Skip to main content

The Statistics Behind Google's Search Average Position

by Daniel James Stoker

Last Updated on October 22, 2024

What is Average Position in Google Search Console

Google defines average position as “…the topmost position occupied by a link to your property or page in search results, averaged across all queries in which your property appeared.” Google makes available this metric of Average position in the Google Search Console (GSC) dashboard, alongside metrics of Total clicks, Total impressions, and Average CTR (click-through rate). By default, the Average position is displayed for all the pages and queries driving search traffic to a website, as seen below.

google search console dashboard metrics average position
Figure 1 – Google Search Console dashboard for Performance on Search results, showing Average position alongside other metrics of Total clicks, Total impressions, and Average CTR.

This definition addresses two of the four types of search average position that Google makes available to filter on and view in the Google Search Console dashboard – the two types of average position that aggregate on multiple queries.

How Google Calculates Average Position

Google's documentation details that when a user’s search query produces the same website multiple times in a Google search result, only the website's topmost position is used in its average position calculation.

For example, if a query produces a search result where a website appears in positions 2, 5, and 9, only the topmost position, in this case, 2, will be used in the average position calculation. Similarly, if another query places the same website in positions 4 and 8, again only the topmost position, which is 4, will be used.

The average position for these two topmost positions would be (2+4)/2 = 3. This example assumes one impression per query, but when a query produces multiple impressions, Google factors this frequency of visibility into its calculations for average position through a weighted average position.

The weighted average position is determined by multiplying each average position by its number of impressions (the number of times it appeared in Google search results), summing these terms, and then dividing by the total number of impressions for all average positions.

This approach gives more importance to the topmost positions that appeared more frequently in Google search results, ensuring they have greater representation in the final weighted average position, compared to those with fewer impressions.

If we assume that the previous example of a topmost position of 2 and a topmost position of 4 were both surfaced through one impression each, the equation for weighted average position is calculated as:

google search weighted average position formula for two queries with one impression each

With both topmost positions getting an equal number of impressions, each query has equal representation in the final calculation and the weighted average position is the same as the average position calculated previously.

However, if the topmost position of 2 had 200 impressions, and the topmost position of 4 had 50 impressions, we would calculate the weighted average position to be:

google search weighted average position formula for two queries with different number of impressions

This weighted average position is lower than the average position, evident from the observation that the first topmost position now has more impressions than the second. This gives the first topmost position more representation in the final calculated weighted average position.

While it’s important to understand how Google calculates weighted average position, moving forward, we will keep impressions equal among topmost positions (which now we'll just refer to as positions) so that our weighted average positions are equal to our average positions. This does not alter the types of underlying statistical scenarios that can emerge for different average position types and their variations over time. A weighted average position only affects how much each average position contributes to these statistical scenarios.

This will keep our analysis simpler, but it should be noted that practically all websites have a wide variance in their number of impressions per query and resulting topmost positions. It should also be noted that while Google presents the metric “Average position” in the Google Search Console dashboard, this is indeed a weighted average position that takes impressions into account.

How Google Search Console Filters Views for Average Position

Filtering in Google Search Console dashboard allows viewing the average position metric by a single URL, by a single query, or simultaneously by both single URL and single query, over a specified time range. It is not immediately apparent to users that each of these different filtered views of average position has distinct statistical characteristics and limitations based on their type of average position.

These varying types arise from the statistical mapping of how the underlying queries drive search traffic to webpages. The four types of Google search average position are defined as:

  1. Site Average Position (SAP): average position viewed for all pages and queries driving traffic to the website.
  2. Page Average Position (PAP): average position viewed for a single page and all queries driving traffic to this single page.
  3. Query Average Position (QAP): average position viewed for a single query and all pages this single query is driving traffic to.
  4. Query Page Average Position (QPAP): average position viewed for a single page and a single query driving traffic to this single page.

While Google doesn’t categorize value changes for any type of average position as good, bad, or neutral for performance, it has become common practice for users of the dashboard to do so. Traditionally, a decrease in the value of any average position type is considered good for performance, an increase is seen as bad, and no change is viewed as neutral. However, this performance categorization may be oversimplified and, in some cases, incorrect for some types of average position.

To fully explore and understand Google Search average position, we will thoroughly examine the statistics behind the four main average position types: Site (Overall) Average Position, Page (URL) Average Position, Query Average Position, and Query Page Average Position. For each of these types, we will answer the question: Can decreasing, increasing, or same value for [Type] Average Position be statistically categorized as good, bad, or neutral performance, respectively?

