Start with intent, not volume

Search volume tells you how often a phrase is typed. It tells you nothing about whether the person typing it is ready to buy, comparing options, researching a problem, or trying to navigate to a specific page. Before a keyword earns a place on a content roadmap, classify it against the four standard intent categories: informational (learning about a problem or topic), navigational (looking for a specific brand or page), commercial investigation (comparing solutions or vendors), and transactional (ready to act, request a quote, or buy). A 10,000-a-month informational keyword and a 200-a-month transactional keyword are not comparable line items on the same spreadsheet; they belong to different stages of a funnel and different page types entirely.

Demand mapping: connecting keywords to the buyer journey

Demand mapping means laying every keyword cluster against the stage of the buyer journey it represents, then checking that stage against the pages you actually have. Most sites over-invest in top-of-funnel informational content and under-invest in the commercial investigation stage, where a well-built comparison, pricing, or “best X for Y” page can capture a searcher who is genuinely close to a purchase decision. The gap is usually visible the moment you plot your existing content against the four intent categories: a long blog archive of “what is” articles and almost nothing that helps someone choose between options.

Keyword-to-page models

Every keyword cluster should map to exactly one page, and every page should have exactly one primary cluster it targets. This is the single most common structural mistake in keyword research: treating keywords as a flat list to be sprinkled across whichever page seems related, rather than as a cluster-to-URL model that determines information architecture. When two pages compete for the same cluster, both usually rank worse than one consolidated page would, because internal links, relevance signals and content depth get split rather than concentrated.

Clustering by SERP overlap

The most reliable way to group keywords into a single page’s cluster is not by looking at the words themselves but by checking whether Google already returns the same or substantially overlapping set of URLs for each query. If eight or more of the top ten results overlap between two keyword variants, treat them as the same cluster and the same page. If overlap is low, Google has already decided these are different intents and you should build separate pages rather than trying to force one page to rank for both. This method is more reliable than semantic grouping tools because it reflects what the ranking algorithm has already concluded, rather than a guess based on surface similarity of the terms.

Difficulty versus authority calibration

Animated view: volume versus difficulty, sized by revenue potential

Question mining

People Also Ask, community forums such as Reddit, Quora threads, and your own sales and support team’s records of prospect questions are all richer sources of real language than keyword tools alone, because they capture the exact phrasing and follow-up concerns a buyer has rather than a normalised search term. Mining these sources for recurring questions, then mapping each one to the page or cluster it belongs to, both improves relevance for AI-driven answer engines and surfaces long-tail transactional questions that volume-based tools under-report because individual phrasings are too low-volume to register.

Seasonality

Pulling twelve months of trend data for every priority keyword prevents two common errors: mistaking a seasonal spike for sustained demand growth, and under-resourcing a page ahead of its predictable peak. A B2B keyword tied to a fiscal-year budgeting cycle, or a retail keyword tied to a holiday period, needs its content and technical work finished well before the seasonal window opens, not during it.

Traffic value estimation

Assign every priority cluster an estimated traffic value by multiplying realistic achievable click-through rate at a target ranking position by monthly search volume, then by your actual site-wide conversion rate for that intent category, then by average order or deal value. This produces a rough monthly revenue estimate per keyword cluster that is far more useful for prioritisation than volume or difficulty alone, because it directly answers the question a budget holder actually asks: what is this cluster worth if we rank for it.

A prioritisation scoring formula

A workable formula: score = (estimated monthly value × intent weight) ÷ (calibrated difficulty × production effort). Intent weight is a simple multiplier (transactional = 1.5, commercial = 1.2, navigational = 1.0, informational = 0.7) that pulls high-value bottom-funnel clusters up the list even when their raw volume is modest. Production effort accounts for whether the page exists and needs updating versus needs building from scratch, including any research, data, or design work required. Run every cluster through the same formula so the roadmap is defensible and comparable, rather than driven by whichever keyword feels most exciting that week.

Common mistakes

The most frequent error is chasing headline volume on purely informational terms while starved of budget for the commercial and transactional pages that actually convert. A close second is building a new page for every keyword variant instead of clustering, which fragments authority and produces a bloated, cannibalising site structure. A third is scoring difficulty from a tool without checking who is actually in the top ten, which either scares teams away from winnable clusters or sends them at competitors far above their current authority. A fourth is ignoring seasonality and building time-sensitive content too late to catch the relevant search window.

Worked example: a local service business

Consider a mid-sized HVAC company. Seed research turns up “furnace repair [city]” (transactional, moderate volume, medium difficulty), “furnace not heating troubleshooting” (informational, higher volume, lower difficulty), “furnace repair cost” (commercial investigation, moderate volume, moderate difficulty) and “best furnace brands” (commercial, decent volume, high difficulty dominated by national review sites). Applying the formula: the transactional city-level page scores highest despite modest volume, because intent weight and realistic conversion rate outweigh its lower search count. The troubleshooting article scores lower for direct revenue but is worth building as a lead-generation and internal-linking asset that funnels readers to the repair and cost pages. The high-difficulty “best furnace brands” cluster is deprioritised initially, not abandoned, since a smaller local business is unlikely to out-rank national review sites in the short term without a distinct content angle. This produces a roadmap of three pages built in a deliberate order, rather than a flat list of forty keywords with no build sequence.

Frequently asked questions

Refreshing the map every quarter

A keyword map is not a document you build once and file away. Search demand shifts: new products enter the category, competitors publish, the SERP layout changes, and a query that was informational last year now returns a comparison grid and three ads. A map that is never revisited quietly stops describing reality, and the content calendar built on it starts producing pages nobody needed.

Set a quarterly review with four checks. First, pull the queries your site now receives impressions for and look for terms that are not on the map at all — Google is telling you where it thinks you belong, and those discovered queries are often the cheapest wins available. Second, re-scrape the SERPs for your priority clusters and note any that changed shape; a cluster that grew a video carousel or a shopping row has less organic real estate than it did, which lowers its effective value. Third, re-run difficulty against your current authority, because a term that was out of reach at the start of the year may now be realistic after two quarters of link acquisition. Fourth, check for pages that have drifted — a post that has been edited five times often no longer matches the intent it was mapped to.

The output of each review should be small: a handful of new targets promoted into the build queue, a handful of pages flagged for refresh rather than replacement, and any clusters demoted because the economics no longer work. Resist the temptation to rebuild the whole map. Stability in the mapping is what lets internal linking, topical depth and authority accumulate on the same URLs over time, and that accumulation is most of where the compounding comes from.

One last discipline: record the assumption behind each priority decision. When you write down that a term was prioritised because it was expected to be worth roughly a certain amount per month at a modest conversion rate, you can check that assumption later against real data. Most teams never do this, which is why their keyword strategies feel like opinion. The teams that do it develop a calibrated sense of what their own demand is worth — and that sense, more than any tool, is what separates keyword research that generates revenue from keyword research that generates spreadsheets.