job
- –Forgetting multipliers stack with dynamic bids
Let automation hold the line
The maths in this guide is not hard. Doing it across four thousand keywords every week, without fatigue and without letting a feeling override the number, is the part humans reliably fail at.
That is the honest case for automation here: not that it is cleverer than you, but that it applies the same discipline on the four-thousandth keyword as the first, and it does not get bored in November.
Master Sifu — The hard part of bidding is not the arithmetic, it is applying the arithmetic to the one keyword you have a feeling about. Whatever holds the line — a rule, a checklist, a colleague — its whole job is to be unmoved by that feeling.
Common questions
How do I calculate the maximum bid for a keyword?
Multiply your revenue per click by your target ACoS. Revenue per click is your average order value times your conversion rate, so a $40 product converting at 10% earns $4 per click; at a 25% target ACoS the most you can justify is $1.00. Halve the conversion rate and the justified bid halves with it.
What is break-even ACoS?
The ACoS at which a sale makes you no profit — equal to your margin after all costs. It is a ceiling, not a target. Your target ACoS should sit below it by however much profit you intend to keep, with the exception of a launch, where spending past break-even buys rank rather than profit.
Do I pay my full bid on Amazon?
Usually less. The industry consensus is that Amazon runs a second-price auction, where the winner pays just above the runner-up’s bid rather than their own — though Amazon has not documented this, so treat it as a working model. Relevance also participates: an ad nobody clicks is worth less than a cheaper one that converts.
Should I use dynamic bidding up and down, or down only?
Down-only is the conservative default and the right choice while you are still learning what a term is worth. Up-and-down lets Amazon raise your bid by up to 100% where a conversion looks likely, which suits terms you have already proven. Fixed bids are for testing, because they keep the variable you are measuring still.
How often should I change bids?
Amazon suggests reviewing roughly every two weeks, which is deliberately unexciting advice. Change bids in steps of 10–20% rather than leaps, and judge the result on accumulated clicks rather than elapsed days — a big move gives you a new number with no way to know which part of it was the change.
Sources
Primary documentation this guide is built on. Amazon revises programme rules and fees, so check the source before acting on a number.
- 1Understand bidding — Amazon Ads
- 2Guide to dynamic bidding with Sponsored Products — Amazon Ads
- 3Bid adjustments — Amazon Ads
- 4Rest of search bid adjustment for Sponsored Products — Amazon Ads
- 5Tips to optimize advertising campaigns — Amazon Ads