Weighted keyword scores that mean something
Not all keywords deserve equal votes. How weighting turns a pile of positions into a single score that reflects business reality — and the traps to avoid.
Any tracker can average your positions. The problem is that an average treats “brand-name query that pays the rent” and “speculative blog keyword from 2023” as equal citizens, which no business actually believes. Weighting fixes the arithmetic to match reality: important keywords count more, and the resulting single score becomes something you can honestly put in front of a decision-maker.
Why equal votes lie
Suppose ten keywords: one high-intent term at #3, nine informational strays around #30. Average position ≈ 27 — sounds dire while the keyword that matters is doing great. Next month the stray crowd drifts to #25 while your money term slips to #6: the average improves as the business outcome worsens. Any aggregate that can move opposite to reality will eventually do so in the exact meeting where it matters.
Choosing weights without astrology
Weights encode “how much do we care” — keep the encoding simple and defensible:
- By intent: transactional and comparison terms high; informational medium; brand terms either high (if contested) or excluded (if you rank #1 by default, they just flatter the score).
- By page value: keywords pointing at money pages outrank keywords pointing at the blog.
- Three tiers beat ten decimals. 3× / 2× / 1× captures most of the truth; precision beyond that is confidence you don’t have. Revisit quarterly, change openly — a silently reweighted score is a small lie with compounding interest.
The same logic powers share of voice, where the weighting is by position value instead of (or alongside) business value; combined, they answer “how visible are we, weighted by what we care about” — which is the actual question.
Traps
- Gaming by pruning. Deleting losers improves the score and nothing else. Keyword-set changes should be annotated like site changes.
- Weight churn. If weights change every month, the trend line measures your indecision.
- One global score hiding segments. Keep per-cluster scores (per product line, per market) alongside the headline — a flat total can hide one segment’s collapse behind another’s boom.
- Forgetting the score is a proxy. It exists to detect movement worth investigating; the investigation still happens at the keyword and page level.
Done plainly — stable set, honest tiers, annotated changes — a weighted score is the rare SEO number that survives contact with a CFO: one line, business-shaped, and hard to argue with because the arguing was done when the weights were chosen.