Site Average Position

The Site Average Position (SAP) is calculated for all pages and queries across the entire website for a specified time range. As mentioned previously, this is the default view in the Google Search Console dashboard. It is the only average position type with a many-to-many relationship, as shown in Figure 2 below. Each page can receive traffic from one-to-many queries over time, and a single query can drive traffic to one or many pages during that same period.

google search site average position many to many pages to queries relationships diagram
Figure 2 – Diagram of Google search Site Average Position many-to-many relationship between multiple pages and multiple queries. All relationships are calculated into the Site Average Position.

Because it captures all pages and queries for a website in its average position calculation, Site Average Position is the most limited of the average position types in representing meaningful descriptions of how individual pages or queries are performing. The specific performance details of individual pages and queries are lost in the overall aggregation of Site Average Position. However, these performance details can be explored at more granular levels through the other average position types.

Before we move on, let’s consider whether a decreasing, increasing, or unchanged value for Site Average Position statistically equates to good, bad, or neutral performance, respectively.

The short answer: the aggregation of all pages and queries into the calculated Site Average Position should make it clear that there are too many individual changes occurring to generalize this level of categorization.

The long answer: would require us to explore the full statistical scenarios behind Site Average Position. We will instead examine Page Average Position and its statistical scenarios, which will be simpler, and then use this information to revisit and fully answer this question for Site Average Position.

Page Average Position

The Page Average Position (PAP) is calculated for a single page (URL) and considers all queries driving traffic to that page over a given time range. This is the first type of average position that dives deeper into the granular behavior underlying Site Average Position and is a one-to-many relationship.

google search page average position one to many pages to queries relationships diagram
Figure 3 – Diagram of Google search Page Average Position one-to-many relationship between a single page and the multiple queries driving traffic to its URL.

To understand how Page Average Position can vary across different pages, a simple example is shown in Table 1 below. This example presents three pages on a website, each with their own three unique queries driving traffic to their respective URLs for a given time period. The positions differ between pages and between each of their respective queries, resulting in varying calculated Page Average Positions for each page.

Table 1 – Three webpages each receiving traffic from three different queries with varying positions.
PageFirst Query PositionSecond Query PositionThird Query PositionPage Average Position
Page 13513
Page 24596
Page 314221818

Based on the individual position of each query for each page, we can calculate the Page Average Positions for Page 1, Page 2, and Page 3 as 3, 6, and 18, respectively. If these were the only pages and queries driving traffic to the site, the Site Average Position would be the average of these three Page Average Positions, which would be (3 + 6 + 18) / 3 = 9. However, websites typically have far more pages and queries driving traffic to their content. Also, we are still treating impressions as equal among all queries, so weighted averages do not need to be calculated, but typically they do.

But as seen, even from this simple example, several statistical stories about the traffic to these pages are lost when we look only at the Site Average Position, which in this case is 9. For example, we lose visibility on the lower value Page Average Position of Page 1, which in this example is a relatively good performance story, and the higher value Page Average Position of Page 3, which is less favorable.

Websites typically have hundreds to thousands of pages, each with dozens to hundreds of queries driving traffic to their content each month. The larger the site in terms of content and traffic, the less representative and meaningful the Site Average Position becomes for analyzing the underlying statistical scenarios. This issue is further compounded when reviewing how Site Average Position has changed between two time periods, as we are unable to track where average position changes are taking place between pages and queries.

Next, we will explore the statistical scenarios that can exist behind the Page Average Position. This analysis will help us determine if changes in value of Page Average Position, or Site Average Position, of increasing, decreasing, or remaining the same, can be statistically categorized as bad, good, or neutral for performance, respectively.

Page Average Position Statistical Scenarios

In Table 2 below, we have a single page (URL) with four different queries driving traffic to it. We include Query Page Average Positions (which will be discussed later), which calculates the average position for the combination of a single page and a single query.

For example, for page 1 and query 1, the Query Page Average Position is 4, and for page 1 and query 2, the Query Page Average Position is 2. If these four queries were driving all the traffic to this page, taking the average of all the Query Page Average Positions would give us a Page Average Position of 4 (again, we are treating impressions as equal among queries).

