The first thing to check is scope: did one page drop or did many drop together? That single count comes before any fix, because the answer routes the entire investigation down one of two roads. A drop confined to a single page and a drop spread across the site have almost nothing in common as problems, and chasing the wrong one wastes the days that matter most.
Start by counting, not fixing. Pull the pages and queries that lost position over the same window and see whether the loss clusters on one URL or scatters across many. Resist the urge to edit the page you happened to notice first, because that is acting before you know what kind of problem you have, and the wrong fix on a site-wide issue leaves the real cause untouched while you congratulate yourself.
If the drop is isolated to one page, the diagnosis points inward to that page. The likely causes are content and intent: the query’s meaning shifted, a competitor produced something stronger, the page was edited in a way that weakened it, or it lost a link that was carrying it. The investigation stays narrow, on that URL and the SERP it competes in.
If many pages dropped at once, the diagnosis points to something shared across them. Now the likely causes are technical or algorithmic: a crawl or indexing problem, a sitewide template or configuration change, a server issue, or an algorithm update that re-weighted the whole site. The investigation widens to systems and dates rather than a single page’s words. The two roads do not overlap, which is exactly why you commit to one only after the scope is known.
So the verb is count first, diagnose second. Knowing the breadth of the drop tells you whether you are looking for a page problem or a site problem before you spend any effort solving the wrong one.
After your next drop, open your rank data and tally how many pages and queries lost ground in the same window before you touch anything, then follow the count to either the single page or the whole site.