The Worst SEO Advice I Ever Gave Cost Someone Two Million Dollars

The call was on a Thursday. I remember this because I had cleared Thursdays for "deep work," which is a concept I borrowed from Cal Newport and which in practice means I block off the day and then spend it answering emails and feeling guilty about not doing deep work. But this Thursday the phone rang and the number on the screen was one I recognized and one I did not want to see, because the last time this number had called me was six months earlier and the conversation had not gone well and the time before that it had gone worse and there is a specific kind of dread that attaches itself to a phone number, a Pavlovian flinch, where the mere appearance of ten digits on a screen can rearrange your internal organs before you've even answered.

I answered.

"Amos," said the voice on the other end. It was David. I'm going to call him David because that's not his name and because what happened to his company is the kind of thing that follows a person around and I don't need to make that worse. David was the VP of Marketing at a SaaS company I'd worked with two years earlier. A good company. Real product. Real users. Real revenue. The kind of company that doesn't usually need to call its former SEO consultant with a voice that sounds like it's been scraped over gravel.

"David," I said.

There was a pause. The kind of pause that has weight. The kind that tells you the person on the other end is choosing their words carefully, which is never a good sign, because people who have good news don't choose their words. They just say them. It's the bad news that requires curation.

"I want you to know that I'm not calling to yell at you," he said. "I'm past that."

This is worse than yelling. Yelling means there's still energy. Yelling means they think it might still be fixable. "I'm past that" means they've moved through anger into exhaustion and out the other side into a flat, gray acceptance that is much harder to sit with because it doesn't give you anything to push against.

"Okay," I said.

"We ran the numbers," he said. "On the migration."

The migration. My migration. The one I'd recommended. The one I'd scoped. The one I'd project-managed and overseen and signed off on and been confident about, not just professionally confident but personally, viscerally confident, the kind of confidence that comes from having done something a dozen times before and having it work every time. The kind of confidence that is, I now understand, the most dangerous substance in consulting.

"Two million," David said. "Give or take."

He said it flatly. No inflection. The way you'd tell someone the weather. It's raining. It's seventy degrees. You cost us two million dollars in lost revenue. He explained the math. Before the migration, they were generating roughly $430,000 a month in revenue that could be attributed to organic search. After the migration, that number dropped to $180,000 and stayed there. For fourteen months. They'd been waiting for the recovery I'd promised - three months, I'd said, maybe four, six at the outside, these things always bounce back - and the recovery hadn't come. Fourteen months at a $250,000 monthly shortfall. Two million dollars, give or take, and the "give or take" was not making anyone feel better.

"I'm sorry," I said, because what else do you say? What is the correct response when someone tells you that your professional advice, delivered with confidence and backed by experience, cost them two million dollars? There isn't one. "I'm sorry" is a placeholder for the absence of an adequate response. It fills the space where a real answer should go, the way packing peanuts fill the space in a box that isn't big enough for what's inside it.

The Recommendation

Here's what happened. I need to tell you what happened because the specific details matter and because without the specific details this is just another story about an SEO migration that went wrong, and there are thousands of those, and most of them are boring and most of them are attributable to obvious incompetence - broken redirects, lost content, botched canonical tags, the kind of errors that show up in a Screaming Frog crawl and make you wonder how someone who calls themselves an SEO professional could have missed them.

This wasn't that. This was worse. This was a migration that went right.

David's company had a problem. Their website had grown organically (in the non-SEO sense of "organically," meaning "without anyone really planning it") over eight years. It started as a WordPress blog. Then someone added a product section. Then someone else added a knowledge base on a subdomain. Then a third person, probably an intern, created a separate blog on Medium and started cross-posting content. Then a fourth person set up a documentation site on GitBook. The result was a digital archipelago - five separate properties, three different CMSes, two subdomains, and a Medium account, all nominally representing the same company but linked together with the structural coherence of a plate of spaghetti thrown against a wall.

I recommended consolidation. One domain. One CMS. One site architecture. Everything under one roof. This is, by the way, correct advice. If you have a fragmented web presence spread across multiple properties, consolidating it will almost always improve your organic performance in the long run. The theory is sound. The logic is impeccable. Google rewards topical authority, and topical authority is easier to build when all your content lives in one place, linking to itself, reinforcing itself, signaling to crawlers that this is a single, coherent entity that knows what it's talking about.

I recommended migrating from WordPress plus the constellation of satellite properties to a single Webflow instance. New domain structure. Clean URLs. Proper hierarchy. Content consolidated, deduplicated, and reorganized into topic clusters that made logical sense. It was, if I'm being honest, a beautiful plan. The kind of plan you put in a deck and present to a room full of executives and they nod because it makes sense, because it's organized, because it has a Gantt chart and a risk mitigation section and all the trappings of professional competence.

