Every number in RankDots is one of three kinds: measured, observed or estimated. Positions, impressions and clicks from your Google Search Console are measured. Google Top 20 and SERP features are observed - snapshots of a results page. Everything else - volume, difficulty, weak spots, potential traffic, estimated impact - is estimated. Knowing which kind you are looking at tells you how hard you can lean on it.
Why we are telling you this
Because every serious SEO asks, and a straight answer is worth more than a confident one. Tools that present modelled numbers as facts set their users up twice: once when the estimate is treated as a measurement and lands in a plan it cannot support, and again when the real, measured numbers arrive and disagree. We would rather you knew the difference from the start.
The three kinds
| Number | Kind | Ground |
| Positions / Average position | Measured | Your Google Search Console |
| Impressions and Clicks | Measured | Your Google Search Console |
| Click-through rate | Measured | Derived from your clicks and impressions |
| Google Top 20 | Observed | Google’s results page, at check time |
| SERP features | Observed | Google’s results page, at check time |
| Search volume | Estimated | Google Keyword Planner, for your search engine and country |
| Trend | Estimated | Monthly search data behind the volume figure |
| Cost per click | Estimated | Modelled from Google advertising data |
| Keyword / Page / Topic difficulty | Estimated | Calculated by RankDots from the results pages |
| Weak spots | Estimated | RankDots’ assessment, relative to your site |
| Easy-to-rank search volume | Estimated | Sum of volumes behind that assessment |
| Potential traffic | Estimated | Built from volume, CTR by position, competitive strength, current traffic |
| Estimated impact | Estimated | Built on your Search Console data — but still an estimate |
Measured: counted by Google, about your site
Positions, impressions, clicks - and the click-through rate derived from them - are facts about your site, counted by Google and delivered through Search Console once you connect it. They are as close to ground truth as search data gets. Even here, be aware of the source’s own behaviour: Search Console data arrives with a delay, and some rare queries are anonymised.
Observed: true at check time
Google Top 20 and SERP features are read directly from Google’s results page for a keyword, in your country and language. Nothing is modelled — but nothing is permanent either. They are photographs, accurate on the day they were taken, and results pages move.
Estimated: modelled, calculated, assessed
Search volume is Google’s own estimate, not ours: it comes from Google Keyword Planner - the advertising tool - for your project’s search engine and country. Google describes the figure as the average number of times people search for a keyword and similar terms, averaged over a twelve-month period. Both halves of that description matter. It is an average, so a seasonal keyword’s busiest month looks nothing like its figure. And it covers similar terms together, because Planner folds close variants into one number - which is why different phrasings of the same request can show identical volumes on the Keywords screen. Tools built on other data sources model volume differently, which is why numbers differ between tools (Why our numbers differ from other tools [link is coming] explains the mechanics). Difficulty is RankDots’ calculation from the results pages. Weak spots is an assessment relative to your site. And potential traffic and estimated impact are estimates built from the layers below them.
What this means for you
Decide with estimates; verify with measurements. Estimates exist to compare options - which cluster, which page, which recommendation first. Measurements exist to tell you what actually happened after you acted. Used that way round, both kinds serve you; swapped, both mislead you.
When a measured number and an estimated number disagree, the measured one is right. If your Search Console shows different traffic than a modelled volume implied, that is not a bug in either number — it is the difference in kind doing exactly what it does. Adjust your expectations to the measurement, not the other way around.
Expect observations to age. A Google Top 20 from project creation describes that day’s results. If a decision hangs on who ranks today, look at today’s results page.
Start every “this number looks wrong” question here. The first diagnostic is always: which kind is it? Most apparent contradictions dissolve once the kinds are separated. For the ones that do not, The numbers look wrong [LINK PENDING: numbers-look-wrong] walks the remaining cases.
What next
- Why our numbers differ from other tools [link is coming]
- What is potential traffic?