Content Quality Scoring: Give Every Article a Checkup With 6 Dimensions

You publish 12 articles in a month and organic traffic barely moves. You flip through the dashboard article by article — each one looks fine — yet you can’t say which to keep, which to rewrite, and which to merge. What’s missing isn’t effort, it’s one unified ruler: content quality scoring.

Here’s the bottom line: break quality into six dimensions — search match, information density, structural readability, credibility, reading experience, and conversion guidance — each scored 0 to 5, out of 30 total. Below 18, rewrite first; 18 to 22, do incremental optimization; 23 and up, just add internal links and a cover. One person can score 20 articles in about 90 minutes, and what you get is a processing list you can drop straight onto the calendar. For how to define low-quality pages, the boundary thinking in the UGC quality scoring piece is a good reference.

1. Search match: does it precisely catch one real search intent, with the main keyword and two or three variants naturally distributed, and give the answer in the first paragraph.
2. Information density: how much usable information per unit of length — every section has numbers, steps, or reproducible details, and after reading you know what to do.
3. Structural readability: can you scan it in 30 seconds and grasp the thread — H2s use action sentences, and key conclusions are carried by lists and tables.
4. Credibility: why should readers believe this — it needs data sources, real cases, or experiment records you ran yourself.
5. Reading experience: is it tiring to read on a phone — first screen goes straight into the body, images carry captions, paragraphs stay under five lines.
6. Conversion guidance: is there a clear next step after reading — an action list at the end, relevant entry points in the body that are easy to click.

Why judging quality by feel always goes wrong

Feel gets pulled off course by three things: the writer has sunk cost in their own draft, the reviewer gets influenced by layout and cover, and in team discussions the loudest voice wins. So revision resources forever flow to whatever got complained about most recently, not to the page where one change yields the biggest payoff.

Scoring isn’t about how precise the number is — it’s about compressing a vague argument into six answerable questions. In a review meeting, “this article isn’t deep enough” becomes “information density 2/5: not a single concrete number in the whole piece, not one reproducible step”. The writer hears the second sentence and immediately knows what the next version needs.

The ruler also protects writers. Scores are built from dimensions, anyone can re-check against anchors, and an editor can’t bounce a draft with a bare “doesn’t feel right” — rework counts drop noticeably.

What the six dimensions each measure

Each dimension answers exactly one question, avoiding mutual contamination. Anchors are locked in advance: 0 means not done at all, 3 is peer average, 5 is top three in search results for the same topic.

The six items default to equal weight. In a growth phase, give search match and information density a 1.5x multiplier each; when you enter the monetization phase, raise conversion guidance’s weight. Once the weights are set, don’t touch them for at least a quarter, or scores from the two batches can’t be compared horizontally.

The operating flow for scoring 20 articles at once

• Pull data: export impressions, clicks, and average ranking for the last 90 days from Search Console, merge into one table by URL, with the scoring columns to the right of the data columns.
• Calibrate anchors: first trial-score three known-good and three known-bad articles so everyone agrees on what 3 means, then start batch work.
• Time-box each article to 4 minutes: scan subheadings for the skeleton, read the first 200 words of the first screen, spot-check two facts, and if time runs out give the score now and mark it for re-review.
• Stick scores onto data: the batch that scores high but clicks low is usually stuck on title and snippet — fix the meta description first following the title and meta description CTR writing approach; it’s the cheapest change there is.
• Two-person spot-check: randomly pull 5 of the 20 for a second person to re-score; if any dimension differs by more than 2 points, go back and rewrite that dimension’s anchor description.

Batch scoring should be done in one sitting. If you come back the next day, the ruler quietly loosens. In practice, the score distribution of the same batch is usually high in the middle and low at both ends — if 80% of your articles score above 24, the anchors are almost certainly set too loose.

How scores become scheduled actions

Don’t pile all the low-scoring articles into one week. Handle two or three a week and leave an observation window, or you can’t tell whether the score lift actually brought click growth. For the specific rewrite steps, just apply the old-content refresh checklist — no need to redesign the flow every time.

Four calibration moves that keep the ruler straight

• Build an anchor sample library: store two on-site examples per dimension, one scoring 2 and one scoring 5; newcomers read all six pairs before starting.
• Re-score 10 old articles every quarter: if the same batch’s scores rise collectively, it’s usually the ruler loosening, not content genuinely improving.
• Bind with review standards: the judgment basis for structural readability and credibility should quote the content style guide and review directly, so the score sheet and the review checklist don’t become two sets of rules fighting each other.
• Record before and after: each rewrite leaves a row with the before-score, after-score, and change date; after three months you can calculate which type of change pays best, and fold labor costs in using the content ROI data standard.

The score sheet itself needs iteration too. After two full quarters, look back: if one dimension never correlates with traffic changes, it means what it measures isn’t important to your site — swap it out decisively. Don’t let the table grow into a formality you just go through.

When it really comes to landing, the actions are concentrated: today, build the table — six dimensions plus a total column — and paste in the URLs of your last 20 published articles; this week, finish the first scoring round, scoring only without editing, to see what the distribution looks like; pick 3 articles scoring 18 to 22 for incremental optimization and check click changes two weeks later; send the 0-to-11 list to the editor-in-chief to confirm merge or take-down, and don’t leave them hanging.