I built the redirect map. 3,400 redirects. Every old URL mapped to a new URL. Every piece of content accounted for. I verified every redirect with a custom script that checked status codes, followed chains, and flagged loops. I tested the entire map in staging. I tested it again. I tested it a third time because I am, despite the evidence you're currently reading, a thorough person.

The canonical tags were correct. The sitemaps were updated. The robots.txt was clean. The internal linking was rebuilt from scratch to reflect the new architecture. The page speed was better. The mobile experience was better. The content was better - we'd rewritten 40% of it during the migration to improve quality and relevance. We'd even updated the schema markup, because I'm the kind of person who updates the schema markup.

By every metric that the SEO industry uses to evaluate a migration, this was a good migration. A textbook migration. The kind of migration that gets written up as a case study and presented at a conference with a title like "How We Consolidated Five Properties Into One and Increased Organic Traffic by 47%."

Except we didn't increase organic traffic by 47%. We decreased it by 58%. And it stayed decreased. For fourteen months. And counting.

The Autopsy

I spent two weeks after David's call doing what I should have done before the migration: understanding why this specific site worked the way it worked, rather than assuming it worked the way sites in general work.

Here's what I found.

The spaghetti architecture - the fragmented, chaotic, multi-platform mess that I'd looked at and immediately diagnosed as a problem to be solved - was not a bug. It was, in the specific and peculiar context of this specific and peculiar company, a feature. The fragmentation wasn't hurting them. It was protecting them.

The WordPress blog had been accumulating backlinks for eight years. Not high-quality backlinks, necessarily, but many backlinks, from many different domains, pointing at many different URLs. The link profile was distributed across hundreds of individual blog posts, each of which had its own small cluster of referring domains. No single page was a powerhouse. But collectively, the blog was an authority engine, generating domain-level trust through the sheer breadth of its link distribution.

The knowledge base on the subdomain had a different link profile entirely. It was linked from Stack Overflow. From GitHub issues. From developer forums. From technical documentation on other companies' sites. These were high-quality, high-relevance, impossible-to-replicate links that came from real developers referencing real technical content. The knowledge base didn't rank for commercial keywords. It ranked for technical queries that drove developer traffic that converted at a rate three times higher than the blog traffic.

The Medium posts - the ones I'd dismissed as "duplicate content" that needed to be consolidated - were being syndicated and shared across communities that never would have found the main site. They were functioning as a distribution channel, not a content property. Their value wasn't in the SEO equity they carried but in the audiences they reached.

When I consolidated everything, I didn't just merge the content. I merged the link profiles. And when you merge link profiles, you don't add them together. You average them. Three hundred blog posts, each with 10-20 referring domains, got redirected to 60 consolidated pages. The link equity didn't triple. It diluted. Each new page had more links pointing at it, technically, but the links were pointing at redirect chains, and the redirect chains were passing equity at a diminished rate, and the pages they landed on were different enough from the pages they'd originally linked to that the contextual relevance of the links was compromised.

The knowledge base links from Stack Overflow and GitHub - the most valuable links in the entire portfolio - were now pointing at URLs that returned 301 redirects to a different domain on a different CMS. The links still passed some equity. But the signals were muddied. A link from a Stack Overflow answer about a specific API endpoint, pointing at a knowledge base article about that specific endpoint, now pointed at a redirect that landed on a consolidated "documentation" page that covered twelve endpoints. The specificity was gone. The contextual match was gone. The thing that made those links valuable was gone.

I didn't break the redirects. I didn't lose the content. I didn't make any of the obvious mistakes. I made the sophisticated mistake. The mistake that requires expertise to make. The mistake that a junior SEO couldn't have made because they wouldn't have had the confidence to recommend a migration of this scope in the first place.

That's the part that keeps me up at night. Not that I failed, but that the failure was a product of my experience, not a deficiency in it.

The Expertise Trap

There's a concept in psychology called the Einstellung effect. It describes what happens when your existing knowledge actively prevents you from seeing a better solution. You have a hammer. You've used the hammer successfully a hundred times. You encounter a new problem. The problem looks like a nail. You reach for the hammer. But the problem is actually a screw, and the hammer makes it worse, and you keep hammering because the hammer has always worked before and it is inconceivable to you that this time it won't.

Everyone knows the Dunning-Kruger effect - the curve where beginners overestimate their competence because they don't know enough to know what they don't know. What people talk about less is the second peak on that curve. The expert peak. The point where you've been right so many times, where your framework has been validated so consistently, where your experience has been rewarded so reliably, that you stop questioning whether the framework applies to the current situation. You just apply it. Confidently. Because you're an expert.