Table 2 – A single page with four different queries driving traffic to its respective URL.
Page (URL)QueryQuery Page Average Position
page 1query 14
page 1query 22
page 1query 34
page 1query 46

Now if the Page Average Position changes, say from 4 to 5 or from 4 to 3, can we confidently say that change is bad or good for the performance of this page, respectively? Remember, this is at the page level, not the site level for average position type. Can we classify changes (or lack thereof) in the Page Average Position as good, bad, or neutral for performance? Let’s take a closer look.

There are four queries driving traffic to this single page. We’ll need to consider the average position changes for each query and how these can vary in combination. Then for each unique statistical case, we'll need to identify its performance categorization.

For example, if the positions of all four queries driving traffic to this single page have decreased in value (all QPAPs have decreased), this is a good performance story for each query, so we can statistically identify this scenario as good for performance for the Page Average Position. If the majority of queries have decreased in average position value while a minority have increased, we likewise identify this as a good performance outcome, as we are only considering the quantitative changes for each page-query position, treating all queries equally in terms of impressions and value.

It should be noted, we must also include scenarios where half of the queries are increasing and/or half are decreasing in their average position values, as the scenario of exactly half increasing and half decreasing would result in a quantitative nullification, which we identify as neutral for performance.

Continuing this logical analysis, we find that there are 15 major quantitative statistical scenarios that can occur for the underlying queries for a single page (for the Page Average Position), in consideration of their individual average position value changes between a pre and a post time period.

Table 3 – A list of Page Average Position statistical scenarios for combinations of individual average position value changes for all queries driving traffic to a single page.
PAGE AVERAGE POSITION STATISTICAL SCENARIOS FOR EXISTING QUERIES
no queries positions change
all queries positions decreased
all queries positions increased
majority of queries positions decreased, minority of queries positions increased
minority of queries positions decreased, majority of queries positions increased
half of queries positions decreased, minority of queries positions increased
minority of queries positions decreased, half of queries positions increased
minority of queries positions decreased, minority of queries positions increased
majority of queries positions decreased, no queries positions increased
no queries positions decreased, majority of queries positions increased
half of queries positions decreased, no queries positions increased
no queries positions decreased, half of queries positions increased
minority of queries positions decreased, no queries positions increased
no queries positions decreased, minority of queries positions increased
half of queries positions decreased, half of queries positions increased

It’s important to differentiate each of these scenarios because each carries its own statistical categorization of good, bad, or neutral for performance. Looking purely at these changes from a quantitative standpoint, where all queries hold the same impressions and value, the categorization mapping would be as follows.

Table 4 – A list of Page Average Position statistical scenarios for combinations of individual average position value changes for all queries driving traffic to a single page with performance category added.
PAGE AVERAGE POSITION STATISTICAL SCENARIOS FOR EXISTING QUERIESCATEGORY
no queries positions changeNeutral
all queries positions decreasedGood
all queries positions increasedBad
majority of queries positions decreased, minority of queries positions increasedGood
minority of queries positions decreased, majority of queries positions increasedBad
half of queries positions decreased, minority of queries positions increasedGood
minority of queries positions decreased, half of queries positions increasedBad
minority of queries positions decreased, minority of queries positions increasedNeutral
majority of queries positions decreased, no queries positions increasedGood
no queries positions decreased, majority of queries positions increasedBad
half of queries positions decreased, no queries positions increasedGood
no queries positions decreased, half of queries positions increasedBad
minority of queries positions decreased, no queries positions increasedGood
no queries positions decreased, minority of queries positions increasedBad
half of queries positions decreased, half of queries positions increasedNeutral

Again, this analysis focuses solely on the quantitative comparison of these queries, without considering their qualitative value, which can be subjective. We have mapped out 15 statistical scenarios for the combinations of changes in average position values of the underlying queries for a single URL between two time periods. Our statistical categorization defines decreased search position as good for performance, increased search position as bad, and unchanged search position as neutral for a single page-query (QPAP). And we have applied this categorization to all scenarios, however, we have yet to identify the scenarios in which the Page Average Position can statistically increase, decrease, and/or remain the same.

