Raise a bid when ACOS sits well below your target, cut it when ACOS sits well above, and leave it alone in a band around the target. Draw the lines wider when the ACOS rests on few orders. With three orders, one sale more or less turns a 60% ACOS into 45% or 90%, so lines drawn for busy keywords act on noise there.
A rule with one line raises at 49% and cuts at 51%, every week, on the same keyword.
Three zones, not one line
The simplest bid rule has one line: below the target ACOS, raise; above it, cut. A keyword sitting near the target crosses that line back and forth on ordinary variation, and the rule follows it. The bid never settles, and every change resets what the next fortnight measures.
A working rule set has three zones: raise well below the target, cut well above it, and do nothing in between. An example for a 50% target, read over 30 days:
| Orders in 30 days | Raise | Leave alone | Cut |
|---|---|---|---|
| 3–17 | ACOS ≤ 37.5%: +4% | 42.5–55% | ACOS ≥ 75%: −6% |
| 18 or more | ACOS ≤ 45%: +3% | 50–58% | ACOS ≥ 60%: −4% |
Between the zones nothing fires on this window. A keyword at 65% ACOS with 5 orders is neither healthy enough to hold nor bad enough to cut by the 3–17 standard, and that is deliberate: the evidence does not support a move yet.
Why the lines move with the order count
ACOS is ad spend divided by ad revenue, and ad revenue arrives one order at a time. With $25 orders:
| Keyword | One order fewer | As measured | One order more |
|---|---|---|---|
| 3 orders, $45 spent | 90% | 60% | 45% |
| 30 orders, $450 spent | 62% | 60% | 58% |
Both keywords read 60%. One sale takes the first to 90%, deep in the cut zone, or to 45%, back inside the leave-alone band; the second stays within two points. Lines that suit the second fire on the first's noise, which is why the 3–17 row above raises only below 37.5% and cuts only above 75%, and the 18+ row narrows to 45% and 60%. The steps shrink too, from +4% and −6% to +3% and −4%: the lines sit closer to the target, so smaller corrections keep the bid inside them.
Below three orders the ratio barely exists: no orders means no ACOS at all. Judge those keywords by the money they have spent instead, as in why an ACOS threshold cannot tell $8 from $800.
Why a high-ACOS keyword is not switched off
Pausing is the strongest thing a bid rule can do, so it should ask for more than a cut. In the same example set, pause rules read 90 days and need ACOS of at least 90% at 3–17 orders and 70% at 18 or more. A keyword at 80% with 5 orders is above the target and above the cut line, so its bid is being stepped down, but it is under the pause bar. The rule is not broken. It cuts before it stops, and it asks for the most extreme reading where one order can still change the picture.
What the do-nothing zone protects
Leaving a keyword alone matters most when several rules can reach it. A rule that calculates a bid from a target ACOS, like the one in the bid formula article, would otherwise reprice a keyword that the step rules have just called healthy. InVendlen Automation the do-nothing zone is a rule of its own: a Bidding rule with a Percentage action of 0, a Hold, which keeps the bid and stops lower-priority rules from touching it in that run. It does not stop a higher-priority cut or raise; it only stops the rules ranked below it.
Why the bid can still go up and down
Zones on one window do not stop swings between windows. A keyword at 40% ACOS over 90 days meets a 90-day raise rule; after one bad fortnight, a 14-day cut rule takes it. Each rule is right about its own window. When both match on the same day, let the cut win: a wrong cut costs a little traffic, a wrong raise keeps paying. Across days, the swing is the two windows disagreeing, and the question to ask is which one describes the keyword now, the subject of why ninety days hides the fortnight that broke.
Where this still needs a person
Zones are about evidence, not strategy. A launch you are deliberately overspending on, a keyword you defend against a competitor, or a product about to go out of stock all want a different answer from the one the ACOS gives. So does any keyword where the target itself is wrong, because the lines are only as good as the target ACOS they are drawn around.
What to do with this
- Write down your target ACOS: break-even ACOS from the break-even ACOS calculator, less the margin you want to keep. Draw raise, leave-alone and cut lines around it, wide for keywords with few orders and narrower for keywords with many.
- Count orders before reading ACOS. Under three, look at spend instead.
- Move bids a few percent at a time, and wait for new clicks before the next change.
- Set the pause bar above the cut bar, and higher still for keywords with few orders.
- When two of your windows disagree, cut first and find out which window is current.
Questions sellers ask
How much should I raise a bid at a time?
A few percent per change, and less as the keyword gathers orders. On a $1.00 bid, 3% is three cents. Wait for new clicks before the next step, so each change is judged on traffic it actually received.
How often should I change bids?
When the keyword has gathered enough new clicks since the last change to show its effect, not on a fixed day. In Vendlen a cooldown, five marketplace days by default, holds a keyword after a change and can release it early once it gathers a set number of clicks or amount of spend.
Should I pause keywords with high ACOS?
Cut the bid first and pause only at a higher bar. With few orders, one sale can move ACOS by tens of points, so the pause bar should be higher still there. For keywords with no orders, ACOS does not exist; decide on the money spent.