Keyword Performance Analysis: A Quarterly Audit Routine for Pruning, Fixing and Scaling Keywords
A practical how-to for deciding which keywords earn their keep - defining success metrics, tiering by intent and value, classifying into keep/scale/rewrite/cull, and building a review cadence that survives the next algorithm update.

Who This Keyword Audit Is For - and When to Run One
This guide is for the in-house SEO or agency analyst with thousands of tracked keywords facing a harder question: which of them earn their keep? The job is not to build another dashboard - it is to walk into a quarterly review with a defensible answer for a marketing manager, grouped into keep, fix, promote or cut.
Run a keyword performance analysis when a quarter has closed, a core algorithm update has rolled through, a product line has launched or been retired, or the tracked list has grown past the point where anyone reads it. If you are still assembling raw data for the first time, that is a different task - this is the evaluation cycle that begins once the data exists.
The deliverable of an audit is not a spreadsheet. It is a short list of decisions with named owners - pages to rewrite on the same URL, topics to expand, and keywords to stop tracking.
Prerequisites: What Must Exist Before Keyword Performance Analysis Is Useful
Four things must be in place. First, at least 90 days of Search Console data - Google retains 16 months, so quarter-on-quarter and year-on-year comparisons are possible, but the first month of a new site or page is usually noise. Second, a conversion event tying a landing page to money or an enquiry, because keyword value must be expressed in Rands or qualified leads, not sessions. Third, a keyword-to-landing-page mapping, so you analyse page + query pairs rather than queries floating alone. Fourth, a brand/non-brand split so a single campaign spike does not distort every comparison.
Step 1: Define What 'Performing' Means for Each Keyword
A keyword only performs relative to a goal. Fix three success metrics before looking at data: expectation (position band or share of top-10 appearances), engagement (click-through against the impression baseline), and outcome (conversions, revenue or qualified leads attributed to the landing page). Write them down per intent tier, because a transactional query at position 8 may be worth ten times an informational query at position 2.
Reading Keyword Metrics Without Being Fooled: Position Averages, Impression Gaps and the 90-Day Window
Average position is an average, and averages hide distribution. A keyword reported at position 12 could sit at 3 for some searchers and 40 for others. Prefer percentile position or the share of queries appearing in the top 10. Watch the impressions-versus-clicks gap: high impressions with near-zero click-through usually means the SERP answers the query without a click, or your title is the wrong promise.
| Signal | Likely meaning | First action |
|---|---|---|
| Impressions up, clicks flat | AI answer or rich result absorbing clicks | Rework title/meta and check AI visibility |
| Position 15+ with impressions, zero clicks over 90 days | Page not competitive or wrong intent | Rewrite or cull |
| Two URLs alternating for one query | Cannibalisation | Consolidate or differentiate |
Google's documentation on Search Console data notes a 2-3 day lag and sampling on large properties, and low-volume queries are anonymised - so long-tail judgement needs rank-tracking or page-level aggregation alongside.
Step 2: Tier Keywords by Intent and Commercial Value, Not Volume
Sort every keyword into navigational, commercial-investigation, transactional or informational tiers, then weight each tier by revenue or lead value. A Cape Town ecommerce store with a R2,500 average order value and roughly 3% conversion rate can calculate that a transactional query converting at half the site average is worth around R37 per conversion-adjusted click - far more persuasive in a review than a volume figure. Volume tells you how crowded a room is; intent tells you whether anyone in it is buying.
Step 3: Analyse at Page + Query Level to Catch Cannibalisation
One query can map to several URLs, and two pages splitting the same query rarely rank as well as one focused page. Group Search Console rows by landing page, flag any query where two or more URLs each collect meaningful impressions, and decide which URL owns the intent. This is where scalable search growth either compounds or collapses into internal competition.


