Building a Keyword Plan
Building a keyword plan means producing a document that can be checked, not a list of queries with volume and difficulty. Every column names its source, the unit of a row is a user task rather than a phrasing, and each forecast carries the assumptions behind it. Three months later you then see not "did it work" but which factor was wrong.
It rests on the core updates documentation and the performance report.
The unit of a plan is a task, not a query
One row = one user task = one target page. If 1,200 queries collapse into 60 tasks, the plan has 60 rows, not 1,200. The opposite approach — a page per phrasing — runs into the spam policies: scaled content abuse is many pages generated primarily to manipulate search rankings. How to group is worked through in the lesson on content strategy.
The columns and their sources
| Column | What is in it | Where it comes from |
|---|---|---|
| Task | What the person wants to do | Query grouping, verified against the results |
| Format | The page type the results expect | The composition of the top — see the lesson on intent |
| Target URL | One address, existing or new | A decision about site structure |
| Demand | Order of magnitude and seasonality | The approximate average plus the monthly breakdown |
| Expected value | Money per month | The arithmetic from the lesson on valuation |
| Effort | Person-days, data required | The team's estimate |
| Assumptions | CTR, relevance share, conversion | Your own data at the time of the estimate |
| Review date | When forecast meets outcome | Publication plus the crawl allowance |
"Volume" and "KD" are deliberately absent: the first is not a traffic forecast and the second does not exist at Google. "Demand" and "format" take over their job.
How long until the review
The review date is not invented. Crawling can take anywhere from a few days to a few weeks, after which the page still has to accumulate impressions. The working minimum is a month before the first review on impressions and a quarter before conclusions about clicks and conversions. Seasonal rows count differently — see the lesson on seasonality.
The entry bar
A topic enters the plan when we have something to add. The documentation frames that as questions about original information, reporting, research or analysis and about whether a piece adds substantial value when it draws on other sources. No contribution, no row — however high the demand.
Handling conflicts
- Two tasks point at one page. Either the tasks are really one, or there should be two pages. Settled before publishing, not after: conflicts in live data are what cannibalization finds.
- Two pages already exist for one task. They merge under the duplicate consolidation rules, leaving one target URL in the plan.
- The task exists, the page does not. That is a creation row, with effort and a date.
- The page exists but is sliding. That is an update row: content decay surfaces those.
The review loop
- Pull the outcome: queries and pages from the Search Console export, the overview from the pages report.
- Compare against the forecast in three separate steps: impressions, clicks, conversions. A gap at the first means indexing or format, at the second the snippet and position, at the third the page itself.
- Compare periods through wins and losses, and uncovered topics through coverage gaps.
- Update the assumptions: your CTR by position bucket moves as the site does.
- Rewrite the priorities. A plan is a living document, not a commitment made last quarter.
What a plan does not promise
Positions are not part of the commitment, and that is documented. Core updates are broad in nature and don't target specific sites or individual web pages, and the same page states plainly that there's no guarantee that changes you make to your website will result in noticeable impact in search results, and that positions in Google Search results aren't static or fixed in place.
Hence the shape of the plan: a portfolio of bets with recorded assumptions, not a schedule of positions. What can be promised is the work and the review, not a place in the top.
Mistakes
- A row per query instead of a row per task.
- A KD column as the basis for priority.
- A forecast with no recorded assumptions: later there is nothing to analyse.
- Reviewing a week after publication.
- Treating the plan as a commitment on positions.
- A quarterly plan nobody opened after it was approved.
The module in one line
The whole path: the user's task → the format the results expect → one page → expected value in money → priority against effort → comparison with the outcome. Not one step asks for faith in a metric without a source — and that is what separates a plan from a list.