Use when you have a Search Console export and want the near misses rather than a fresh keyword plan.
gsc-striking-distance-mining.md
You are mining a Search Console export for queries that are close to paying off. You are not doing keyword research from scratch.
Query export (query, landing page, clicks, impressions, CTR, average position): {{GSC_EXPORT}}
Our own CTR by position band: {{CTR_BENCHMARK}}
Impression floor worth bothering with: {{MIN_IMPRESSIONS}}
Pages that matter commercially: {{PRIORITY_PAGES}}
Produce three ranked lists, each as a table.
1. Near misses: rows from {{GSC_EXPORT}} at average position 5 to 15. Columns: Query | Page | Impressions | Position | What is likely holding it back (intent mismatch, thin coverage, weak internal links, title) | First action.
2. Click leaks: rows where CTR sits well under {{CTR_BENCHMARK}} for that position band. Columns: Query | Page | Position | Actual CTR | Expected CTR | Likely reason (title, description, a SERP feature above us, brand mismatch).
3. Wrong page ranking: queries landing on a URL other than the one in {{PRIORITY_PAGES}} that should own them. Columns: Query | Ranking page | Intended page | Fix.
Constraints:
- Exclude every row under {{MIN_IMPRESSIONS}} and state how many rows were dropped.
- Average position is an average across many impressions, not a rank on one SERP. Do not describe it as one.
- Where a query appears against several pages, say so rather than quietly picking one.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When do I mine Search Console rather than do keyword research?
When the site already has impressions. Keyword research tells you what exists; this tells you what you nearly have, which is cheaper to act on because the page is already built and already ranking. Reach for research instead when a whole topic is missing rather than under-performing.
What does the export need to contain?
Query and page level rows, not query only, or the wrong page ranking table cannot be built at all. You also need your own click-through rate by position band, an impression floor worth bothering with, and your commercially important pages. Filter to one country and one device before exporting.
Which of the three tables should I work from?
Near misses at position five to fifteen, click leaks against your own curve, and queries landing on the wrong URL. The third is usually the cheapest, since it is an internal link or canonical fix rather than writing. The second is only trustworthy if the benchmark is your own curve, not a published one.
What is the mistake that costs me here?
Reading average position as a rank. A query showing 8.4 may never once have appeared at eight; it is a blend across many impressions, and mixing countries and devices in the export flattens it further, so pages look closer to the top than they ever were. Filter first, then trust the bands.