Finding out what actually caused the drop before anyone starts fixing things

Most drops are not what the owner thinks they are

The call is nearly always the same. Traffic fell, somebody said it was a penalty, and now there is an urge to change everything at once. That urge is the most expensive part of the situation, because rebuilding a site before the cause is known destroys the only thing that would have told you whether the fix worked.

There are four genuinely different causes and they need completely different responses. A manual action is a penalty applied by a human reviewer, and it is written plainly in Search Console — bought links, cloaking, empty spam pages, fabricated review markup. If the manual actions report is empty, this is not what happened to you. An algorithm update is Google reassessing broadly, and updates are announced and dated, so the first move is matching your decline against the published dates. Something the site did to itself is the largest category and the one owners forget — a redesign, a migration with incomplete redirects, a staging rule that reached production, a plugin update, a hosting problem. And finally the drop may not be a drop at all, but a change in how the numbers are collected.

That last one is worth spelling out because it looks exactly like a catastrophe. It shows up as impressions dropping away while clicks stay roughly flat and average position appears to improve. Nothing happened to the site. The way results were being counted changed, removing impressions that were never seen by a person and taking the deepest positions out of the average along with them. Read that chart in the usual way and you would spend three months and a large budget fixing nothing.

How the diagnosis and recovery run

1. Triage in the first week. The manual actions report. The exact date the decline started, taken from Search Console rather than from memory. That date matched against Google's published update history and against your own release calendar. Then the shape of the loss — which pages, which queries, which countries, which device types, informational or commercial. Five days.
2. Confirm the cause with evidence, or rule it out. A crawl compared against the indexing report to find pages that stopped being indexed and why. Checks for the self-inflicted causes — robots rules, noindex directives, redirect chains left from a migration, server errors, pages that disappeared in a content clear-out. Manual checks in a clean browser on the queries that lost clicks while holding position. One week.
3. Write the remediation plan and rank it. Every finding with what it is, where it is, what we believe it cost, the effort to fix, and who does it. Ordered by expected effect rather than by ease, and explicitly separating what we are confident about from what is a hypothesis. You should be able to see which items are being done because we know and which are being done because they are cheap. Two to three days.
4. Implement, one class of change at a time. Not everything at once. Technical blockers first because they are provable, then content and structural work, each batch given enough time for pages to be recrawled before the next begins. Where the cause was a manual action, the offending material is genuinely removed rather than hidden, and only then is a review requested. Three to six weeks.
5. Verify against the baseline, and keep watching. Indexing counts, non-branded clicks, query-level movement, and the same manual result-page checks repeated on the same queries. Recovery from an algorithmic reassessment often becomes visible over following weeks rather than on a single day, so monitoring continues past the implementation. Four to eight weeks of monitoring.

What sets the price

The range is $800–$6,000 over 1–3 months, and it moves with three things.

Which of the four causes it turns out to be. A self-inflicted technical fault found in week one and fixed in week two is at the bottom of the range. A site facing an algorithmic reassessment of its content quality across several thousand pages is a months-long piece of work and sits at the top.

How much history has to be reconstructed. Sites with a saved crawl, exported query data and a known release calendar are quick to diagnose. Sites where nothing was archived and nobody remembers when the redesign shipped need that timeline rebuilt from what survives before any conclusion is safe.

Whether we implement or specify. As everywhere else here, a report costs less and does nothing on its own. Recovery work in particular tends to stall when it is handed over, because the individual items look small and none of them are anybody's priority.

The three ways recovery money gets wasted

Work starts on the assumption that it is a penalty. Manual actions are rare and they are announced in writing. Buying a recovery engagement without opening that report means paying to solve a problem the site does not have, while the actual cause — very often a migration that left a section unreachable, or a directive that shipped from staging — carries on unaddressed.

The damage is measured in impressions and average position. These are the two numbers most easily moved by things that have nothing to do with your site, and the two most likely to be quoted in a panicked email. Clicks and conversions are what pay for the business. A diagnosis built on impression charts can conclude that a site has collapsed when its actual commercial performance is unchanged.

The affected pages get deleted. The instinct after a content-related decline is to remove everything that dropped. Those pages frequently carry years of accumulated references and positions on queries that still convert, and deleting them throws that away permanently while doing nothing about the reason. Improving a page keeps what it has. Merging it into a stronger page and redirecting keeps most of it. Deleting keeps none of it, and it is the one action that cannot be reversed once the address stops resolving.