Case Study 2 — The Disavow That Dug the Hole Deeper

A complementary angle. Case Study 1 watched a real, self-created link scheme draw a real manual action — the danger this chapter takes seriously. This one watches the opposite failure: a capable team, hit by an algorithmic drop with no link problem at all, misdiagnoses it as a link penalty and "fixes" it into a deeper, longer disaster. It teaches the chapter's discipline from the failure side, where it bites hardest — the confusion of §26.6, the danger of the disavow tool (§26.3), and the negative-SEO paranoia of §26.7, all in one cautionary sequence.

This is a labeled composite. It is not one named company; it is a pattern that plays out constantly, in agencies and in-house teams, assembled from the common, well-documented ways link panic goes wrong. Every number below is illustrative — round figures chosen to show the shape. What is real is the sequence of reasonable-seeming mistakes, and how each one felt right in the moment.

Background: a good site, a bad morning

A mid-size independent publisher — call it a hobby-and-crafts site with a decade of genuinely useful how-to content — earns about 300,000 organic visits a month. It has real expertise, a real audience, and a respectable, honestly-earned link profile. It has never bought a link, never had a manual action, never so much as opened the disavow tool.

Then comes the bad morning from Chapter 6: over about a week, organic traffic falls from 300,000 toward 190,000 and keeps sliding. It is a real, painful drop. The timing, had anyone checked, lines up precisely with a broad core update Google confirmed on its status dashboard that same week — which makes this, by every marker in §26.6, an algorithmic demotion: silent, site-wide, no notice, a quality reassessment. The correct response is the slow one from Chapter 6 (diagnose honestly, improve the weakest content, wait for reassessment). Here is how the team talks itself out of that, one reasonable step at a time.

The issue: four moves, each of which felt like taking action

  • They skip the ten-second check, and reach for "penalty." Nobody opens the Manual Actions report; if they had, it would have said "No issues detected," ending the whole misadventure in seconds (§26.4). Instead, "our traffic got penalized" becomes the working assumption, and "penalty" points them at links.
  • A toxic-link scanner supplies a villain. Under pressure to find the cause, someone runs a third-party backlink audit. It obligingly reports 1,900 "toxic" links and a lurid toxicity score. This feels like the answer — here is a number, here is a list, here is something to fix — even though (§26.1–§26.2) those flags are a vendor heuristic full of false positives, and the site never built a single one of them. The scanner has manufactured a diagnosis to match the panic.
  • They disavow aggressively — including good links. They upload a disavow file covering all 1,900 flagged domains. Buried in that list are perfectly legitimate links the model misjudged: a link from a well-known crafts magazine, two from popular blogs in their niche, a university resource page. In one upload, they tell Google to stop counting real votes they had earned over years (§26.3). Authority they were genuinely receiving is now switched off — by their own hand.
  • They file a reconsideration request into the void. Believing they have "cleaned up the penalty," they submit a reconsideration request. There is no manual action, so there is nothing to reconsider; the request is paperwork mailed to an office with no case open (§26.5, §26.6). They wait on it anyway, treating "we're waiting to hear back from Google" as progress, which stalls the real work for weeks.

text FIGURE C26.2 — "How a recoverable drop became a lost quarter" [constructed teaching example] WEEK 0 (before) THE MISDIAGNOSIS CASCADE THE RESULT ~300,000 organic visits/mo Core update hits (algorithmic) real ~35% dip — recoverable No manual action ...but never checked the report chased a "penalty" that didn't exist Honestly-earned links toxic scanner flags 1,900 "toxic" disavowed good links among them Zero disavow history mass disavow + reconsideration req removed real votes; waited on nothing (All figures illustrative — the SEQUENCE of avoidable mistakes is the lesson, not the numbers.)

The deeper hole: how the "fix" made it worse

Here the case earns its place as the failure study, because the second wave of damage is self-inflicted and more instructive than the first.

The core update had cost the site perhaps a third of its traffic — a genuine but recoverable algorithmic dip, the kind Chapter 6 says you climb out of by improving quality and waiting for reassessment. But now the site has also amputated a slice of its real link authority by disavowing good links. So when the next reassessment comes, the site is not merely being judged on quality — it is being judged with weaker earned authority than it had before, because it voluntarily switched some off. The disavow did nothing about the actual cause (there was no link cause) and actively worsened the site's standing. Traffic that might have begun recovering instead sags further and flattens at a lower level.