In order to identify the possible value changes for PAP for each statistical scenario, we can create specific examples of how the search positions of each underlying query can change between the two time periods for this single URL, as shown in Table 5 below. For instance, in a scenario where the majority of query positions have decreased and a minority have increased, we can map out the following page-query average position changes:

Table 5 – Two examples of Query Page Average Position changes for a single page for the scenario of majority queries positions decreasing and minority queries positions increasing
Pre Query Page Average PositionsPost Query Page Average Positions 1Post Query Page Average Position 2
211
433
237
655

Both examples illustrate a post period where the majority of query positions decreased and the minority increased. However, the Page Average Position for each example varies in directional change. In the first post-period example, the average position changes are from 2 to 1, 4 to 3, 2 to 3, and 6 to 5, resulting in a change in the Page Average Position from 4 in the pre period to 3.5 in this post period - a decrease in PAP. In the second post period example, the average position changes are from 2 to 1, 4 to 3, 2 to 7, and 6 to 5, resulting in a change in the Page Average Position from 4 in the pre period to 4.5 in this post period - an increase in the PAP.

While each example statistically represents a positive quantitative performance scenario, with the majority of query positions decreasing and the minority increasing, one example shows a decrease in Page Average Position, while the other shows an increase in Page Average Position. This granular statistical evaluation of average position changes for each query supporting a single page reveals that even at this level, changes in PAP value cannot be statistically mapped to good, bad, or neutral performance.

It should be noted that for this scenario, these average position changes could also occur in a way that results in the Page Average Position remaining the same value.

Carrying out this analysis for all statistical scenarios, we can identify whether the value of Page Average Position can increase, decrease, or remain the same for each of the 15 statistical scenarios. The findings are shown in Table 6 below.

Table 6 – A list of Page Average Position statistical scenarios for combinations of individual average position value changes for all queries driving traffic to a single page with performance category and PAP value change possibilities added.
PAGE AVERAGE POSITION STATISTICAL SCENARIOS FOR EXISTING QUERIESCATEGORYPAGE AVERAGE POSITION VALUE CHANGE POSSIBILITIES
no queries positions changeNeutralsame
all queries positions decreasedGoodsmaller
all queries positions increasedBadlarger
majority of queries positions decreased, minority of queries positions increasedGoodsmaller, same, larger
minority of queries positions decreased, majority of queries positions increasedBadsmaller, same, larger
half of queries positions decreased, minority of queries positions increasedGoodsmaller, same, larger
minority of queries positions decreased, half of queries positions increasedBadsmaller, same, larger
minority of queries positions decreased, minority of queries positions increasedNeutralsmaller, same, larger
majority of queries positions decreased, no queries positions increasedGoodsmaller
no queries positions decreased, majority of queries positions increasedBadlarger
half of queries positions decreased, no queries positions increasedGoodsmaller
no queries positions decreased, half of queries positions increasedBadlarger
minority of queries positions decreased, no queries positions increasedGoodsmaller
no queries positions decreased, minority of queries positions increasedBadlarger
half of queries positions decreased, half of queries positions increasedNeutralsmaller, same, larger

From Table 6, it is evident that attempting to map good, bad, or neutral performance categories to value changes in the Page Average Position—whether decreasing, increasing, or remaining the same—is statistically incorrect. Therefore, in addressing our question, we can now answer that a decrease, increase, or unchanged value in Page Average Position does not equate to categorizing performance as good, bad, or neutral, respectively.

And because Site Average Position is an average of all the Page Average Positions, this statistical limitation on mapping value changes exclusively to performance categorizations carries through to Site Average Position.

Thus we have quantitatively shown from these statistical scenarios of Page Average Position that both Site Average Position and Page Average Position value changes do not statistically equate to performance.

Next, we look at Query Average Position.

Query Average Position

The Query Average Position (QAP) is calculated for a single query and considers all pages that query is driving traffic to over a given period of time. This is another one-to-many relationship like Page Average Position.

google search query average position one to many pages to query relationships
Figure 4 – Diagram of Google search Query Average Position one-to-many relationship between a single query and the multiple pages that this query is driving traffic.

Perhaps switching the filter to Query Average Position will change the statistical mappings and allow us to finally assign good, bad, or neutral performance categories to the value of Query Average Position decreasing, increasing, or staying the same, respectively.

However, when we examine how Query Average Position is filtered, we find that the previous 15 statistical scenarios of Page Average Position still persist with Query Average Position, with the same descriptions, categories, and mix of value change possibilities of Query Average Position increasing, decreasing, and staying the same. The only difference is that we have reversed the one-to-many relationship between query and page.

