The weekly optimizations

Every change you make either expands or contracts. Expanding is raising bids, adding keywords and testing new campaigns, to find more sales. Contracting is lowering bids, pausing and negating, to waste less.

You’ll do both every week on individual keywords. The account-level question is which way to lean. If ACOS is well under target and you want more sales, lean toward expanding. If it’s over target, lean toward contracting.

Make the changes at the level the data lives. Bids and pauses on targets, negatives on search terms, pauses on ads. A campaign’s totals blend everything in it together, so a campaign that looks bad can have great keywords hiding inside it, or one search eating most of the budget that a single negative would fix. You’ll rarely need to pause a whole campaign.

The rules below use 30 days for bid changes and 90 days for pausing and negating, unless lesson 6 says otherwise.

Raise bids on winners

Look at	Keywords and targets with at least 2 orders
When	ACOS is below your target by a few points (I use a 3 to 5 point buffer, so 22% or lower for a 25% target)
Do	Raise the bid 10%
Watch out for	Leave anything close to target alone. Set a max bid so a keyword that keeps qualifying can’t climb forever.

Lower bids on losers

Look at	Keywords and targets with at least 2 orders
When	ACOS is above your target plus a 3 to 5 point buffer
Do	Lower the bid 10 to 20%, more the further it’s off. With a 25% target: 10% for 30 to 40% ACOS, 15% for 40 to 50%, 20% above 50%
Watch out for	Small bids. A 10% cut on a $0.25 bid is under 3 cents.

Get impressions on targets that aren’t showing

Look at	Keywords and targets over the last 7 to 30 days
When	Zero impressions
Do	Raise the bid 10% a week, with a max bid
Watch out for	If a target still has zero impressions after 90 days of weekly raises, the bid probably isn’t the problem. Start it over in a new campaign or ad group. Check the bidding strategy too (lesson 4).

Percentages barely move small bids. A 15% weekly raise takes a $0.25 bid about 10 weeks to reach $1. Give low bids a floor (anything under a set amount jumps straight to it) or a minimum step of a few cents.

Pause targets that spend without selling

Look at	Keywords and targets over 90 days
When	0 orders and more than 20 clicks (exact and product targets) to 40 or more (phrase, broad, auto and category)
Do	Pause the target
Watch out for	Targets with 1 order get the same treatment at about double the clicks, 50 to 80. On broad targets, try negating the bad searches first.

Negate search terms that spend without selling

Look at	The search term report, 90 days
When	0 orders and about 20 clicks (use your own number from lesson 6)
Do	Add the search as a negative exact keyword in that campaign or ad group
Watch out for	Look for words that keep showing up across bad searches. A negative phrase on the shared word can catch dozens at once.

Promote searches that sell

Look at	The search term report from auto, phrase and broad campaigns
When	2 or more orders. Ask for more (3 or 4) before adding it as a phrase or broad keyword.
Do	Add it to your manual campaign as an exact keyword (or a product target, if it’s an ASIN)
Watch out for	Don’t negate it where it came from while it’s still selling there (lesson 7).

Adjust placements

Look at	The campaign’s placement report, 30 days
When	A placement’s ACOS is above target, or below it with room to grow
Do	Move that placement’s adjustment down or up 10 to 25 points (from 50% to 30%, say)
Watch out for	Adjustments can’t go below 0%. If a 0% placement is still too expensive, lower the bids instead.

Pause products that don’t sell

Look at	The advertised products in each ad group, 90 days
When	A product has plenty of clicks and no orders while others in the ad group sell
Do	Pause that product’s ad
Watch out for	Check the listing (price, stock, reviews, the main image) before blaming the ad.

How often

Weekly is a good rhythm by hand. Pick a day, run the rules, and you’re most of the way to well-run campaigns. Smaller changes more often respond faster to what the market is doing, which is where automation earns its keep, because nobody wants to do this every morning.