After managing hundreds of brands on Amazon, the most common thing we find inside a new account isn’t a bidding problem or a creative problem. It’s a leak. Money pouring into search terms that have never produced a single sale — and that nobody is watching.
The industry numbers back up what we see in the field: brands waste an average of 25-35% of ad spend on irrelevant search terms, and in accounts with no negative keyword discipline, 30-40% of spend goes to queries that have never converted. We audited a kitchen brand last quarter that was burning roughly $9,400 a month on terms with zero attributed sales over 90 days. Their agency was reporting a “healthy 22% ACOS.” The ACOS was healthy because the converting terms were subsidizing the leaking terms inside the same blended number.
This is the most boring, least glamorous, highest-ROI work in Amazon PPC. Here’s how we run it.
Why the leak hides in plain sight
Negative keywords are the lever nobody wants to own. Adding keywords feels like growth. Pulling search terms feels like janitorial work. So it doesn’t get done — and the leak compounds.
It hides for three reasons:
- Blended ACOS masks it. Your winners carry your losers. A 22% blended ACOS can contain a cohort of search terms running at 80%+ ACOS and another cohort at 0% conversion. The average looks fine. The components are bleeding.
- Auto and broad campaigns are leak machines by design. They’re supposed to spray across queries to find new ones. That’s their job. But without a harvesting and negation cadence, they keep paying for the same dead-end queries month after month.
- It’s invisible on the dashboard. Spend without sales doesn’t show up as a red flag in Seller Central’s headline view. You have to go pull the Search Term Report to see it, and most operators don’t, weekly.
The two halves of the system
People treat negative keywords as a defensive cleanup task. That’s half of it. The real system has two moves that run together:
Harvesting — pulling converting search terms out of auto/broad campaigns and promoting them into their own exact-match campaigns where you control the bid.
Negation — blocking non-converting and off-target search terms so auto/broad campaigns stop paying for them.
Done together, every dollar in your discovery campaigns is doing one of two things: finding a new winner to graduate, or being cut off. Nothing sits in the middle leaking indefinitely. Run one without the other and you either starve your exact campaigns or let the leak run.
The negation thresholds we actually use
Don’t negate on gut feel and don’t negate too early — you’ll kill terms before they’ve had a fair chance. These are the rules we run:
- Zero-sale spend threshold. A search term that has spent 2-3x your target CPA (or your product’s price) with zero conversions is a negation candidate. For a $30 product, a term that’s spent $60-90 with no sale is leaking. Negate it.
- High-ACOS, low-volume terms. A term converting at 3x your target ACOS that isn’t strategically valuable (not branded defense, not a launch term) — negate or it bleeds slowly.
- Click volume without conversion. Industry guidance lands around 15-30 clicks with no sale as the action threshold. We use a spend-based trigger instead of pure clicks because a $5 product and a $200 product have completely different click economics.
- Irrelevant-intent terms. Anything where the intent is clearly wrong — competitor’s exact-model-number queries you can’t win, “free” or “cheap” modifiers on a premium product, wrong-use-case terms. Negate on sight; you don’t need data to know intent is mismatched.
Match type matters: use negative exact for a specific dead query so you don’t over-block, and negative phrase for a recurring junk modifier (like “for kids” on an adult product) you want gone across variations.
The cadence: weekly, not “when we remember”
The single biggest difference between accounts that hold a tight TACoS and accounts that drift is cadence. The leak is a flow, not a one-time mess, so a one-time cleanup doesn’t hold.
Here’s the rhythm we run on managed accounts:
- Weekly (15-20 min per account): Pull the Search Term Report for the last 7-14 days. Negate anything past the zero-sale spend threshold. Harvest any new converting term into exact match. This is the workhorse pass.
- Bi-weekly: Run an n-gram analysis — break search terms into individual words and look for losing patterns, not just losing terms. If “refurbished” appears across twelve different leaking queries, you negate the word once at the campaign level instead of chasing twelve terms.
- Monthly: Audit your negative list itself. Over-aggressive negation throttles discovery. Pull terms you blocked 60-90 days ago and sanity-check you didn’t wall off a category you should be in.
The weekly pass is the one that pays. Brands that work the Search Term Report consistently cut wasted spend 20-30% within a single quarter — and that savings drops almost entirely to the bottom line, because you’re not losing sales, you’re losing waste.
What the recovered spend should actually do
Cutting 25% of waste doesn’t mean cut your budget 25%. That’s the mistake. The point is reallocation:
- Push the recovered budget into your harvested exact-match winners, where ACOS is controlled and incremental.
- Fund your top-of-funnel discovery deliberately instead of accidentally — the leak was funding discovery badly; now you fund it on purpose with guardrails.
- Watch TACoS, not just ACOS. When you plug the leak and reinvest into winners, ACOS may even tick up slightly while TACoS drops — because more of your total revenue is now coming from efficient spend and organic lift from the terms you graduated. That’s the number that ties to profit.
The Rufus wrinkle (2026)
One 2026 note: as more discovery shifts into conversational and AI-surfaced results, auto campaigns are picking up a messier spread of long-tail and question-shaped queries. We’re seeing more junk variety in search term reports than two years ago, which means the negation cadence matters more now, not less. The leak has more places to hide. If you set negatives once in 2024 and walked away, your account has almost certainly developed new leaks you haven’t seen.
FAQ
How much ad spend is normal to “waste”?
Some discovery spend on non-converters is the cost of finding winners — that’s healthy. The problem is repeat spend on the same dead terms. If 30%+ of your spend is going to terms with zero sales over 60-90 days, that’s a leak, not discovery.
Will negative keywords hurt my organic ranking?
No. Negatives only stop your ads from showing on a query. They don’t touch organic. If anything, reallocating spend to converting terms strengthens the organic rank lift on the terms that matter.
Should I negate inside auto campaigns or at the account level?
Negate at the campaign/ad-group level for control. Account-level (portfolio) negation is a blunt instrument — useful for truly universal junk (competitor brand names you’ll never win, profanity, irrelevant categories), risky for anything borderline.
Can’t software just do this automatically?
Rules-based tools help with the volume, but auto-negation without a human reviewing intent will block terms you actually want. We use tooling to surface candidates and a human to make the call on anything strategic. Full automation is how you accidentally wall yourself out of a profitable category.
How fast will I see the impact?
Wasted spend drops the week you start. The TACoS improvement and organic lift from reinvesting into winners shows up over 30-60 days.
The bottom line
The negative keyword audit isn’t sexy and it doesn’t make a good case study slide. It just quietly returns 20-30% of a leaking ad budget to the brands willing to do it every week. Most aren’t. That’s the opportunity.
If you’re looking for a team that manages every lever — creative, advertising, and operations — Velocity Sellers works with brands doing $100K+/month on Amazon. Contact us for a free account audit.