Query Average Position Statistical Scenarios

Table 7 – A list of Query Average Position statistical scenarios for combinations of individual average position value changes for all pages driving traffic to a single query with performance category and QAP value change possibilities.
QUERY AVERAGE POSITION STATISTICAL SCENARIOS FOR EXISTING QUERIESCATEGORYQUERY AVERAGE POSITION VALUE CHANGE POSSIBILITIES
no pages positions changeNeutralsame
all pages positions decreasedGoodsmaller
all pages positions increasedBadlarger
majority of pages positions decreased, minority of pages positions increasedGoodsmaller, same, larger
minority of pages positions decreased, majority of pages positions increasedBadsmaller, same, larger
half of pages positions decreased, minority of pages positions increasedGoodsmaller, same, larger
minority of pages positions decreased, half of pages positions increasedBadsmaller, same, larger
minority of pages positions decreased, minority of pages positions increasedNeutralsmaller, same, larger
majority of pages positions decreased, no pages positions increasedGoodsmaller
no pages positions decreased, majority of pages positions increasedBadlarger
half of pages positions decreased, no pages positions increasedGoodsmaller
no pages positions decreased, half of pages positions increasedBadlarger
minority of pages positions decreased, no pages positions increasedGoodsmaller
no pages positions decreased, minority of pages positions increasedBadlarger
half of pages positions decreased, half of pages positions increasedNeutralsmaller, same, larger

In our third type of average position, QAP, we have found the same relationships from our analysis of PAP -  performance categorizations of good, bad, and neutral do not strictly map to value changes in Query Average Position.

This should be evident for both PAP and QAP as both types of average position are taking multiple average position changes, either from queries or pages, through a one-to-many relationship. So then individual average position changes can statistically contribute with increasing, decreasing, or same value changes to the overall average position type along with how much the value changes are differentiated between the post and pre periods.

Now there is one average position type left to review, Query Page Average Position.

Query Page Average Position

We review our last average position type, Query Page Average Position (QPAP). Because we are mapping a single page to a single query, we could equally call this Query Page Average Position, but we’ll choose the former because its acronym is easier to say. This is our only one-to-one relationship for average position type.

google search page query average position one to one page to query relationships
Figure 5 – Diagram of Google search Page-Query Average Position one-to-one relationship between a single query and the single page this query is driving traffic to.

Since Query Page Average Position tracks average average position changes for a single query and for a single page, we no longer encounter the underlying statistical scenarios found with Page Average Position and Query Average Position (and which statistically both contribute to Site Average Position). Because a change in Query Page Average Position results from the granular, or quantum, one average position change for one query and one page, we can now map a decrease, increase, or no change in value of Page-Query Average Position to the categories of good, bad, or neutral performance, respectively.

We now have an answer for the statistical mapping of performance categories to value changes for all average position types and we show this below in Table 8.

Table 8 – All four Google search average position types of Site Average Position, Page Average Position, Query Average Position, and Query Page Average Position and their statistical mapping of value changes to performance categorization.
Google Search Average Position TypeDoes decreasing, increasing, or same value of [TYPE] Average Position statistically equate to the categories of good, bad, or neutral performance, respectively?
Site Average PositionNo - multiple statistical scenarios limit this exclusive mapping.
Page Average PositionNo - multiple statistical scenarios limit this exclusive mapping.
Query Average PositionNo - multiple statistical scenarios limit this exclusive mapping.
Query Page Average PositionYes - a single statistical scenario supports this exclusive mapping.

It should be noted that this analysis of statistical scenarios for changes in the average positions of individual queries only considered existing queries—those present in both the pre and post periods. This analysis did not account for lost queries, which appear only in the pre period, or new queries, which appear only in the post period. Despite this simplified focus on existing queries, we found that only the Query Page Average Position supports the statistical mapping of performance categories to its value changes.

There is, however, always churn and acquisition of queries, with lost and new queries comprising a sizeable percentage of total queries in the pre and post periods, respectively. These churn and acquisition rates mostly reflect the appearance and disappearance of rare keywords that occur at very low frequency over time. The high-frequency keywords, which hold higher value in terms of volume of traffic, usually persist as existing queries. By focusing only on existing queries, this analysis provided a simple foundational statistical examination of average position types, allowing for the identification of type characteristics and limitations at their most fundamental level.

