Taxonomy and scoring
How GrowthOS classifies and scores every page: separate health and quality scores, a six-dimension quality framework, and the diagnosis that turns scores into priorities.
GrowthOS scores every page on two axes it keeps deliberately apart: Health, which asks whether the page is well built, and Quality, which asks whether it's the right page for the searcher. Keeping the two separate is the heart of the methodology. A broken page and a thin page need opposite fixes, and a single blended score would hide which one you're looking at.
This page covers the full chain: how a page gets classified, how the two scores work, what quality measures, and how scores turn into a ranked set of decisions. The exact weights and formulas stay under the hood. The reasoning is the part worth understanding, and it's the part that makes the scores mean something. For the plain definitions without the methodology, see reference: scores.
Classify before you score
A score only makes sense once the system knows what a page is and who it's for. So classification comes first, on three layers.
| Layer | What it is |
|---|---|
| Personas and ICP | Personas describe the people you sell to. The ICP describes the companies they work at. Together they're the full target. |
| Taxonomy | The vocabulary you classify work with: categories you define, and the values inside each category. |
| Content Clusters | Named strategic territories you're betting on, like "spend management" or "developer onboarding," each one a first-class object tied to pages, opportunities, and briefs. |
These classifications are why Quality can be judged per persona. The same page can read strong for one buyer and weak for another, and the system only knows that because it knows which personas the page is meant to serve. Classification is the lens; scoring is what you see through it. Every classification here lives on the page's record in the page portfolio.
Two scores, never blended
Health and Quality answer different questions, draw on different evidence, and lead to different work. GrowthOS reports them as two numbers on purpose.
| Health | Quality | |
|---|---|---|
| The question | Is this page well built? | Is this the right page for the searcher? |
| What it looks at | Crawlable, fast, structured, machine-readable | Format fit, trust, structure for extraction, originality |
| Measured against | Technical standards | The page's search intent |
| Output | A score from 0 to 100, plus an A to D grade | A rating, judged per persona where relevant |
Why not roll them into one number? Because a clean-but-empty page and a strong-but-broken page can average to the same middling score, and yet one needs a rewrite while the other needs an engineering fix. Blend them and you throw away the one piece of information you needed to act. Keep them apart and the scores tell you not just that a page is weak, but what kind of weak. To act on these two scores for a real page, read a page's scores walks the page-level read.
What Quality measures
Quality is intent-relative, so it breaks into dimensions that each ask a different question about whether the page serves the searcher. Intent alignment carries the most weight, because the wrong format fails no matter how well the page is written.
| Dimension | What it asks |
|---|---|
| Intent alignment | Does the format match what this query's stage calls for? |
| Entity trust | Are author, organization, and first-hand-experience signals present and verifiable? |
| Information gain | Does the page say something the top results don't? |
| Content structure | Can a machine extract the answer? |
| Brand presence | Do domain-level authority signals back the page up? |
| Engagement craft | Is it written to hold attention, starting with the hook? |
The highest-leverage element inside content structure is the answer capsule: a short, self-contained summary right after the headline that states the answer before the article elaborates on it. It's the piece an AI engine can lift and cite, so it does more for citation than any other single edit. Information gain is the other differentiator, the reason an engine cites you instead of the three pages that say the same thing.
The exact weight on each dimension stays internal. What's worth keeping is the order of operations: get the format right first, then earn trust and say something new, then structure it so a machine can read it.
The Big Picture label
Underneath the detail numbers, GrowthOS rolls four signals (Health, Quality, impressions capture, and traffic capture) into one readable status. The label is graded against the page's own potential, not an abstract bar.
| Label | Read it as |
|---|---|
| Optimal | Capturing close to its potential |
| Good | Performing well, with minor headroom |
| Developing | Climbing, not there yet |
| Needs Work | Underperforming its potential |
| Incomplete | Missing the data or content to judge |
"Graded against its own potential" is the idea that matters here. GrowthOS asks how much of what a page could capture it actually captures. That reframe turns a wall of scores into a list of decisions, because it surfaces the gap between where a page sits and where it could be.
From scores to a diagnosis
Scores describe a page. A diagnosis decides what to do about it. On top of the two scores, GrowthOS reads a page on six axes and compresses the result into a short read and a single highest-leverage action.
| Axis | The question |
|---|---|
| Lifecycle | Where is this page in its arc, from new to declining? |
| Momentum | Which way is it moving right now? |
| Traffic tier | How much does it matter by volume? |
| Query health | Is the keyword strategy working, and how? |
| Readiness | Is the content and technical foundation sound? |
| Competition | How contested is the result page for its target query? |
Two pieces of the reasoning are worth carrying, even though the exact math behind them stays under the hood.
The first: readiness is a matrix, not an average. It combines Health and Quality as a pair. A page at 95 Health and 15 Quality and a page at 55 and 55 average to the same number, but they need opposite work, so the system treats the pair, never the mean.
The second: urgency is multiplied, not added. Factors compound instead of summing, so a high-traffic page that's collapsing rises far above a quiet page with the same symptoms. The highest-leverage fixes are the ones that repair a whole section of the site at once.
A Health score and a Quality rating ship today, and so does the per-page read. The full Quality model that resolves Quality into a single 0-to-100 score, and the diagnosis layer that ranks the whole portfolio into one work queue, are on the roadmap. We label what's live and what's coming so you always know which is which.
Where to go next
Last updated at June 3, 2026
The page portfolio
Every URL on your site as one inventory of scored assets, where each page carries a type, a search intent, and its own scores.
Opportunities
An opportunity turns one piece of keyword research into a decision you can triage: a query, the intent behind it, where it came from, and the page that would answer it.