Worse, the team spends six weeks in the wrong posture entirely: waiting on a reconsideration reply that will never resolve anything, congratulating themselves on the "cleanup," and not doing the one thing that would have helped — the honest content audit and quality work of Chapters 6 and 12. By the time someone senior asks "wait, did we ever confirm this was a link penalty?" and opens the Manual Actions report to find the serene green "No issues detected," a recoverable quarter has been lost, and undoing the bad disavow (removing the good domains from the file and waiting for Google to re-crawl and re-credit them) is itself a slow, months-long reversal.

What it shows

  1. The ten-second check is the whole ballgame. Every downstream disaster traces to the un-opened Manual Actions report. One look — "No issues detected" — would have ruled out the entire link-penalty theory and pointed the team at Chapter 6 instead of the disavow tool. Look before you say "penalty."
  2. A toxicity score is a villain generator, not a diagnosis. Under pressure to find a cause, a team will accept the first number that looks like one. The scanner's 1,900 "toxic" links were a heuristic answer to a question the team never should have been asking, and treating the score as truth (§26.1) is what loaded the gun.
  3. Disavow is subtraction, and you cannot un-subtract quickly. Disavowing good links removed real authority (§26.3), and reversing it is slow. The tool has no upside for a site with no manual action and a real downside — this case is that downside, realized.
  4. "Doing something" is not the same as doing the right thing. Every move felt like action: run the scanner, upload the file, file the request, wait for Google. None of it addressed the actual, algorithmic cause — and two of the moves made things worse. The disciplined response (open one report, then do the slow quality work) looks like inaction and is in fact the only thing that helps.

Outcome

In the version a calm practitioner produces, the story ends on the first morning: someone opens the Manual Actions report, sees "No issues detected," checks the timing against Google's status dashboard, recognizes a core-update algorithmic demotion (Chapter 6), and never touches the disavow tool. They begin the honest quality program — prune and deepen the weakest content, strengthen the genuine expertise signals — and wait for reassessment, diversifying traffic in the meantime. Traffic dips, holds, and climbs back over the following updates, because nothing that was working was thrown away.

In the version that actually happened, a recoverable algorithmic dip became a lost quarter plus a self-inflicted authority loss that took months to reverse — and the lasting, expensive lesson the team learned was every heading in this chapter: check the report, distrust the score, keep your hand off the disavow tool, and never confuse the silent disease (algorithmic) with the notified one (manual).

Connect it to Rivertown. This is the disaster the Delgados' Strategy File is built to avoid. Rivertown inherited some genuinely spammy links (unlike this publisher, it does have a small self-procured cluster) — and still the decision is ignore, not disavow, precisely because the same failure logic applies: there is no manual action, the links are not the bottleneck, Google already ignores the junk, and disavowing carries downside for zero upside (§26.3, Strategy File). Rivertown has 8,000 visits it cannot afford to jeopardize by amputating a link it misjudged. The publisher in this case had the clean bill of health and threw it away by acting; Rivertown reads the same report, draws the same "no manual action" conclusion, and has the discipline to stop.

Discussion questions

  1. The entire cascade begins with one un-opened report. Name it, say what it would have shown, and explain why that single check makes the link-penalty theory collapse.
  2. The toxic-link scanner reported 1,900 "toxic" links. Reconstruct why that number felt like a diagnosis to a panicked team, then dismantle it using the provenance question (§26.2) and the false-positive problem (§26.1).
  3. Disavowing is described as "subtraction you cannot un-subtract quickly." Explain, using how links pass authority (Chapter 22), what was actually lost when the team disavowed legitimate links — and why reversing it is slow.
  4. This site had no manual action, yet filed a reconsideration request. Explain precisely why that request could never help, and what the team was mistaking it for.
  5. Contrast this case with Case Study 1 (JCPenney). One site should have used the disavow-and-reconsideration machinery and (in the modern version) would; the other should never have gone near it. What single diagnostic question separates the two — and why is getting it wrong the most expensive error in this whole chapter?