Thus, introducing lost or new queries into the analysis would not change these results but would only intensify their complexity and strengthen the statistical conclusions. Calculating churn and acquisition rates of queries would result in an increase in the possible statistical scenarios underlying Page Average Position, Query Average Position, and Site Average Position. 

It should be observed that the performance tracking of statistical scenarios of Page Average Position and Query Average Position are only accomplished because each scenario takes into the account the individual performances of all single page to single query average position changes. It's these individual Query Page Average Positions that build our statistical scenarios for PAP and QAP. So a major take away from all this analysis should be that performance categorization for any Google search average position type is based in calculated Query Page Average Positions, no matter the type of analysis. 

Also, while we reviewed these average position types within the Google Search Console dashboard, these are statistical definitions that exist regardless of the format in which they are viewed. These average position types can be easily queried in other Google tools like Google Big Query, or in standalone databases or data warehouses.

Lastly, while there are ways to further segment average position in Google Search Console, these do not create new types of average position, as we comprehensively and statistically reviewed. For example, segmenting your search data by type of device will not change the definition of Site Average Position, or any other average position type. It only changes which devices you are viewing Site Average Position for on your website.

Case Study in Google Search Average Position 

In the following study, we analyzed a large single website and examined the Site Average Position between two consecutive months, along with a comprehensive analysis of the underlying statistical scenarios of Page Average Position for every URL.

Table 9 – Site Average Position for a website between February and March.
MonthSite Average Position
February22.2
March22.1

The Site Average Positions between the two months are essentially the same, with a less than 1% decrease in value. However, upon examining the Page Average Position statistical scenarios more closely, we find the following percentage changes for each:

Table 10 – Page Average Position statistical scenarios percentages found present when comparing between the pre and post periods of February and March for a case study website.
PAGE AVERAGE POSITION STATISTICAL SCENARIO FOR EXISTING QUERIESSCENARIO PERCENTAGES FOUND IN THE POST PERIOD
no queries positions change1.38
all queries positions decreased26.22
all queries positions increased26.09
majority of queries positions decreased, minority of queries positions increased14.74
minority of queries positions decreased, majority of queries positions increased17.04
half of queries positions decreased, minority of queries positions increased0.61
minority of queries positions decreased, half of queries positions increased0.68
minority of queries positions decreased, minority of queries positions increased1.35
majority of queries positions decreased, no queries positions increased0.01
no queries positions decreased, majority of queries positions increased0.01
half of queries positions decreased, no queries positions increased0.5
no queries positions decreased, half of queries positions increased0.52
minority of queries positions decreased, no queries positions increased0.02
no queries positions decreased, minority of queries positions increased0.02
half of queries positions decreased, half of queries positions increased10.81

While Site Average Position has remained essentially constant between these two months, the percentage breakdown of Page Average Position statistical scenarios reveals that several pages have fluctuating search positions. For example, 26.22% of all pages experienced a decrease in the positions of all queries driving traffic to their respective pages, which is a positive outcome. Conversely, 26.09% of all pages saw an increase in the positions of all their queries, which is a negative outcome. Referring back to our categorization of statistical scenarios, we summed all good scenarios, all bad scenarios, and all neutral scenarios. From this, we find the following:

Table 11 – Case study website percentages for statistical scenarios categorized as good, bad, or neutral performance for changes in Page Average Position between pre and post periods.
Good Scenarios Percentage42.09%
Bad Scenarios Percentage44.35%
Neutral Scenarios Percentage13.56%

Breaking down the Site Average Position into the Page Average Position statistical scenarios, we see that while Site Average Position remained essentially the same between these two months, more than 85% of the URLs experienced net average position changes in their underlying queries, either shifting under a good scenario or a bad scenario.

This breakdown of statistical scenarios only looked at percentages of categorized good, bad, and neutral performance scenarios. But we know that at the Page Average Position level, these scenarios do not map value changes of PAP to performance categorization. So then a natural question would be - what percentage of each scenario maps to PAP increasing, decreasing, or staying the same in value. 

If you recall the added column for PAP value change possibilities across all statistical scenarios, there are 27 in total. However, the distribution of these scenarios matched to PAP value changes (or no change) is not equal. Therefore, we will focus only on the statistical scenarios matched to PAP value changes that account for at least 5% or more of all these matched scenarios. The results are shown in Table 12 below.

