Where keywords actually come from
The best keyword list isn't built from guessing or a generic keyword tool alone — it's built primarily from your own search term report, which shows the actual searches that led real shoppers to click your ad. A term a keyword tool suggests as high-volume is far less valuable than a lower-volume term your own data shows converts.
The three match types, and when to use each
Broad match matches your keyword loosely — related searches, synonyms, and variations can all trigger your ad. It casts the widest net and is best used for ongoing discovery of new converting terms, at the cost of some spend on less relevant matches.
Phrase match requires the search to contain your keyword phrase in that order, though other words can appear around it. It's a middle ground — narrower than broad, but still catches reasonable variations.
Exact match matches only the specific keyword (and very close variants, like plurals or minor misspellings on some platforms). It gives you the tightest control and is where you want your proven, best-converting terms living with a correspondingly higher bid, since you're confident about their performance.
A practical progression
- Start an automatic campaign (or a broad-match manual campaign) to gather real search term data.
- After a meaningful sample, identify terms with strong conversion performance.
- Move those terms into an exact-match campaign or ad group with an elevated bid — you know they convert, so it's worth paying more to make sure you win those auctions.
- Keep a phrase-match layer for terms that are promising but not yet proven enough for the aggressive exact-match bid.
- Continue running broad match or automatic targeting alongside this, purely for ongoing discovery of new terms as the search term report keeps producing new candidates.
Negative keywords: the other half of the job
Negative keywords tell the platform not to show your ad for specific searches, even if they'd otherwise match. This matters most for two situations: excluding irrelevant terms an automatic or broad-match campaign has matched to (wasting spend on searches that will never convert for your product), and preventing your own campaigns from competing against each other on the same term (e.g., excluding a term from your broad-match campaign once you've moved it into a dedicated exact-match campaign, so you're not bidding against yourself).
Negative exact blocks only that specific search term. Negative phrase blocks any search containing that phrase. Use negative exact for a single clearly irrelevant term, and negative phrase when an entire family of related searches is consistently irrelevant (e.g., excluding "kids" and "toy" phrase variants from an adult-product listing that keeps getting mismatched).
Worked example
A seller advertising a leather laptop sleeve reviews their automatic campaign's search term report after three weeks and finds: "13 inch laptop sleeve leather" (12 clicks, 3 sales — strong), "leather laptop bag" (8 clicks, 1 sale — decent), and "leather ipad case" (22 clicks, 0 sales — clearly a mismatch, since the product doesn't fit an iPad). They move the first term into an exact-match campaign with a bid raised from the automatic campaign's average, keep the second as a phrase-match candidate to watch longer, and add "leather ipad case" as a negative exact keyword to the automatic campaign so it stops paying for that irrelevant match.
How often to review and update
Weekly is a reasonable cadence while campaigns are newer or lower-volume; sellers with larger, more established accounts often move to a biweekly or monthly deep review once the keyword lists have stabilized, supplemented by automated bid rules for day-to-day adjustments. Reviewing too infrequently lets wasted spend accumulate on obviously poor terms; reviewing too frequently on tiny data samples leads to overreacting to noise.
Mistakes to avoid
- Building a manual keyword list entirely from a third-party keyword research tool without cross-checking against your own search term report data.
- Bidding on the same term in both a broad-match discovery campaign and a dedicated exact-match campaign without negating it in the broad campaign, causing your own campaigns to compete against each other.
- Adding negative keywords too aggressively based on a single non-converting click — a term needs a real sample of clicks with no sales before it's confidently a poor match, not just one unlucky click.
- Ignoring the search term report for months at a time, letting avoidable wasted spend accumulate.
Best practices
- Review the search term report on a fixed cadence and treat it as your primary keyword research source.
- Negate a keyword in your discovery campaign as soon as you've promoted it into a dedicated exact-match campaign, to avoid self-competition.
- Use negative phrase match for entire families of irrelevant searches rather than adding many individual negative exact terms one at a time.
FAQs
How many clicks does a term need before I can judge it as good or bad? There's no universal number, but treat anything under roughly 10-15 clicks with real skepticism — a single early sale or non-sale on that few clicks isn't a reliable signal yet.
Should I ever remove a negative keyword once added? Occasionally — shopper search behavior and your own catalog can change over time, so a term that was irrelevant a year ago might become relevant if you add a new, related product. Periodically re-check negative keyword lists rather than treating them as permanent.
Do negative keywords cost anything? No — adding a negative keyword doesn't cost money itself; it prevents future spend on a match you don't want, so there's little downside to using them once you're confident a term is genuinely irrelevant.