I was at the expert peak. I'd done migrations before. I'd consolidated fragmented sites before. I'd seen the traffic graphs go up after consolidation, the hockey-stick charts, the case studies with percentages in the titles. I had a framework. The framework worked. I applied it.

But frameworks are generalizations. And generalizations are, by definition, approximations of reality that sacrifice accuracy for applicability. "Consolidation improves organic performance" is true in general. It was not true for David's company, with David's link profile, distributed the way David's links were distributed, pointing at the specific URLs they were pointing at, carrying the specific contextual relevance they were carrying.

I didn't ask. That's the thing. I didn't ask "does this framework apply here?" I didn't ask "what's different about this site that might make the standard approach dangerous?" I didn't ask "where does the authority actually live, and what will happen to it when we move it?" I didn't ask because I was an expert, and experts don't ask. Experts prescribe.

I looked at the spaghetti and saw a mess. A person who knew less than me might have looked at the spaghetti and seen something I couldn't: a system. An ugly, unplanned, accidental system, but a system that worked. A system that had evolved over eight years to occupy a specific niche in a specific way, like a mangrove forest that looks chaotic from above but whose root structure is the only thing holding the coastline together.

I paved the mangrove forest and built a parking lot. A very nice parking lot. Great URL structure. Beautiful architecture. Clean redirects. And the coastline eroded.

The Confession

I want to tell you that I learned this lesson and never made the mistake again. I want to tell you that I now approach every client engagement with appropriate humility, that I question my frameworks, that I treat my experience as a starting point rather than a conclusion. I want to tell you that the $2 million mistake turned me into a better consultant.

And it did. Sort of. In the way that all painful experiences make you better, which is to say: slowly, incompletely, and with frequent relapses. I still catch myself applying frameworks without questioning them. I still catch myself being confident when I should be curious. The difference is that now, sometimes, I catch myself. Before David's call, I didn't even know there was something to catch.

The more insidious lesson is this: the $2 million mistake wasn't a mistake in the way we usually think about mistakes. I didn't do anything wrong. Every individual decision I made was defensible. The redirect map was perfect. The content strategy was sound. The technical execution was flawless. If you showed my migration plan to a hundred SEO professionals, ninety-five of them would approve it. The other five would say "looks good, but did you check the link distribution?" and they would be the ones worth listening to, and I was not one of them.

The system worked perfectly. The system was wrong. And the system was wrong not because of a flaw in the system but because of a flaw in the assumption that systems are universally applicable. That what works in general works in specific. That expertise in a domain means expertise in every instance of that domain.

It doesn't. It can't. Every website is an ecosystem. Every ecosystem has its own rules. And the rules of one ecosystem are not the rules of another, even when the ecosystems look similar from the outside, even when they share the same CMS and the same industry and the same target keywords. The way authority flows through a site - the specific, idiosyncratic, path-dependent way that links and content and history have combined over years to create a particular pattern of organic performance - is as unique as a fingerprint. And you can't consolidate a fingerprint.

The Industry's Problem

I'm not telling you this story to flagellate myself, though the flagellation is warranted and, honestly, slightly therapeutic. I'm telling you this story because the mistake I made is the most common serious mistake in SEO, and it's the one we talk about least, because acknowledging it undermines the entire value proposition of expertise.

The SEO industry sells frameworks. That's what we do. We take the complexity of Google's algorithm and the chaos of the internet and the unpredictability of user behavior, and we reduce it all to frameworks. Frameworks for keyword research. Frameworks for content strategy. Frameworks for technical audits. Frameworks for link building. Frameworks for migrations. Frameworks for everything. And then we sell the frameworks, confidently, because we've used them before and they've worked before and therefore they'll work again.

Except when they don't. Except for the cases where the framework doesn't fit, where the specific situation is the exception to the general rule, where the thing that works for most sites is the thing that kills this site. And those cases are more common than we admit, because admitting them means admitting that our expertise is contingent, that our frameworks are approximate, that the thing we're selling is not certainty but probability, and probability comes with a downside that no case study ever mentions.

We present our predictions as forecasts when they're really horoscopes. We present our frameworks as science when they're really heuristics. We present our confidence as competence when it's really pattern-matching, and pattern-matching fails precisely when you need it most: at the edges, in the exceptions, with the clients whose situations don't fit the pattern.

David's company was an exception. I should have seen it. The signs were there. The fragmented architecture that shouldn't have been working but was. The link profile that was distributed in an unusual way. The knowledge base with the developer community links that couldn't be replicated. All of it was visible to someone who was looking for exceptions rather than confirmations. I wasn't looking for exceptions. I was looking for confirmations. I saw the mess and confirmed that it needed to be cleaned up. I saw the fragmentation and confirmed that consolidation was the answer. I saw what I expected to see because my expertise had taught me what to expect.

