I Was Wrong About Where My Money Comes From

By Ryan Nichols

I had a number I was certain about. This week it broke.

The pattern

Five book sales in a row came in behind the same referrer. Every one of them said t.co. That is X.

FACT. Orders on August 30, August 30, August 31, August 31 and September 1 all carried t.co in their attribution data. The one before them, August 29, came from Google.

Five for five. In business that is not a coincidence. That is a finding.

So I wrote it down as settled. X makes the money. Facebook does not.

What a settled finding does to you

It makes you stop looking.

For a week my own notes carried the same sentence every morning. Every recent dollar came from X. Work X.

Then seven days went by with no sales at all. September 2 through September 7. Zero.

And I still had the same sentence in my notes.

The sixth sale

FACT. September 8. Early release digital edition, $29.99. Referrer: facebook.com.

Not X. Facebook. The channel I had quietly written off eight days earlier on the strength of five data points.

The numbers I should have pulled sooner

FACT. Last seven days on my own site, real humans only, after stripping out both the traffic my system tags as automated and a test suite that was hitting the site from a developer machine:

• Direct: 177 views, 111 sessions
• X, t.co: 102 views, 72 sessions
• Facebook: 97 views, 76 sessions

Facebook sent me more sessions than X did. Not fewer. More.

And here is the part that matters if you are checking your own: Facebook does not arrive under one name. It arrives under four. facebook.com, m.facebook.com, lm.facebook.com, and l.facebook.com. The mobile app and the link shim are separate rows. Read only the first one and you are looking at less than half your Facebook traffic.

What actually went wrong

Nothing was wrong with the data. The data was right the whole time.

What was wrong was that I took a real pattern out of a small window and let it harden into a rule. Five is a real signal. Five is not a law.

The moment it became a rule it stopped being a question. I quit checking the thing I was most confident about, which is exactly the thing that should get checked hardest.

What this proves

DOCUMENTED INFERENCE. Both channels work. Neither one is the whole answer. The seven quiet days line up with both of my engines being quiet, not with one platform dying.

What this does not prove

It does not prove Facebook beats X. One sale is one sale, and I just finished saying five is not a law. Six is not either.

It does not prove my attribution is complete. Direct is my single biggest bucket at 177 views, and direct means I do not know. Somebody typed the address in, or came from an app that strips the referrer. That is a hole in my picture, not an answer.

The part worth stealing

If you run a business and there is a number you are certain about, go pull it again today. Not the one that worries you. The one you are sure of.

Mine had four different names in it and I was reading one.

Most owners never find this, because nobody is looking. The report says what it said last month and everyone nods. The money is in the rows nobody queries.

---

If you want a system that tells you this before it costs you a week of sales, that is the work I do.

Work with me

Call or text (903) 500-8898