Average position is easy to read too quickly. A number gets smaller, and a report calls it an improvement. A number gets larger, and somebody opens a task to fix it. Before making either decision, establish which impressions contributed to the average and whether the comparison still describes the same part of the business.
This guide uses an invented example to show a practical investigation. The goal is to help an analyst decide what to examine next. It does not predict a traffic change or suggest that a single number can explain one.
Know what the metric represents
In Google’s explanation of impressions, clicks and position, Google explains that Search Console position reflects the topmost result for the property or page in an impression, averaged across the recorded impressions. Result layouts and aggregation affect its interpretation. A manual search is one observation under its own conditions, so it need not match an account’s average.
Keep that definition close to the report. A client reading “position 6.4” may imagine the same page holding the same sixth place every time someone searches. That mental picture is too simple for the decisions the report is trying to support. Explain the scope beside the number so the reader can judge what it describes.
The useful initial question is: what exactly does this row include? Write down the property, search type, dates and filters before describing the change. This small habit prevents a later screenshot from becoming detached from the comparison it was intended to support.
Make the question narrower
Suppose a company sells commercial storage equipment. The owner wants to understand visibility for a specific product family in the United Kingdom. Start with that customer question. Choose a documented query group related to the product family and record why each term or matching rule belongs there.
A group built with a text rule needs a quick relevance check. A product word might appear in an unrelated informational query or a different range. Inspect the returned queries before interpreting the group. If the rule changes after review, use the same revised rule for both periods and keep a note of the change.
Avoid comparing a tightly filtered view this month with an unfiltered total from the previous report. The two numbers may look compatible because they share a label, but they describe different sets of observations. Save the exact configuration so another person can reproduce the intended comparison.
Hold device and country steady
For the first pass, select one country and one device group, such as UK mobile. Use two complete intervals of equal length and note holidays, campaigns or other context relevant to the business. Equal duration is a starting convention, not a guarantee that demand is comparable.
Then inspect desktop separately. If mobile and desktop behave differently, describe that difference rather than hiding it inside an account-wide average. A useful report might say that the question is concentrated on mobile and that the next review will focus on the pages receiving those impressions.
The Search Console Performance report documentation describes filtering and comparison options in the Performance report. Use the available controls to preserve scope, then document the business choices the interface cannot make for you: the intended product group, the reason for selecting the interval and what would make the finding worth acting on.
Read the other measures beside position
Imagine that UK mobile average position changes from 8 to 11 while impressions rise from 1,000 to 3,000 and clicks rise from 80 to 120. These are fictional figures for a reasoning exercise. The headline “rankings worsened” leaves out information the owner would need to judge the situation.
One possibility to investigate is that the group gained exposure for additional queries. Another is that the existing queries changed. You cannot choose between those explanations from the aggregate numbers alone. Compare the query rows and page groups, keeping the original question in view.
Now imagine the opposite: average position looks better, but clicks decline. A report should not celebrate automatically. The business may have lost exposure in a part of the query set while the remaining impressions came from stronger positions. Again, the next step is to examine the composition, not to assign a cause from the summary.
Separate stable queries from new exposure
Create two working lists. One contains queries visible in both intervals. The other contains newly observed or no-longer-observed queries. Treat these lists as a way to organize investigation, with the data limitations clearly stated. They are not a complete inventory of every search that happened.
Review a small, relevant subset of the stable group first. Which landing pages received the activity? Did the business change those pages? Were products removed or renamed? A change log and the team’s operational context may explain which questions deserve closer inspection.
For newly observed queries, ask whether the audience fits the offer. A broader mix can be commercially useful or irrelevant. The answer requires looking at the actual intent and the page serving it. A position average alone cannot make that business judgment for you.
Keep page and property questions distinct
A company-wide question and a landing-page question should have different report labels. At the property level, the owner may be asking about overall visibility for a product group. At the page level, the editor may be checking whether a particular resource is reaching its intended audience. Decide which question the current analysis serves.
If multiple pages are involved, list the relevant URLs in the working record. Check whether a changed URL, canonical choice or content move affects how the team understands the comparison. Do not assume a page has lost all activity just because a familiar URL is absent from one view.
When uncertainty remains, write it as a specific follow-up. “Check how this page group is represented after the migration” gives someone a task. “Google is confused” is an interpretation that does not tell a colleague what evidence to gather.
Turn the finding into a bounded action
A useful conclusion has three parts: the observed pattern, the unresolved explanation and the next step. For the fictional equipment example, the analyst might report that UK mobile exposure expanded, that the source of the broader mix needs review and that a named colleague will examine the new queries before recommending page changes.
The team can then agree what would change the recommendation. If the new exposure is relevant, it might support developing the product information. If it is irrelevant, the analyst may need a narrower reporting group before drawing a commercial conclusion. Those are different decisions that should not be concealed by one average.
Save the comparison dates, filters, selected queries and notes together. At the next review, reopen that record instead of starting from a new account-wide chart. The benefit is a more consistent investigation, with less time spent arguing about what last month’s number meant.
Start today with one query group, one device and one country. Check the scope, read the surrounding measures and write one sentence explaining what you still need to learn. That is a practical use of average position: a prompt for a better question and a clearer next action.