Rules and ethics
How to trace keyword research data lineage before you trust a report
Trace every keyword metric back to its tool, export file and date stamp so reports can be verified before they justify attribution decisions.
How to trace keyword research data lineage before you trust a report
What to take away
- Keyword research data lineage means knowing the tool, the export file and the pull date behind every number in a report.
- A written source register turns scattered exports into a keyword research audit trail you can show a client or a finance team.
- Search volume and difficulty scores differ between providers, so a metric without a source and a date stamp cannot be verified.
- Trace the inputs before anyone uses the report to justify a budget or an attribution decision.
- Keep raw exports, not only the cleaned spreadsheet, so a reviewer can reproduce your figures.
Why keyword metrics drift between providers
Google Keyword Planner, Ahrefs, Semrush and Moz each estimate demand from their own panels. The same term can show 1,200 searches in one tool and 2,400 in another. Both are modelled estimates, refreshed on different schedules. The general definition appears in the Wikipedia article on keyword research. Record the provider name beside every metric you publish.
Build a keyword research source register
A source register records where each metric came from. Keep one row per export rather than one row per keyword, and store it beside the raw files. The register settles provenance questions in seconds when a report is challenged.
- Name the tool and the account or project used for the export.
- Record the filters, country and language settings exactly as applied in Google Ads.
- Save the raw CSV or API response before any cleaning.
- Note the download date and the colleague who pulled the file.
- Log every manual edit, merge or volume adjustment made later.
Keep a source tracking log for each client site you manage.
| Register field | Example entry | Why a reviewer needs it |
|---|---|---|
| Tool | Google Keyword Planner | Names the model behind the volume |
| Export date | 14 March 2025 | Demand estimates age within weeks |
| Geography | England | Regional demand differs from United Kingdom totals |
| Filters | Web search, English | Explains why the term list looks narrow |
| Contact | Content lead | Gives a name for follow-up questions |
Search Console supplies impressions and clicks from your own site, which is first-party evidence rather than a modelled estimate. Background on its features sits in the Wikipedia article on Google Search Console. Note the property, date range and filters used for those exports as well.
Where keyword research data provenance breaks down
- A colleague re-sorts a CSV and saves over the original file.
- A provider updates its volume model mid-quarter, so two exports disagree.
- A client hands over a keyword list with no tool name attached.
- A dashboard refreshes nightly while the report cites a monthly pull.
- A translated term list loses its original language setting.
Each of these breaks the link between the number and its origin. Write the fix into the register: version the file name, keep the original download, and note model changes when a provider announces them.
Example: tracing one keyword from export to slide
Take the term "air source heat pump grant". A Keyword Planner export for England in March 2025 showed 1,900 monthly searches. The cleaned client spreadsheet showed 2,400, matching an Ahrefs export pulled the same week. The slide showed only "2,400 searches", with no provider and no date. A reviewer could not tell which tool produced it.
Tracing one keyword
- March 2025Keyword Planner export 1,900 searches
- Same weekAhrefs export 2,400 searches
- After cleaningClient spreadsheet 2,400 searches
- Slide shown2,400 searches, no provider or date
Use a filename pattern that carries the tool, the geography and the date, such as keyword-planner_england-heat-pump-grant_2025-03-14.csv.
If a reviewer cannot find the file, the number is a claim rather than a measurement.
Check the audit trail before attribution sign-off
Keyword research reporting accuracy depends on the weakest input, not the best-looking chart. Pick three headline keywords and trace each one back to its export. Open the file, find the row, and confirm the number and the date match the report. A mismatch means the report is not ready.
Before you credit a channel with a sale, check the keyword set that model relies on. The keyword research attribution methods page compares first-touch, last-touch and data-driven models, and shows the evidence each one gives you. Ask which export an attribution model is reading before you accept its output.
Common questions
How far back should I keep keyword exports?
Keep the files behind any live report, plus the previous comparable period. Twelve months is a workable default for seasonal terms.
Which date matters more, the export date or the report date?
The export date. Search volumes are modelled and refreshed, so the pull date shows which version of the estimate you used. Record the report date as well for context.
Is a source register worth it for a one-off report?
Yes. Three fields, tool, date and geography, take a minute to fill in and remove most later guesswork. Add the file name once the export is saved.
