Average position is the most misleading number in SEO
Average position is the number every SEO report leads with, and on its own it is close to useless. Not wrong — useless. It moves constantly for reasons unrelated to performance, and it stays still through changes that matter enormously.
Three ways it lies
It improves when you lose visibility. Search Console weights position by impressions, so take a case where they are even: one query at position 3 with 100 impressions, and ninety queries at position 40 with 100 impressions each. The average is about 39.6. Now Google stops showing you for those ninety weak queries. Your average leaps to 3. Nothing improved — you lost 9,000 chances to be seen.
It worsens when you gain. Publish something genuinely good and you start appearing for two hundred new queries, initially at position 30. Your average position gets dramatically worse. You are doing better than you were last month.
It is unweighted. A query with 40,000 impressions and one with 4 count the same. The number treats a phrase your business depends on identically to a typo somebody searched once.
What it is actually good for
One thing: watching a single query, or a single page, over time. Segment first, then trend. "Average position for our twelve commercial queries" is a real number. "Average position" is not.
What to watch instead
Impressions, because it is the size of the demand Google thinks you are relevant to. Rising impressions with flat clicks means you are being shown more and chosen less — usually a snippet problem, not a ranking problem.
Clicks against the ceiling. For the impressions you already earned, how many clicks would you have at position one? That ratio is honest in a way average position is not: given the visibility you already have, how much of it becomes visits?
Queries crossing thresholds. The count of queries that moved from page two to page one summarises a month's work far better than any average, because it maps directly onto traffic.
The habit worth breaking
If a number would move in a direction you like for a reason you would hate, it is not a good headline metric. Average position fails that test in both directions.
Report it if a client expects it. Do not make a decision with it.