Average position is the most quoted number in Google Search Console and the most misread. It is not "where you rank." It is the mean of the highest position your site occupied across the impressions Google recorded — and each of those words is doing work.
Three things it is not
It is not a ranking. It is an average of many rankings, across devices, locations, personalization states, and time. A query where you sit at position 3 for half the impressions and position 20 for the other half reports as 11.5, a position you have literally never held.
It is not recorded when you do not appear. If your page drops out of the index for a week, those days contribute nothing to the average — they are absent, not zero. This produces the single most common misreading in SEO reporting: average position improves while traffic falls. Losing visibility on the queries where you ranked poorly raises the mean of what remains. The metric went up because you lost coverage.
It is not the position of the page. It is the position of your best-ranking URL for that query. If a stronger page cannibalizes a weaker one, the reported position can improve while the page you were actually working on falls out entirely.
The skew problem
Position data is not normally distributed. It is bounded at 1, has a long right tail, and clusters heavily in the ranges where you have real presence. A mean over that shape is pulled hard by the tail.
The fix is to stop using the mean. Two alternatives are more honest:
- Bucketed share of queries. Count how many queries sit in positions 1–3, 4–10, 11–20, and 21+. Track the movement between buckets. This survives the skew and makes progress legible: "we moved 34 queries from the 11–20 bucket into 4–10 this quarter" is a real sentence about real change.
- Median position, weighted by impressions. Less familiar to stakeholders but far more stable month to month, and much harder to accidentally improve by losing coverage.
Pair it with impressions, always
Average position alone is uninterpretable. Average position plus impressions is diagnostic, because the two together tell you which of four things happened:
| Position | Impressions | Likely story |
|---|---|---|
| Up | Up | Genuine gain — more visibility, better placement |
| Up | Down | Lost coverage on weak queries; check for deindexing |
| Down | Up | Ranking for a broader, more competitive query set |
| Down | Down | Real decline — investigate immediately |
The second row is the trap. It looks like a win in a dashboard that tracks position alone, and it is very often the early signal of an indexing problem.
The date dimension hides a rounding artifact
Search Console rounds average position to one decimal place and computes it over whatever date range you select. A 28-day window and a 7-day window over the same period can tell different stories, not because anything changed but because the query mix inside each window differs. When you compare periods, hold the window length constant and compare like to like — a 28-day block against the preceding 28-day block, not against "last month," which varies in length.
What to do with the number
Average position is useful for exactly two jobs.
The first is anomaly detection at the property level. A sharp, sustained move in site-wide average position is worth investigating, because whatever caused it was large enough to overwhelm the noise. Treat it as a smoke alarm, not a diagnosis.
The second is relative comparison within a fixed query set. Define a set of queries you care about, hold the set constant, and watch its average position over time. Because the set does not change, the coverage problem disappears and the number starts meaning something.
Everything else — client reporting, forecasting, proving the value of a content program — is better served by clicks, impressions, and bucketed query counts. Those metrics are harder to move accidentally, which is precisely what makes them worth reporting.
The reporting habit
If a monthly SEO report leads with average position, it is optimizing for a number that can improve while the business gets worse. Lead with non-brand clicks. Show impressions alongside. Put bucketed position movement in as the leading indicator, because rankings move before traffic does. Keep average position in an appendix if a stakeholder is attached to it, with a one-line note about what it measures.
That note is the whole point. A metric everyone misunderstands is worse than a metric nobody looks at.