To bid for a target ACOS, multiply the keyword's revenue per click by the target: bid = (sales ÷ clicks) × target ACOS. A keyword that earned $300 from 200 clicks makes $1.50 a click, so a 30% target gives a 45-cent bid. Revenue per click is an estimate, though, and on a keyword with few orders it is a loose one.
The formula is exact. The number you feed it is not.
Where the formula comes from
Amazon defines ACOS as ad spend divided by ad revenue. Spend is clicks times cost per click, so ACOS is the cost of a click divided by the revenue a click brings in. Turn that around and you have the most a click can cost at your target:
The arithmetic
ACOS = (CPC × Clicks) ÷ Sales = CPC ÷ Revenue per click
Revenue per click = Sales ÷ Clicks = Order value × Conversion rate
Bid = Revenue per click × Target ACOS
A bid is the most you are willing to pay for a click, so a bid set this way keeps every click at or under the price that hits the target, as long as revenue per click holds. Which target to use is a margin question: the break-even ACOS calculator gives the line a product cannot cross, and the target sits below it by the profit you want to keep.
How much one order moves the bid
Revenue per click comes from orders. Take a keyword with $25 orders that earned 3 orders from 40 clicks, priced for a 30% target, and give it one order fewer or one more over the same clicks:
| Orders from 40 clicks | Revenue per click | Bid at 30% |
|---|---|---|
| 2 | $1.25 | $0.38 |
| 3, as measured | $1.875 | $0.56 |
| 4 | $2.50 | $0.75 |
One sale either way moves the bid by a third. Give a keyword with 30 orders from 400 clicks the same one-order difference and its bid stays between $0.54 and $0.58, about 3% either side. The formula gives the first keyword a precise-looking number resting on almost nothing. Use it once a keyword has enough orders for one more or one fewer not to matter much, and use small percentage steps below that, with the lines drawn as in when to raise or lower a bid.
Read it off the data, not off the current bid
A common shortcut scales the current bid instead: new bid = current bid × target ACOS ÷ current ACOS. It looks equivalent and is not. The current bid is often not the bid that produced the window's clicks, because the bid has changed since. Run the shortcut daily on a window that has not caught up, and it applies the same correction again on every run. A keyword bid at $1.20 that paid $1.08 a click at 60% ACOS, against a 30% target, run three times on the same 30 days of data:
| Run | Bid × 30% ÷ 60% | $1.80 a click × 30% |
|---|---|---|
| 1 | $1.20 → $0.60 | $1.20 → $0.54 |
| 2 | $0.60 → $0.30 | $0.54, no change |
| 3 | $0.30 → $0.15 | $0.54, no change |
The revenue-per-click version reads only the window, so the same data gives the same bid, and the second run has nothing to do. The Target ACOS action in Vendlen Automation works this way: it sets a click-billed target's bid to Sales ÷ Clicks × Target ACOS over the rule's Data Window, and a target with no clicks, no spend or no sales is skipped rather than guessed at.
Two brakes on the move it asks for
The formula says where the bid should end up. It says nothing about how fast to get there, or whether the destination points the right way.
- Approach the target in steps. A keyword at 80% ACOS with a 50% target can be priced as if its target were 74% this time: 80% + (50% − 80%) × 0.2. That closes a fifth of the gap per change, so a wrong estimate costs a small step instead of a collapse in position. In Vendlen this is Target ACOS Glide, and a per-run cap, 25% by default, limits any single move.
- Never move against the evidence. A keyword bid at $1.20 that paid $0.90 a click at 45% ACOS is under a 50% target. Its revenue per click is $2.00, so the formula says $1.00: a cut, on a keyword doing better than its target. A direction gate holds the bid instead: below target the formula may not lower it, above target it may not raise it.
Where this still needs a person
The formula assumes revenue per click stays put when the bid moves. It may not: a different bid can change where the ad shows and who sees it. A new keyword, a seasonal one or one whose price just changed has no revenue per click worth trusting. And a ratio has no size: the formula prices a $3 keyword and a $3,000 one the same way, which is why the loss side of the account is better judged in money, as in why an ACOS threshold cannot tell $8 from $800.
What to do with this
- Work out your target ACOS per product: break-even ACOS less the margin you want to keep.
- For each keyword, take sales and clicks over 60 to 90 days. Divide to get revenue per click, and multiply by the target.
- Check the order count first. If one order more or less would move the bid by a fifth or more, use a small percentage step instead.
- Move part of the way: a fifth of the gap, or a fixed cap per change.
- If the keyword is under target and the formula says cut, or over target and it says raise, leave the bid where it is.
Questions sellers ask
What is revenue per click?
Ad sales divided by clicks over the same period. It is also average order value times conversion rate: $25 orders converting at 10% earn $2.50 a click. It is the most a click is worth in sales, and the input every target ACOS bid starts from.
Can I use the bid formula on a new keyword?
No. With no sales, revenue per click is zero and the formula would bid zero. Open a new keyword at a bid priced from similar terms, such as a share of their cost per click, and switch to the formula once it has orders of its own.
What target ACOS should I use in the formula?
Your break-even ACOS less the profit you want to keep on each sale. Break-even ACOS is the share of the price left after cost of goods, Amazon fees and other costs. A target copied from another product assumes that product's costs.