Skip to content
Ecomwith

Google Search Term Mining: Expand, Clean, and Segment Intent

  • Intermediate
  • 55 minutes
  • Pro lesson · free with an account until 2027-04-26

From diagnosis to queries

From diagnosis to queriesRead orderStarting point
20oz leak-proof tumbler for commuteBuying intent, page specs, margin, and query cost.Expand or split only when evidence supports it; one order is not a budget case.
free tumbler patternFree pattern or free-shipping context; do not act on the root alone.Clean only when the meaning is stable; pause before touching the shared negative list.
custom logo tumbler bulkBulk/custom intent, MOQ, quote page, and lead quality.Split into the B2B/quote route instead of mixing it into the DTC query pool.
20oz tumbler lid replacementSupport/parts intent, order value, and service path.Route to parts or support; do not negate it account-wide.

Google search-term mining: choose the action before expanding or cleaning

The search terms report is not a negative-keyword trash bin. Return to the raw query, query cluster, page/feed fit, and order value before choosing expand, clean, split, or observe. When complete queries are hidden, record the aggregated theme and uncertainty instead of treating it as line-level proof.

Google Ads Pro Lesson 2

If the search terms report is only used for cleanup, half of its value is lost. The better read is finding promotable high-intent queries, themes worth separating, and low-quality intent quietly absorbing budget.

Fix the report scope first

A search term is what the shopper typed; a keyword is only the trigger you configured

Keep using the non-brand Search Campaign for the 20oz leak-proof bottle. Export a fixed Search terms report and keep the query, triggering keyword, match type, Campaign, cost, conversions, and order value. The search term is what the shopper actually typed; the keyword is the condition in your ad group that helped trigger the ad. They are different, so a keyword label alone cannot decide intent.

Choose one query cluster, then ask what kind of demand it bought with the budget. “20oz leak-proof tumbler for commute” may be purchase intent. “Free tumbler pattern” may mean a free template or free shipping. Read the raw phrase, landing-page fit, and order quality before choosing expand, clean, split, or observe.

Without a fixed window, raw query, cost, and at least one order or page signal, classify only. Do not turn one term directly into a budget action or negative.

Set action boundaries first

Use the same search terms report to separate expand, clean, split, and observe

Search term work fails when action comes too fast: low conversion becomes a negative, high conversion gets more budget, or a repeated root edits the shared list. Sort each term into four actions before doing anything else.

Exercise · pick one

Try it first

Use when
High intent, stable performance, page fit, and profit support.
Evidence
Order quality, margin, new customers, stable query pattern, and page fit.
Avoid
Leaving it only inside broad match and waiting.

Pro lesson · free with an account until 2027-04-26 · Parts 3-24 · 22 parts

Members: full exercises and records

This part includes

  • Fix the report scope before sorting
  • Turn each cluster into a testable action
  • Close on one worksheet

An account is free; sign in to read this part.

Checking your membership…

What you finished

FAQ

What is the difference between search terms and keywords?

Keywords are targeting inputs you set or allow the system to match. Search terms are the real phrases shoppers typed before your ad showed. Read the real search terms, sort them into waste, potential, brand, competitor, compatibility, and bulk/custom groups, then choose expand, clean, split, or observe.

Should every low-converting search term become a negative keyword?

No. First inspect raw query, match type, spend, page behavior, add_to_cart / purchase, order quality, and whether the page or Feed answers the demand. Low conversion can mean bad intent, weak page fit, thin sample, or long conversion lag. With only one-system evidence, observe or collect more proof.

Governance routes after term mining

Sources

Edition notes

  • Author Ranfeng Wei
  • Published
  • Updated
  • Reviewed
  • Next review

This update: Restored the FAQ and tidied the source list.