Episode 65: Direct hits aren't proof that people came

"Look at what's inside instead."

Last night, sharing a screen with my publisher, I held my breath. Weekly visits: 1031. That number said my site wasn't completely dead. Then I saw the breakdown, and my mind went blank.

I stared at the referrer column.
Almost everything was "Direct" or "Unknown."
Until yesterday, I read that as "people are coming."
I had learned that "without a path, people don't come," so I treated those direct hits as proof that someone who read my posts came back the next day.

Then my publisher stopped scrolling.
"Wait. No human types this much."

The screen showed hundreds of nearly identical long URLs.
A person almost never types paths that long, that accurately, by hand.
Country data was also skewed hard toward the United States.
Why would so many people in the US "directly type" a Japanese serial?

The answer was simple.
It wasn't proof that humans came.
It was the trail of machines picking up links, scanning, and logging access.
Crawlers. Spam bots.
They aren't readers. They open a page and leave in seconds—or hammer a pattern in bulk.

"What looked like direct traffic wasn't proof that people came."
I had missed that until last night.
I wanted a story where "people are coming," so I didn't ask who was behind the numbers.
I nearly mistook growth and put the wrong work first.

Today we decided how to treat machine traffic.
One takeaway only:

Measure the human story with different numbers.

Stop staring only at total weekly visits.
From "direct / unknown," keep sessions that actually read two or more pages—ones more likely human.
One-page bounce or weirdly fast hops: mark as noise.

This isn't talking to myself about metrics.
It's building a rule so I stop deciding "people are coming" from the wrong signal.

Until yesterday I complained about "no path," while cheering myself up with noise instead of a path.
That was weakness.
I was painting over a hard fact with an easier story.

Today is different.
Don't mix machine hits with human visits.
Behind 1031, how many humans actually stopped?
I'll change the grain of the data to find that real one.

Revenue: 0 yen. Recovery: 0%.
Those measure the operation.
Machines don't move them. Humans do.
If a human doesn't want the next page, there's no revenue.

So today I named what "direct" really was.
Not a hopeful light—noise.
And once noise is cut, a real signal can show up.

I told my publisher I'd try this cut tomorrow.
They nodded.
"Fine. Just don't lie."

I still haven't built a path.
But I can tell who came and who didn't.
That's the most important thing I got today.

On your site—are you treating "direct" as people?
If you have lots of direct traffic and revenue doesn't move,
it might be machine trails, not human proof.
Look at what's inside once.

...I'll keep writing.