Amazon ad waste isn't a bad keyword — it's a bad pattern

Every account we inherit this time of year has the same story: Prime Day prep window is open, budgets are getting scaled up, ACoS is climbing, and the first instinct is to raise bids or reallocate budget. That’s the wrong move. The money isn’t leaking at the bid level. It’s leaking at the search-term level — and not where most operators look.

The mistake: scanning the Search Term Report row by row

The default way to work a Search Term Report is top-down by spend: sort by cost descending, scan the top 50, negative-match anything that didn’t convert. This catches the obvious waste.

It misses the real leak.

The real leak is a single low-intent root word appearing across 20, 30, 50 different search terms. Each individual row looks fine — 8 clicks, no orders, too little data to flag. All of them together represent hundreds of clicks and zero orders. Per-row review never catches it because no single row crosses the threshold.

It’s not that one keyword is expensive. It’s that you’re repeatedly buying a whole category of wrong traffic.

Three patterns to negative-match first

1. Low-purchase-intent roots

Words that attract clicks from people who are not looking to buy your product today:

• repair, replacement, manual, instructions, how to, parts, troubleshoot

Volume is usually decent. Conversion is near-zero. Category-dependent — if you sell repair kits, obviously different — but for most brands, these are pure spend drains.

2. Wrong-need-attribute roots

Words that make the traffic look relevant but indicate a different purchase intent:

• accessory, accessories for [your product], substitute, alternative, compatible with, replacement for, used, refurbished, DIY

The buyer wants something adjacent to your product — not your product. If you sell the main product at full price, the person searching “accessories for X” is not your buyer today.

3. Repeating-root-word patterns that look OK per row

The compound leak. Example: a kitchenware seller’s Search Term Report shows recipe, recipes, recipe ideas, recipe for [dish], easy recipes, quick recipes, how to cook, meal ideas — each with 6–12 clicks, zero orders, across 20+ rows. Per row: doesn’t cross the threshold. Combined: significant monthly spend, zero revenue.

The only way to find this pattern is to pivot the report by root word, not by the full search term.

The SOP we use

This is a weekly 20-minute exercise that saves more than raising bids would have lost:

1. Pull the Search Term Report for the last 14–30 days.

2. Don’t read it row-by-row. Pivot it by root word / word pair. Export to a spreadsheet. Tokenize the search term column. Pivot clicks and orders by root word. Now you see which root words have 100+ clicks and 0 orders.

3. Prefer phrase-match negatives over exact-match negatives. Phrase negatives kill all variants of the pattern in one move. Exact negatives leave 50 cousins still running.

4. For low-risk patterns, consider negating at the campaign level (not just ad group). If repair never converts in this account, put it at the campaign or even the ad group level across the account, not row-by-row.

5. Observe 3–7 days after negatives before scaling bids. Negatives change impression share, conversion rate, and ACoS simultaneously. Don’t conflate the effects by also raising bids — wait a week, then make the next decision with clean data.

The order most teams get wrong

Raise bid → add budget → launch new campaign → then go looking for waste. That’s gas pedal down before you fix the leak in the tank.

Reverse it: negatives first, then scale.

Tonight’s one task

If you’re running an ad account right now, open the Search Term Report for the last 14 days. Don’t read it row by row. Pivot it by root word. Look for the words that have high aggregate clicks and zero aggregate orders.

That’s where your Prime Day budget is currently being set on fire. Fix that first, then raise bids.