Table 12 – Percentage breakdown of top 10 Page Average Position statistical scenarios matched to PAP changes.
PAGE AVERAGE POSITION STATISTICAL SCENARIOPAGE AVERAGE POSITION VALUE CHANGE% PAGE AVERAGE POSITION STATISTICAL SCENARIO &
half existing queries positions decreased, half queries positions increasedPage Average Position decreased5.59
half existing queries positions decreased, half queries positions increasedPage Average Position increased5.22
All existing queries positions decreasedPage Average Position decreased17.32
All existing queries positions decreasedPage Average Position increased8.90
All existing queries positions increasedPage Average Position decreased9.86
All existing queries positions increasedPage Average Position increased16.23
Majority existing queries positions decreased, minority of queries positions increasedPage Average Position decreased9.13
Majority existing queries positions decreased, minority existing queries positions increasedPage Average Position increased5.61
Minority existing queries positions decreased, majority queries positions increasedPage Average Position decreased6.73
Minority existing queries positions decreased, majority queries positions increasedPage Average Position increased10.31

It may be surprising, but there is a mix of good and bad scenarios with both increasing and decreasing Page Average Position. For example, in the quantitative good performance scenario of "majority existing queries positions decreased, minority existing queries positions increased," this scenario made up 14.74% of all statistical scenarios. However, when matching to PAP value changes, 61.9% of these scenarios are matched to a decreasing PAP, while 38.1% are matched to an increasing PAP. In fact, among these top scenarios, not a single one matches to a specific PAP value change (increasing or decreasing) with a percentage as high as two-thirds or more. Only the "all existing queries positions decreased" scenario matched to PAP decreased came close to this at 66.1%.

You may have noticed in the last example that we have PAP increased for 33.9% of the "all existing queries positions decreased" scenario, which wasn't previously mapped as a possible PAP change. This discrepancy arises because the previous mapping of statistical scenarios to performance categorizations and possible value changes in PAP were found only for existing queries. This real-world data includes lost and new queries, which also contribute to PAP in both the pre and post periods.

This illustrates the increasing complexity behind Page Average Position, Query Average Position, and Site Average Position as lost and new queries are taken into account. The Query Page Average Position remains statistically shielded from this complexity, as evaluating only a single page and single query allows for a granular performance review for not just existing queries, but also lost and new queries.

So, let's now discuss lost and new queries by reviewing the churn and acquisition rates of queries over time.

Google Search Existing Queries, New Queries, & Lost Queries

When analyzing Google average position over time for any of the four main types of search average position, we observe the carryover of existing queries, the gain of new queries, and the loss of queries between two time periods.

The churn rate of lost queries and the acquisition rate of new queries can fluctuate between time periods. However, for a site with steady traffic, we would statistically expect the percentage of existing queries that carry over to make up at least half of all queries. Additionally, we expect the percentage of gained queries to be relatively close in number to that of lost queries.

In Figure 6 below, we have the website from our previous case study showing existing queries, new queries, and lost queries by month. 

case study website search traffic existing new lost queries over time graph
Figure 6 - Case study website showing search traffic existing queries, new queries, and lost queries over time by month.

Existing queries are typically composed of high-frequency searched queries that consistently drive traffic to a site over time. New and lost queries, on the other hand, tend to be rare queries with less frequent appearances in search. However, of interest in Figure 6 above is the decline in new queries—the query acquisition rate has gone down for two consecutive months. This has resulted in a decrease in total queries over these months, as existing queries and lost queries have remained relatively stable in their percentages.

This trend could indicate the the site was affected by a recent Google search algorithm change, among other factors. 

But this is one of the reasons our initial analysis of statistical scenarios only considered existing queries—they tend to make up the most stable and valuable search traffic to a site. Additionally, we aimed to keep the initial analysis simple to establish a foundational understanding of the complexity behind Google average position types, even in their simplest form.

If we were to also account for the statistical scenarios of new and lost queries, the complexity of the analysis would increase exponentially, as we would need to factor in a product of existing queries statistical scenarios and lost queries statistical scenarios and new queries statistical scenarios. While this could provide valuable insight into a website's search performance, our primary interest lies in how the visibility and positions of existing queries vary over time.

Adding another layer of complexity, we have yet to consider the qualitative differences between queries, position changes, and statistical scenarios. For example, while statistical scenarios offer a quantitative analysis and comparison across pages and queries, the value of these changes may shift with qualitative analysis.

