KGR Analysis Worksheet: Test Whether the Heuristic Works for Your Site
The Keyword Golden Ratio is useful only if it improves decisions on your site. This worksheet turns KGR from a confident-looking decimal into a testable hypothesis.
The conventional formula is:
KGR = estimated allintitle results / estimated monthly search volume
Do not multiply the result by 1,000. A previous version of this page used that incorrect formula and presented invented success rates, conversion ranges, and case-study outcomes without supporting data. Those claims have been removed.
For a conceptual explanation and limitations, read Keyword Golden Ratio: a filter, not a ranking formula. This page focuses on experiment design.
1. Record the raw inputs
Create one row per candidate query. Preserve the source and date instead of saving only the calculated ratio.
| Field | What to record |
|---|---|
| Query | Exact phrase being evaluated |
| Market | Country, language, and device assumption |
| Volume | Value or range reported by the selected tool |
| Volume source | Tool, match settings, and collection date |
| Allintitle estimate | Observed result estimate and collection date |
| KGR | Allintitle estimate divided by volume estimate |
| Intent | Informational, commercial, transactional, local, or mixed |
| Proposed page | Guide, product, comparison, tool, category, or other |
Google explains that Keyword Planner historical metrics are rounded, depend on location and network settings, and fluctuate with seasonality and events. Treat them as estimates, not observed organic demand. See Google Ads Help.
2. Add an intent and evidence gate
Before selecting a topic, inspect the live result page and answer:
- Does the proposed format match what people appear to need?
- Can this site add first-hand experience, original data, a working tool, or a clearer process?
- Is the topic inside the site’s established purpose and expertise?
- Is there a meaningful next action for the reader?
- Can the team maintain the page when facts change?
Reject the candidate if these answers are weak, even when the KGR is low. Google’s people-first content guidance explicitly warns against producing content mainly to attract search visits or publishing across many topics without real audience value.
3. Create a matched comparison
Select a small batch, such as six KGR candidates and six non-KGR candidates. Match them as closely as practical on:
- publication window;
- page type and approximate production effort;
- topic relevance to the site;
- internal-link support;
- author and review process;
- market and language.
This is not a scientific ranking experiment because search demand and competition cannot be held constant. It is still more informative than publishing only low-KGR pages and attributing every result to the formula.
4. Define outcomes before publishing
Use a fixed observation window and record both search and business outcomes:
| Layer | Possible measure |
|---|---|
| Eligibility | Indexed canonical URL, no accidental exclusion |
| Discovery | First impression date and query coverage |
| Visibility | Impressions and position distribution by query group |
| Traffic | Search clicks and landing-page sessions |
| Value | Qualified signup, lead, sale, tool completion, or assisted action |
| Cost | Production time, updates, and review effort |
Search Console reports are sampled and limited in several ways. Compare query groups and pages over the same dates, but do not interpret missing query rows as zero demand.
5. Use a stop rule
Decide in advance when to stop or revise the method. For example:
- stop expanding if the KGR batch does not outperform the matched batch on useful outcomes after the agreed window;
- revise pages where impressions appear but intent-aligned clicks or actions do not;
- consolidate pages that attract overlapping queries and fail to provide distinct value;
- abandon candidates that require unverifiable claims or repetitive content.
Do not respond to weak results by producing 100 more pages. A failed small test is useful because it prevents a larger quality problem.
6. Write the review note
At the end of the window, document:
- which inputs proved unstable;
- whether low KGR aligned with weak live results;
- which pages earned new query coverage;
- whether visits completed the intended task;
- what other factors better explained the outcome;
- whether KGR stays, changes, or leaves the editorial workflow.
The result should be a decision about your process, not a universal claim about Google’s algorithm. Continue with Search Console query clustering to group observed demand rather than creating one page for every keyword variation.