That's the trap. Not incompetence. Not negligence. Expertise. The thing that makes you good at your job is the same thing that makes you dangerous. The more you know, the more patterns you recognize, and the more patterns you recognize, the more likely you are to see a pattern where there isn't one, to apply a solution that fit the last problem to a problem that is different in ways you can't see because your expertise is telling you it's the same.

What I Do Differently Now

I have a checklist. I hate checklists. I've written about how crowd behavior and template thinking distort outcomes. But I have a checklist because the alternative is trusting my instincts, and my instincts are the things that cost David's company two million dollars.

The checklist has one item on it. Just one. Before I recommend anything significant - a migration, a consolidation, a major restructure, anything that changes the fundamental architecture of how a site works - I ask myself:

What is working about this site that I can't explain?

Not "what's wrong with this site." I can see what's wrong. What's wrong is obvious. What's wrong is what jumps out at you when you run a crawl, when you look at the architecture, when you compare it to the framework you carry in your head. What's wrong is easy to find because the framework tells you what wrong looks like.

What I need to find is what's right. What's working. What's producing results that the framework doesn't predict. What's generating traffic or links or authority in ways that don't map to the standard model. Because those are the things that will break when I apply the framework. Those are the mangrove roots. Those are the things I'll pave over if I'm not paying attention.

For David's company, the answer would have been: the link distribution. The knowledge base links. The Medium syndication. All working, all generating value, all invisible to the standard migration framework. If I'd asked the question, if I'd looked for what was working instead of what was wrong, I would have seen it. I would have recommended a different approach. Not consolidation but selective consolidation. Keep the knowledge base on the subdomain. Keep the link equity where it lives. Consolidate the blog, sure, but carefully, preserving the distributed link architecture that was quietly doing the heavy lifting.

Would it have been perfect? No. It would have been ugly. It would have left some of the spaghetti on the wall. The Gantt chart would have been less clean. The executives might have nodded less enthusiastically. But it would have preserved the two million dollars that I vaporized with my confidence.

I've used the checklist eleven times since David's call. Three times, it changed my recommendation. Three times, I found something working that I couldn't explain, something that the standard framework would have broken, something that would have cost someone real money if I'd applied my expertise without questioning it.

Three out of eleven. A 27% rate of near-misses. More than one in four times, my initial expert recommendation would have caused significant damage. Think about that. Think about what it means for the industry. Think about how many migrations are happening right now, recommended by experienced consultants, executed with perfect redirects and clean architectures and beautiful Gantt charts, that are quietly destroying the specific, idiosyncratic, impossible-to-replicate things that make those sites work.

The textbook is not wrong. The textbook is incomplete. And the distance between incomplete and wrong, measured in dollars, is two million. Give or take.

The Last Call

David and I still talk sometimes. Not often. The relationship is what it is - we both know what happened, and we both know that "I'm sorry" is a packing peanut, and we both know that there isn't a better word available. He didn't sue me. I sometimes wish he had. A lawsuit has a resolution. A settlement. A number that closes the account. What David and I have is an open balance, unresolvable, the kind of professional debt that accrues interest in the form of 3 AM recollections.

His company is doing fine, by the way. Not because of the migration and not despite it. Because they have a good product and good people and the kind of market position that can absorb a two-million-dollar hit and keep going. They eventually recovered their organic traffic, eighteen months after the migration, not because of anything I did but because they kept producing good content and Google eventually figured out the new architecture and the world moved on. The recovery wasn't a triumph. It was regression to the mean. They got back to where they'd been. Fourteen months late and two million dollars lighter.

Sometimes David sends me an article about SEO migrations. A case study. A best-practices guide. A conference talk. He doesn't add commentary. He just sends the link. I open it and I read about redirect maps and canonical tags and crawl budgets and architecture best practices, and I read about how the migration increased organic traffic by 47% or 63% or 120%, and I look for the asterisk, the footnote, the part that says "but also, here's what we almost broke and didn't notice." It's never there. The case studies never mention what was working before. They only mention what changed after. Survivorship bias in a slide deck, presented as expertise.

I've been in this industry for more than twenty years. I've done good work. I've helped companies grow. I've made people money. And I've cost someone two million dollars because I was confident and competent and experienced and wrong. Not wrong about the theory. Wrong about the application. Wrong about the assumption that what I knew was sufficient to know what to do.

The most dangerous SEO practitioner is not the one who doesn't know what they're doing. The most dangerous SEO practitioner is the one who does.