Thousands of rows and a Friday afternoon
Every week someone downloads the Sponsored Products search term report, opens it in Excel and starts filtering. Terms with clicks and no orders. Terms converting well in an auto campaign that should be added as exact match. Terms that are obviously irrelevant, a competitor's brand, a different product type, a spelling that means something else. Each one gets copied into a negative list or a new campaign by hand.
It is slow, so it gets done every other week, then monthly. The filters depend on who is doing it. And the same irrelevant term keeps spending money in three campaigns because it was only added as a negative to one of them.
Why the review never keeps up
The report is large and grows with your catalogue. Search terms arrive in many forms: typos, plurals, ASINs from product targeting, other languages in European marketplaces. Deciding what to do with each one needs your rules and your product knowledge, which mostly live in one person's head.
- The report covers every campaign at once, but actions are per campaign or ad group.
- Negatives added in one place do not protect other campaigns.
- There is no memory of what was reviewed last week, so the same terms get looked at again.
- Thresholds such as how many clicks without a sale shift from person to person.
- Product targeting terms (ASINs) mix with keyword terms and need different handling.
What an irregular review costs
Wasted clicks on irrelevant searches keep spending until someone adds the negative. Good converting terms stay stuck in broad or auto campaigns where you cannot control their bids. Your campaign structure drifts away from the one you designed. And the person doing the review spends time on sorting rather than on decisions.
When an agency does this for you, the same is true, except you cannot see how consistently it is done.
Picture a seller of stainless steel water bottles. Searches for 'glass water bottle', 'water bottle for kids' and a rival's brand name all get clicks through the auto campaign. None convert. Someone negates 'glass water bottle' in the auto campaign in March, but the broad match campaign keeps paying for it until July, because nobody checked that one. Multiply that by every product line and every irrelevant variation, and the waste is steady rather than dramatic, which is why it survives.
The weekly routine we build
- A weekly pull of search term data for every Sponsored Products campaign through the Amazon Ads API, with a longer look-back window you choose so low-volume terms are judged fairly.
- Rules you set for each product group: clicks without an order before a term becomes a negative candidate, conversion level before it becomes an exact-match candidate, and words that always mean irrelevant.
- Grouping of similar terms, such as plurals and spelling variants, so one decision covers the family.
- A memory of past decisions: terms already negated, already promoted or deliberately left alone are not proposed again.
- A short review list for the campaign manager, with each proposal, the numbers behind it and the campaigns it would apply to.
- On approval, negatives and new keywords are added through the API to the right campaigns and ad groups, and every change is logged.
| Search term pattern | Proposed action |
|---|---|
| Many clicks, no orders, clearly irrelevant | Negative exact or phrase across campaigns |
| Converting well in auto or broad | Add as exact in a manual campaign |
| Competitor brand term | Flag for your policy decision |
| ASIN from product targeting | Target or negate as a product |
| Too little data yet | Keep watching |
Nothing changes in your campaigns without a person approving it, unless you decide later that some clearly defined rules can run on their own.
Your Friday afterwards
The review list is short: perhaps a few dozen proposals instead of thousands of rows. The campaign manager goes through it, approves most, rejects some and adds notes. Rejected proposals are remembered and do not come back next week.
Over time, the rules get tuned to your products. A kitchenware seller might treat any search containing 'toy' as irrelevant; a baby products seller certainly would not. Those decisions are made once and applied consistently.
Because every change is logged, you can also look back and see what happened after a batch of negatives went in, which is the only honest way to judge whether the rules are right.
Does this describe your ad management?
- The search term report is reviewed irregularly or not at all.
- Negatives are added to one campaign but not the others.
- Good terms stay in auto campaigns for months.
- Only one person knows the rules for what counts as irrelevant.
- You cannot say what changed in your campaigns last month and why.