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:
- Self-referrals. My own browser hitting stale bookmarks while I edited the site. Filtered by excluding my IP and my logged-in session.
- Bots. My host's raw logs showed a chunk of traffic from crawlers and preview fetchers. Once I filtered on real browsers with real referrers, the number dropped — but only to 1,102. Still bigger than the homepage.
- A tracking bug. I genuinely suspected this one. My analytics script fires a pageview on every route, including the error route, which means a 404 is logged like any other page. That is not a bug exactly, but it does mean the error page looks like content. Which, apparently, it was.
Where the traffic actually came from
Once I trusted the number, I traced it. Almost all of it fell into four buckets:
- 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.
- A dead project path.
/projects/particle-toywas 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. - 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. - 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:
- 68% of 404 sessions ended in under four seconds
- Median time on the page was six seconds
- Only 9% clicked the link home
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.