Careers
Keyword research, minus the hype in 2027
Keyword research in 2027 maps customer language to useful destinations by preserving each source unit, market, date, limits, task, business fit, and validation plan.
What to take away
- Answer the business question firstname the customer, the market, the action, and the page that would serve it.
- Record source, unit, market, date, settings, and limits beside every number, because a trend index, a paid estimate, and an observed query count are not interchangeable.
- Cluster by destination rather than by wording, then score with visible weights and a stated cutoff.
- Replace estimates with observed queries and qualified outcomes once the page is live.
Keyword research is the record of which customer tasks you will serve, on which page, and why you rejected the rest. Volume is one input, and the least trustworthy one. The method below follows a single seed task to a post-release check, with fields and thresholds written down so a colleague can audit the call.
Decide the task before the query
Write the decision first. Name the customer, market, product, action, value, and exclusions. Then say what the work must change: an existing page, unmet demand, a launch, a market localization, a catalog structure, or support load.
A high-volume term can be unaffordable, off-offer, or held by a different intent. A low-volume task wins when it matches a precise service, an urgent problem, or a repeated support question. Keep business value beside the search estimate at all times.
Keep a source-aware keyword map
One row per query group, with the fields that make the number readable. Without the unit, a value cannot be compared with anything.
| Field | Example entry | Why it stays |
|---|---|---|
| Source | Google Search Console, property sc-domain:example.com | Defines the observed population |
| Unit | Clicks and impressions, 16 months | Counts, not a market estimate |
| Market | Canada, English, desktop and mobile | Locale changes the meaning |
| Date | Extracted 3 March 2027 | Reports and models shift |
| Limits | Brand queries included; low-volume queries hidden | States the blind spots |
Example: one seed through to a live page
- Seeda support ticket asks whether the software handles provincial sales tax. Record the task, not the noun.
- First-party checkthe phrase appears in the site search report, and no page answers it.
- Demand checka Trends comparison of three phrasings shows a flat line, not a spike.
- Paid contextKeyword Planner returns a range and a competition label for the same location.
- Clustertwo phrasings share one destination, a comparison table plus a rules summary.
- Brief and shipaudience, task, subquestions, exclusions, internal path, conversion step, review date.
- Review at day 30 and day 90index state, impressions, clicks, then demo requests.
Score priority with visible weights
Numbers force a conversation. The weights below are ours, not an industry standard, and they should move with the portfolio.
| Criterion | Weight | Note |
|---|---|---|
| Customer value | 30 | Value of the action, not traffic |
| Fit with the offer | 25 | Can we serve and support it |
| Evidence strength | 15 | First-party beats modeled |
| Competitive gap | 15 | Result types and coverage |
| Effort | 10 | Hours to a complete page |
| Risk | 5 | Legal, brand, maintenance |
Act on clusters scoring 65 or above. Record anything between 40 and 64 and revisit it quarterly. Leave the rest out unless support or compliance keeps it alive.
Write the brief before you write the page
A brief names the audience, the task, the primary and supporting language, the page role, the original value only we can add, the subquestions, exclusions, internal path, conversion step, and review date.
Use customer wording in titles and headings. Do not set a repetition target. A reader arriving directly must still get a complete answer.
Serve markets and languages separately
Research each market rather than translating one list. Terms, units, currency, regulations, and search interfaces differ. Track language and region as separate fields. For bilingual Canadian work, annotate each language version as described in this explanation of hreflang attributes.
Demand without a serviceable customer experience is not an opportunity. Where analytics collect query data, follow the consent rules of that market; Canada's federal framework is set out in the Office of the Privacy Commissioner guidance on PIPEDA.
Check the page after release
Record the release date, affected URLs, intended query group, and a guardrail such as a target for qualified enquiries. Watch crawl and index state first, then queries, impressions, clicks, click-through rate, and landing behavior, over a window long enough for the page to settle.
Unexpected queries are the useful signal. They reveal demand you missed or show the page drifting off task. Tie the result to a sales-stage measure rather than sessions, since qualified outcomes are what lead generation work is judged on.
Update the map. Merge groups that converge on one page, split those that prove distinct, retire unsupported ideas, and add language from sales calls.
Keep a cadence you will hold
Keyword work sits inside the annual plan, as set out in our honest guide to SEO strategy, and the cadence below keeps it there.
- Weeklylog new customer questions, product changes, and sudden market events.
- Monthlyreview priority clusters against outcomes and seasonality, then close gaps.
- Quarterlyrefresh markets, competitors, result types, page maps, and weights.
- Before a launch or migrationre-run targeted research and protect high-value demand.
Common questions
How many keywords should one page target?
Is search volume still worth checking?
How long before I judge a new page?
What ends a keyword project?
In this guide
- Common keyword research questions: working answers and a decision recordCommon keyword research questions about volume, intent, page scope, low volume terms, competitors, forecasts, and measurement, with working answers and context.
- How to improve keyword research with a focused business plan: the parts worth your timeHow to improve keyword research in 2027: use Search Console, Keyword Planner, customer tasks, page mapping, and outcome reviews to focus SEO effort.
- Nine keyword research patterns worth adapting, from services to videoKeyword research examples in 2027 offer nine hypothetical patterns across services, ecommerce, local markets, product launches, support, video, and consolidation.
- Why does false precision quietly ruin most keyword research?Keyword research mistakes in 2027 include false precision, mixed markets, assumed intent, volume bias, copied exports, lost raw data, and no refresh.
- Keyword research review checklist: states, evidence, owners, and a completion gateA keyword research checklist for scoring scope, sources, cleaning, clustering, priority, and completion, with states, evidence, owners, and a finished gate.
- A framework for keyword research benchmarksKeyword research benchmarks assess evidence coverage, method completeness, cluster consistency, page-map decisions, demand ranges, and observed performance.
- Keyword research tools compared on source, freshness, and export limitsKeyword research tools in 2027 should be compared by source, unit, market, freshness, coverage, limits, exports, API, security, cost, workflow, and decision value.
- Solid keyword research follows a practical guide, not a single toolKeyword research best practices in 2027 start with decisions, combine customer and search evidence, preserve source units, map tasks to pages, and validate outcomes.
- Should you hire a keyword research agency without a controlled pilot first?Best keyword research agencies in 2027 should be compared through one brief, assigned researchers, source transparency, market fit, security, price, and ownership.
- Keyword research trends: eight signals worth testingKeyword research trends worth testing: AI exploration, AI-answer query data, first-party language, more surfaces, local markets, low-volume tasks, outcomes.
