Negative keywords in Google Ads are the cheapest performance lever most ecommerce accounts never pull. Junk queries drain spend week after week while all the attention goes to bids and budgets. This is the categorization system I run across managed accounts, and the weekly rhythm that makes it compound.
Why Smart Bidding does not solve the negative keywords problem in Google Ads
The pitch sounds plausible: the algorithm learns which queries convert, so it will stop bidding on the bad ones eventually. In practice, smart bidding lowers bids on weak queries; it rarely abandons them. As long as a query carries any predicted chance of converting, it keeps receiving impressions and clicks at your expense.
It gets worse in Shopping and PMax, where the algorithm also decides which queries your products match in the first place, and it casts wide by design. Wide matching plus reduce-but-never-remove bidding equals a permanent tax on your budget unless a human draws hard lines.
Negatives are those hard lines. They are the one targeting instruction Google cannot soften, reinterpret, or quietly expand.
The five categories of wasted spend most accounts miss
Wasted spend hides in categories, not individual queries. These five cover most of what I find in ecommerce audits:
- Competitor names. Shoppers loyal to another brand click, compare, and leave.
- DIY and how-to queries. "How to fix", "how to install", "DIY". These searchers want instructions, not products.
- Wrong-category queries. Your product shares words with something you do not sell: "shower curtain" traffic landing on shower enclosures.
- Job and information seekers. "Salary", "course", "meaning", "wholesale" when you are retail only.
- Bargain-mismatch queries. "Free", "cheap", "second hand" against a premium catalogue.
Each looks small on its own. Together they form a meaningful slice of Shopping spend in most accounts I open, and none of it was ever going to convert.
Reading the search term report to find the worst offenders fast
Do not read the report top to bottom. Sort it three ways and the damage surfaces on its own.
First, sort by cost with conversions filtered to zero: your most expensive dead queries appear immediately. Second, sort by clicks without conversions over a longer window, which catches the slow bleeders that never spike enough to get noticed. Third, scan high-impression terms for intent mismatches that have not cost much yet but will once volume arrives.
Twenty minutes with these three sorts finds most of the waste. The skill that builds with repetition is reading intent from the query itself: "walk in shower for elderly" and "walk in shower diy installation" contain the same product words and opposite buying intent.
Shared negative lists vs campaign-level negatives
Use both, for different jobs.
Shared lists hold the universal exclusions: the competitor list, the DIY list, the job-seeker terms. Build them once, attach them to every relevant campaign, and every new campaign you launch starts pre-protected instead of relearning old lessons with fresh budget.
Campaign-level negatives handle routing and specifics: keeping brand queries out of non-brand campaigns, stopping one product category from absorbing another's queries, and excluding terms that are wrong for one campaign but right for another.
The mistake is putting routing negatives into shared lists, which silently blocks queries you want somewhere else. Shared means universal. If it is not universal, it stays at campaign level.
Shopify-specific negatives: patterns that show up in almost every store
Some query patterns recur across nearly every Shopify store I manage, whatever the niche:
- Platform queries: "shopify", "store", "website" attached to your product terms by people researching stores, not buying products.
- Marketplace intent: "amazon", "ebay", "aliexpress", "temu" variants from shoppers who will buy there anyway.
- Aftermarket queries: "replacement", "spare parts", "repair" when you sell complete products.
- Coupon hunters: your brand plus "discount code" burning non-brand budget.
- Wholesale and bulk queries hitting retail listings.
Seed your shared lists with these on day one rather than paying to rediscover them one click at a time. They cost almost nothing to add and start saving immediately.
The weekly rhythm: what it costs when done systematically
Done as a system, the weekly pass is short: pull the search term report for the last seven days, run the three sorts, categorize the offenders into the right lists, done. On a mature account that is twenty to thirty minutes. On a neglected account the first pass takes longer because you are clearing months of backlog, and that first pass is usually where the biggest single saving of the year comes from.
The compounding is the point. Every excluded junk query redirects its budget toward queries that can convert, and the algorithm's conversion data gets cleaner, which sharpens its bidding everywhere else.
The fashion account holding a 17.3x ROAS runs on exactly this discipline. Nothing about its search term hygiene is clever; it is just never skipped. It is also a fair test of what you are paying an agency for.
Measuring the impact after a clean-up
Three numbers move after a serious negative keyword pass, in a predictable order.
First, spend distribution: the share of budget reaching converting queries rises within days, visible in the search term report itself. Second, click-through and conversion rates tick up as junk impressions and clicks fall out of the mix. Third, ROAS follows over the next weeks as smart bidding relearns on cleaner data.
What does not happen is an overnight ROAS spike. The mechanism is the removal of waste plus a cleaner signal, and both compound rather than jump. Screenshot your baseline metrics before the clean-up so the comparison stays honest, and judge the result on a month, not a weekend.
Maintenance is the strategy
There is no finish line on negatives. New queries appear constantly because search behavior shifts, products launch, and competitors rebrand. The accounts that stay efficient are the ones where the weekly pass is as fixed in the calendar as invoicing.
If you want to know what your account is leaking right now, the free 48-hour audit includes a search term review, and my Google Ads management runs this rhythm as standard.