Once a content library crosses a few hundred pages, 'which pieces deserve more investment' stops being a question anyone can answer by eyeballing a traffic report. Different pages matter for different reasons — a low-traffic bottom-of-funnel page converting at 8% can be worth more than a high-traffic top-of-funnel post converting at 0.1% — and a single metric like pageviews rewards the wrong behavior, pushing teams toward broad, shallow traffic bait instead of the content that actually moves revenue.

A content scoring model solves this by combining multiple signals into one comparable number, so the team can rank hundreds of pages consistently and make update, promote, or retire decisions based on something more defensible than whoever advocates loudest in the planning meeting.

The four inputs: traffic, engagement, conversion, decay

A workable scoring model needs at minimum four inputs, each capturing something the others miss. Traffic (sessions or pageviews over a trailing window) captures reach but says nothing about quality. Engagement (average time on page, scroll depth, or pages-per-session from that page) captures whether readers actually consumed it or bounced in five seconds. Conversion (assisted or last-touch conversions attributed to the page, as a rate per session) captures commercial value. Decay rate (the trend in traffic over the last 2-3 periods, as a percentage change) captures whether the page is still earning its keep or sliding toward irrelevance.

No single input is sufficient alone — high traffic with high decay is a page about to become dead weight; low traffic with high conversion is a page that's quietly carrying pipeline nobody notices. The combination is what makes the score useful.

content_score_worksheet.txt

text

  CONTENT SCORE = (Traffic_norm × 0.3) + (Engagement_norm × 0.2)
                  + (Conversion_norm × 0.4) + (Decay_adj × 0.1)

  Step 1 — Normalize each raw metric to a 0-100 scale within its own
           distribution (percentile rank against all pages in the set).
  Step 2 — Weight conversion highest (0.4) since it's closest to revenue;
           weight decay lowest (0.1) since it's directional, not absolute.
  Step 3 — Decay_adj: +100 if traffic is flat/growing, 50 if declining
           <15%, 0 if declining >15% over trailing 2 periods.
  Step 4 — Bucket by score: 70+ = invest/promote, 40-70 = maintain,
           <40 = candidate for update or retirement (not auto-delete).

  Example:
  Page A: Traffic 85, Engagement 60, Conversion 90, Decay 100
          Score = 25.5 + 12 + 36 + 10 = 83.5  → Invest
  Page B: Traffic 95, Engagement 40, Conversion 10, Decay 50
          Score = 28.5 + 8 + 4 + 5   = 45.5  → Maintain, don't expand

Weight conversion over raw traffic

Teams that weight traffic highest end up optimizing for the wrong thing: a scoring model dominated by pageviews will always rank the viral listicle above the quiet page that converts enterprise buyers, and resourcing follows the score. Weighting conversion highest keeps the model honest about what content is actually for.

Score bucket to action mapping

Score bucket	Interpretation	Recommended action	Review cadence
70-100	High performer across multiple dimensions	Expand: add sections, build supporting cluster pages, promote in distribution	Quarterly
40-69	Solid but not exceptional, or strong in one dimension only	Maintain: light refresh, monitor decay trend	Semi-annual
15-39	Underperforming but may have salvageable intent or backlinks	Audit: rewrite, consolidate with a stronger page, or re-target keyword	Semi-annual
0-14	Negligible traffic, engagement, and conversion	Retire: redirect to a stronger page or remove entirely	Annual cleanup

Decay rate: the input that catches rot early

Decay rate matters because content performance isn't static — search algorithms update, competitors publish fresher pages, and facts go stale, so a page that scored 85 a year ago can quietly slide to 45 without anyone noticing until a quarterly review catches it. Tracking decay as its own input, rather than only looking at current-period snapshots, turns content scoring from a one-time ranking exercise into an early-warning system.

A practical decay signal: compare trailing 90-day traffic to the prior 90-day window for every page, flag anything down more than 15% for review, and separate 'seasonal dip' (same pattern last year) from 'real decline' (no prior-year precedent) before deciding whether to refresh.

Don't score pages that are too new

A page published six weeks ago hasn't had time to rank, accumulate backlinks, or build a conversion sample size — scoring it against a page that's been live for two years and penalizing it for low traffic just produces noise. Exclude content under roughly 90 days old from scoring, or score it on a separate 'early signal' track (indexation status, initial ranking position) until it has enough history.

Operationalizing the score

A scoring model is only useful if it feeds a recurring decision process, not a one-off spreadsheet exercise. The practical pattern: pull the underlying metrics monthly from analytics and the CMS, recompute scores, and route the bottom and top deciles into the content calendar — bottom decile as refresh/retire candidates, top decile as expansion candidates (new cluster pages, updated CTAs, increased internal linking). Most teams find a monthly compute cadence with a quarterly strategic review hits the right balance between staying current and not re-litigating the same pages every few weeks.

What's next

Scoring tells you which content is working, but the same prioritization discipline applies to format choice — video and podcast content need their own performance lenses because their engagement signals don't map cleanly onto blog metrics.

Next: Video Marketing Strategy →

I build these systems professionally.

Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.

Work With Me →

Need production AI or MarTech expertise? I build custom LLM workflows, server-side GTM tagging, and analytics infrastructures professionally.

Need custom AI or MarTech setup? Let's build together.

Let's talk →