Operations
Part of What keyword research operations means for content delivery
Seven checks in a keyword research launch review
A desk review of what a keyword research launch review should assess, separating vendor claims from independent evidence, with a seven-point checklist.
What to take away
- The commonest mistake is reviewing the tool rather than the launch: teams compare features and never check whether the keyword set, owners and records are fit to publish against.
- A launch review tests seven areas. The checkable ones are scope, data sources, records, prompt coverage and creative standards. Ownership and rollback need names and dates.
- Vendor claims about accuracy or AI coverage are claims, not evidence. This review ran no hands-on tests, so treat marketing pages as a starting point.
- The output is a signed record of what was checked, what was excluded and who owns each gap.
- Where a rule differs across the UK, name the nationinformation, advertising and search standards do not always move together.
Why a launch review is not a tool comparison
A keyword research launch review drifts into feature tables because features are easy to score. The harder question is whether the keyword set can carry a content programme for a year.
A review that starts with software misses the operational failure points: unowned keyword lists, no version history, briefs that cite search volume without a source. For the wider workflow, the operations and delivery guide sets out how the pieces fit before a launch is signed off.
This is a desk review. It assessed published vendor, standards and industry documentation; it did not test tools or audit live accounts, so the checks below are editorial judgements.
The seven checks, and what each one catches
Scope. State which markets, languages and nations the launch covers. England-only assumptions break when a campaign also runs in Scotland or Wales under a different code or regulator.
Seven keyword launch checks
- Scopemarkets, languages, nations
- Data sourcesnamed owner, refresh date
- Naming and recordsone convention
- Prompt coveragebuyer prompts, not just head terms
- Creative standardstitles, descriptions, ad copy
- Ownershipone accountable owner per cluster
Data sources. List every input: search volume estimates, internal search logs, sales queries, social listening. Each needs a named owner and a refresh date.
Naming and records. Keyword lists, briefs and approvals should follow one naming convention so records stay findable years later. BSI's BS 10008 standard for electronic information management is the reference many records teams cite when they want evidence that information can be trusted and traced.
Prompt coverage. AI assistants answer many queries directly, so a launch should include the prompts buyers type, not only head terms. Semrush's prompt research feature page introduces this emerging tool category; the page is a vendor description, so weigh it as such.
Creative standards. Titles, descriptions and ad copy sit under industry expectations. The IAB UK search principles for creative best practice give a benchmark for how search content should be written and presented.
Ownership. Every keyword cluster needs one accountable person, plus a deputy. Shared ownership is how clusters go stale.
Rollback. Agree in advance what happens if a launch underperforms: pause, revise or retire, and who decides.
What a reviewer should ask
Ask for the keyword list, the source of each volume figure and the last approval date. A screenshot without a date is unverified.
What a reviewer should not do
Do not score tools you have not used, and do not sign off a launch on vendor benchmarks alone. Record the gap, then assign someone to close it.
Where evidence ends and judgement starts
Vendor pages, standards documents and your own logs are different kinds of evidence. Mixing them in one spreadsheet hides which claims are checkable.
Three evidence types kept separate
Vendor claims
- Example
- Difficulty score
- Checkable
- No
- Owner
- Vendor
- Cost signal
- £400 a month
Independent standards
- Example
- BS 10008
- Checkable
- Yes
- Owner
- Records team
- Cost signal
- Free to cite
Internal observations
- Example
- Search logs
- Checkable
- Yes
- Owner
- Internal team
- Cost signal
- Already held
Keep them separate: vendor claims in one column, independent standards in another, internal observations in a third. For the people and sources behind each column, see the guide to keyword research team roles and data sources, which explains who should hold each input.
For example, a team paying £400 a month for a keyword platform may treat its difficulty score as fact, when a reviewer should label it a vendor estimate instead.
Glossary
- Launch review:
- a documented check of a keyword programme before it goes live.
- Prompt research:
- studying the questions people put to AI assistants.
- Rollback:
- the agreed action when a launch misses its targets.
Common questions
Does a launch review replace a tool trial?
No. A trial tests software behaviour; a launch review tests whether the keyword set, records and owners are ready. Run both, but keep the findings separate.
How long should the review take?
For a single-market launch, a half-day workshop plus a week of evidence gathering is a workable illustrative example. Multi-nation launches need longer because the rules differ.
What if the launch is already live?
Run the same seven checks and record gaps as remediation items with dates. Do not rewrite history to make the original launch look reviewed.