For instance, the scenario "majority existing queries positions decreased, minority existing queries positions increased," which we previously reviewed, is quantitatively considered a good scenario since all queries are treated with equal impressions and value. However, in a qualitative analysis, we might find that the minority of existing queries that increased in position are among the most valuable queries to the page from a business perspective. This could potentially change the scenario from good to bad, highlighting the impact of qualitative analysis.

Such qualitative analysis is subjective to the business, its goals, industry, past performance, and other factors. While these considerations are important, they go beyond the scope and intention of this paper.

Site Average Position vs Query Page Average Position

We'll take one more look at our case study website and compare its Site Average Position (SAP) to its Query Page Average Positions (QPAPs) over time. Figure 7 below shows the SAP, which factors in search traffic to all pages and from all queries. The Site Average Position for the case study website has remained relatively steady, typically around 20 on weekdays and dipping to around 25 on weekends.

case study website search site average position over time
Figure 7 - Case study website Search Average Position over time.

There is limited performance information we can gain from the Site Average Position in this graph. However, by examining individual Query Page Average Positions, we can view specific performance gains and losses over time.

It should be noted that Site Average Position (SAP) is displayed as a single value because it includes all pages and queries for the site. In contrast, Query Page Average Position (QPAP) is calculated for each individual page-query combination driving traffic to the site. As a result, the two graphs show different data formats: SAP appears as a single line over time, while QPAPs are represented as data points for each unique combination that appears daily within our study period.

case study website search page query average positions over time
Figure 8 - Case study website multiple Query Page Average Positions over time.

While there is a minor fluctuation in Site Average Position in early March, the overall trend appears to stabilize. However, as shown in Figure 8 above, tracking multiple Query Page Average Positions reveals that thousands of content pages are experiencing a wide spectrum of value changes in QPAP.

In fact, we can see that several of these pages are losing performance ranking in search beginning in March and continuing through April and May. This crucial performance information highlights how individual content pages are performing in search over time, specifically in terms of their average positions. This level of detail could not be obtained by looking at the Site Average Position alone.

Our statistical analysis, which identifies Query Page Average Position as the only search average position type that maps value changes to performance categorization, coupled with our real-world data of Site Average Position and Query Page Average Positions from our case study website, indicates that performance analysis of a website's average position can only be accurately conducted by calculating the QPAPs of the site.

Summary

In this comprehensive statistical analysis of Google search average position, we reviewed Google's definition and calculation of average position, along with the ways to filter and view this metric in the Google Search Console dashboard. We emphasized the significance of impressions in calculating the weighted average position and identified the four types of search average position: Site Average Position, Page Average Position, Query Average Position, and Query Page Average Position. We further examined how each of these average position types has its own unique statistical characteristics and limitations in performance analysis and reporting.

Through detailed examples and statistical scenarios, we illustrated that changes in average position types cannot be simplistically categorized as good, bad, or neutral performance, except at the granular level of Query Page Average Position. If there is one key takeaway from this paper, it is that statistical analysis and reporting on the performance of a website based on value changes in any average position type must be grounded in Query Page Average Position. This applies whether reporting directly from a single or multiple QPAPs, or identifying the type and frequency of statistical scenarios for Page Average Position or Query Average Position, and extending further to Site Average Position.

In summary, understanding Google's search average position metric necessitates a detailed and nuanced approach that considers the four types of average position, statistical scenarios of the aggregated types, and how these types and scenarios map to search performance. These insights are critical for any company striving to understand the performance of content on their websites and how the visibility and ranking of this content changes over time. Webmasters and SEO professionals aiming to optimize their site's visibility and performance in search results would be more effective in using the analysis and reporting of Query Page Average Positions and PAP/QAP statistical scenarios, the latter which performance categories are identified through varying combinations of QPAPs. 

About the Author: Daniel James Stoker is a member of the FindLaw Performance Team and holds degrees of B.S. in Physics, B.S.H.S. in Physiological Sciences, a M.S. in Computer Sciences Database Technologies, and has completed a PGP in Artificial Intelligence and Machine Learning: Business Applications. 

LinkedIn

Schedule a consultation online

Let’s discuss your marketing challenges and how firms like yours solve those problems. If nothing else, you’ll walk away with legal marketing knowledge you didn’t have previous to the meeting.

An accurate Zip Code ensures your appointment is set with the correct local marketing consultant.