Joseph Khan All articles
Site Notes

The Month My 404 Page Outdrew My Homepage

Joseph Khan
The Month My 404 Page Outdrew My Homepage

Photo: Mark Neyman, CC BY-SA 3.0, via Wikimedia Commons

I opened analytics on the first of the month with a lukewarm coffee and no particular agenda. Then I saw a line item that made me put the coffee down: /[404](https://en.wikipedia.org/wiki/HTTP_404).html sitting at the top of the page list. Not near the top. At the top. For thirty-one days in September, my error page pulled 1,412 views. My homepage pulled 1,309.

The error page was winning. I had a small, quiet crisis about it, then I started pulling data.

First, the boring explanations

Before you build a narrative out of a number, you have to rule out the dumb reasons. I checked three:

Where the traffic actually came from

Once I trusted the number, I traced it. Almost all of it fell into four buckets:

  1. A 2021 post that got shared hard. I wrote a long piece about rebuilding my site over a weekend. It got picked up by a couple of newsletters and a forum thread, and it kept getting linked for years. Then in 2023 I changed my URL structure and that post moved. Every one of those old links now pointed at nothing. That single post accounted for roughly 600 of the 404 hits.
  2. A dead project path. /projects/particle-toy was a tiny canvas experiment I killed in 2024. People kept linking it — including one Stack Overflow answer I have never been able to find and edit.
  3. URL guesses. A surprising number of people type /blog, /writing, /work, /resume, /hire. I have none of those paths. I have /notes, /projects, and /about.
  4. My own newsletter archive. Four issues from 2022 and 2023 linked to pages I later deleted. I broke my own links and never noticed, because I never click old issues. My readers apparently do.

The embarrassing part is not the 404 traffic. It is that the 404 traffic was a list of everything I had neglected.

What people did once they got there

My old error page said Not found in 32-point type with a single link back home. That was it. The data on it was bleak:

I had built a dead end and then acted surprised that people turned around.

Rebuilding the error page

I gave myself four rules and tried not to break them.

Assume they wanted something specific. Nobody lands on a 404 by accident and hopes to see your homepage. They wanted a post, a project, a file. So the top of the page is a search box, not a hero image.

Show the five pages people actually link to. Not my favorites — the ones with the most inbound links. That meant the site rebuild post, my notes index, the about page, my toolbox post, and a contact link.

Say sorry like a person. One line: This one is on me — the link you followed is dead. Then move on.

End with an exit that is not a bounce. I added a single email field. No popup, no exit intent, nothing that follows the cursor. It sits at the bottom under the heading If you want the next thing I write.

I also built a redirect map for the top twenty stale URLs, which dropped the 404 volume by about 40% in the first two weeks. That is the unglamorous fix, and it did more work than the redesign.

Did it turn lost visitors into subscribers?

Sort of, and with a caveat the size of a house.

Over six weeks after launch, about 1,100 people hit the error page. Thirty-four of them subscribed. That is roughly 3%, which is meaningfully higher than my homepage's 1.1% and wildly higher than the zero my old error page produced on purpose.

Thirty-four people is not a business. It is a decent dinner party. But the rate is the interesting part: people who arrive through a broken link are already looking for something. They are not browsing. They are on a mission, and if you hand them a plausible next step, a surprising number take it.

What I actually learned

The long tail of old links is not a historical artifact. It is a live part of the site, and it grows whether or not I pay attention to it. Every redesign, every slug change, every deleted project leaves a little scar somewhere on the internet that I do not control.

The other thing: a 404 is a landing page. It is often the first page a stranger sees, which means it deserves the same care as the homepage and gets none of it. Mine had been sitting there for years like a broken screen door.

I check the error page report every Monday now. It is the fastest way I know to find out what the internet thinks my site is supposed to be — and where I have let it down.

All Articles

Related Articles

Five Pages Gone, More Email: What Happened When I Cut My Portfolio Down

Five Pages Gone, More Email: What Happened When I Cut My Portfolio Down

A Week With JavaScript Off: What Actually Broke on My Portfolio

A Week With JavaScript Off: What Actually Broke on My Portfolio

Thirty Days Behind a Login Screen: I Gated My Blog and My Most Loyal Readers Let Me Have It

Thirty Days Behind a Login Screen: I Gated My Blog and My Most Loyal Readers Let Me Have It