Skip to the indexed answers
Digital Marketing Index

usdigitalmarketingindex.com SEO

Why can average search position improve while rankings for business-critical queries fall?

Because average position is an average over every impression the site received, and the set of impressions changes from one period to the next. If a site starts appearing near the top for many new low-competition queries, those impressions pull the average toward 1, however the old queries are doing. The queries that bring revenue can each slip several places in the same period, and the site-wide number still improves. Search Console’s help describes the metric as one that “averages the position value for all impressions” (Search Console help on impressions, position, and clicks). It answers the question “where, on average, did our links appear across everything we appeared for?” It does not answer “how are we ranking for the terms we care about?”, and treating it as though it did hides exactly the loss a business would want to see first.

Changing query weights

An average over impressions is a weighted average, and the weights are the impression counts of each query. Two things move it.

Change Effect on the average
A query's position changes Moves the average in proportion to that query's impressions
A query's share of total impressions changes Moves the average even if no position changed

The second effect is the one that misleads. A query at position 30 that loses most of its impressions, through seasonality or a change in demand, stops dragging the average down, and the average improves with no ranking having improved. A query at position 2 that gains impressions does the same.

The help page shows the mechanics for a property across queries: it takes the topmost position for each query and averages them, giving the example of positions 2 and 3 averaging 2.5.

New long-tail impressions

Publishing more content is a reliable way to change the mix. New pages that answer specific questions can rank high for specific queries, each with a small number of searches. Added together they contribute many impressions at good positions.

An illustrative two-period query table. All figures are invented, chosen so the arithmetic can be followed.

Period 1:

Query group Impressions Average position Impressions x position
"boiler repair" (priority) 10,000 4 40,000
"boiler service cost" (priority) 8,000 5 40,000
Other queries 2,000 30 60,000
Total 20,000 140,000

Average position, period 1: 140,000 / 20,000 = 7.0

Period 2:

Query group Impressions Average position Impressions x position
"boiler repair" (priority) 9,000 7 63,000
"boiler service cost" (priority) 7,000 8 56,000
Other queries, including many new specific ones 24,000 3 72,000
Total 40,000 191,000

Average position, period 2: 191,000 / 40,000 = 4.8

What the two tables show together:

Measure Period 1 Period 2 Reading
Site average position 7.0 4.8 Improved
"boiler repair" 4 7 Fell 3 places
"boiler service cost" 5 8 Fell 3 places
Total impressions 20,000 40,000 Doubled

The headline metric improved by more than two positions while both priority queries fell by three. The new queries, 24,000 impressions at position 3, outweigh everything else.

Whether period 2 is better or worse for the business depends on what the priority queries are worth compared with the new ones. Positions do not settle that. Clicks and conversions by query group do.

Segmenting key terms

Protect the queries that matter from being averaged away.

  1. Define the priority set: the queries tied to revenue, agreed with the business. Keep it short enough to read.
  2. Fix it. The list changes by decision, with a date, not by drift.
  3. Report each priority query on its own row: impressions, clicks, CTR, position, for both periods.
  4. Report the remainder as groups: branded, other non-branded.
  5. Show the count of queries and impressions in each group, so shifts in mix are visible.

Practical points in Search Console:

Point Detail
Filter Use a query filter, or a regular expression, for the priority set
Same conditions Identical country, device, search type, and date length in both periods
Anonymized queries Queries searched very rarely are omitted from the query table, so long-tail totals are understated in query-level data
Low volumes A priority query with few impressions has an unstable average; note it
Page dimension Check which URL ranks for each priority query; a change of URL can accompany a change of position
Variants Decide whether close variants of a priority query are grouped with it

Position is also an average within a single query. The help page notes that “Position value is the average position for all searches”, and that an individual search can differ because of variables such as search history and location. A fall from 4 to 7 on one query is itself a summary of many searches.

Safer scorecards

Replace the single number with a small set that cannot move in opposite directions unnoticed.

Scorecard line Why
Clicks from priority queries The outcome that matters, for the terms that matter
Position of each priority query, listed No averaging across them
Count of priority queries in positions 1 to 3, 4 to 10, beyond 10 Shows distribution
Non-branded clicks, total Breadth of performance
Number of queries with impressions Shows footprint growth
Impressions by query group Makes mix shifts visible
Site average position Shown last, with the groups beside it, or omitted

A scorecard for the example:

Line Period 1 Period 2 Change
"boiler repair" position 4 7 Down 3
"boiler service cost" position 5 8 Down 3
Priority queries in positions 1 to 5 2 of 2 0 of 2 Down
Impressions, priority queries 18,000 16,000 Down
Impressions, other queries 2,000 24,000 Up
Site average position 7.0 4.8 Up, driven by other queries

Read this way, the report says something accurate: the site’s footprint grew, and its two most important rankings declined. Both are true and each needs its own response.

Rules for anyone presenting position data:

  • Never report average position without impressions next to it.
  • Never compare average position across periods without checking whether the query mix changed.
  • Use a fixed query set for trend lines.
  • Treat a site-wide average as a description of the mix, not as a score.
  • When the average and the priority set disagree, believe the priority set.

The same caution applies in reverse. A worsening average, caused by new impressions at low positions for queries the site has only begun to appear for, can accompany steady or rising priority rankings.

A related puzzle, impressions rising while click-through rate falls, has the same root in changing mix and is worked through in the guide to how to investigate falling organic CTR when search impressions are rising.

The summary: average position describes whatever the site appeared for, weighted by impressions, so a change in what it appears for can move the average in the opposite direction to the rankings that matter.

Link trail

Reader comments on this answer

Add your comment to this answer

Your email address will not be published. Required fields are marked *