How to Prioritize Keywords: Can’t Pick 5 Out of 300 Candidates? Use 3-Dimension Scoring

Your keyword list has 300 candidates, and the topic meeting argued for half an hour without settling on which five to write this month. Keyword prioritization is essentially the problem of missing one unified ruler — put each keyword into a three-dimension score of “traffic × difficulty × business value”, let a single number decide the order, and the argument naturally ends. For how intent layers map to business value, see intent mapping by funnel; for what else to mine beyond tool scores, see zero-search-volume high-conversion keywords.

Here’s my advice: most sites should first concentrate on the batch of keywords with “medium search volume, KD below 20, high business value” — far more stable than chasing head terms from the start, and it pays off faster. The math is simple — total = traffic score × weight + inverse difficulty score × weight + business score × weight, with the three weights summing to 1, tuned by site stage. A spreadsheet takes half an hour to build, and every batch after that is just sorting and taking the top ten.

How to quantify each of the three dimensions

Traffic score uses monthly search volume or Search Console impressions, normalized to 0–10. Difficulty takes the inverse: the lower the KD, the higher the score, so all three add in the same direction. Business score looks at how close a keyword is to a sale — “buy / price / discount / supplier” words get 8–10, “what is / how to” words get 2–4. To make it easy to execute, just use these tiers:

• Traffic score: monthly volume 0–100 = 2 points, 100–1000 = 5 points, 1000+ = 8 points
• Difficulty score: KD under 10 = 10 points, 10–30 = 7 points, 30–50 = 4 points, 50+ = 1 point
• Business score: purchase intent = 9 points, comparison intent = 6 points, pure informational = 3 points

Of the three dimensions, difficulty is the easiest to get wrong, because tools only look at backlinks for KD, not the actual makeup of the search results page. To really judge, glance at whether the top ten are all dominated by big platforms and official sites — that’s more reliable than the number itself. Business score is the same: don’t guess it; first classify keywords with the four search intent types, then apply fixed scores by category, so three people on the team produce consistent numbers.

Distribute weights by site stage

With the same scoring table, a new site and a mature site should get completely different rankings. A new site has no accumulated authority, so difficulty weight should go above 0.5, actively avoiding keywords with KD above 40; e-commerce sites should push business weight to 0.4, because orders are the point; content sites can lean on traffic, building scale with informational keywords first. See this table for reference weights:

Weights aren’t fixed. From month three, domain authority accumulates gradually, so the difficulty weight should come down month by month; if you don’t adjust, you’ll keep avoiding keywords you could actually win, handing the opportunity to competitors.

Four steps from scoring to execution

Import the keyword list into a spreadsheet, score the three columns, add a weighted-sum column, and sort by total descending. Take only the top 10 into each writing queue, then revisit rankings and clicks two weeks after publishing and fine-tune the weights. Batches that are too big lengthen the feedback cycle — what you want is fast validation of whether this weight mix works.

1. Export candidates to a spreadsheet, score the three columns, using one source to avoid inconsistent definitions
2. Weighted sum gives the total, sort descending, re-sort ties by business score
3. Take the top 10 into this batch’s writing queue, leave the rest in the pool without deleting
4. After two weeks, revisit impressions and rankings in GSC, fix the weight mix, and rank the next batch

After sorting, don’t skip one more step: confirm which page each of the 10 keywords lands on. Two high-scoring keywords with similar intent each getting their own article is self-competing — neither page ranks. Follow the keyword-to-URL mapping approach: one keyword one page, and merge overlaps into one deeper piece.

Don’t let a single dimension mislead you

A keyword with 5,000 monthly searches but a KD of 70 often has a worse ROI than one with 300 searches and a KD of 8: the former may sit silent for six months after writing, while the latter can reach the top three in two months and steadily deliver traffic. That’s the point of three-dimension scoring — it forces you to separate the “looks tempting” head terms from the “actually winnable” ones.

Intent mismatch is another hidden killer. Forcing an informational keyword into a conversion page gets flagged for intent mismatch and suppressed; users who come in and can’t find an answer bounce immediately — those two signals stack, and the page has a hard time recovering.

The three most common pitfalls

These practices will make the whole ranking system fail — check against them before every batch lands.

• Looking only at search volume, ignoring difficulty — writing for half a year and stuck on page three
• Guessing business scores, treating pure informational keywords as conversion pages — losing on both conversion and rankings
• Setting weights and never revisiting — the site stage changed but the mix didn’t
• Keyword list only grows, never prunes — old terms imported two years ago still muddy the overall judgment

One more practical detail: add a “last scored date” column. A score from six months ago reflects the competition and your domain authority at that time — it’s long expired. Re-scoring the top 50 keywords every quarter takes under two hours and keeps an entire batch of topics from resting on stale judgment.

When it comes down to execution, it’s really a fixed set of moves: first export all candidate keywords and score them on traffic, difficulty, and business; then pick a weight set by site stage, starting with defaults instead of fine-tuning up front; compute totals, sort descending, and put the top 10 into this month’s writing queue; then assign each of the 10 a landing URL and check for intent overlap; finally mark on the calendar the time to revisit data and fix weights 14 days later. After one round, you’ll have a solid sense of whether this mix works.