I logged 155,000 searches through my own product and found out
I'd been marketing the wrong half
I build an AI search engine for WooCommerce stores. Over four months
my API logged 155,730 queries with actual text in them.
Under 3% came from a real store. The rest was bots hitting public
search endpoints. One source alone sent 65,000 queries.
But the 3% is what got me.
Every demo I have ever shown uses a sentence. "Something for oily skin
that isn't greasy." Those queries are real, I have them in the log.
Then I checked WHEN they arrive: clustered in the first day or two
after a store connects, in a burst, from one location.
That is the shop owner, testing. Not a customer.
What real visitors type, weeks later, steadily:
drifter salts
xros 5
epson emc800
Names, model numbers, part codes, usually misspelled. I built the
sentence half beautifully and spent months marketing it, and it is
the smaller half of what my customers actually get.
The other thing is worse. I found a query that came back empty on
eight separate attempts and concluded the store did not sell that
product. It did. I had not finished importing their catalogue. I was
one step from writing that up as a customer insight when it was a gap
in my own pipeline.
Same store, three different zero-result rates depending on what I
stripped out: 48.9%, then 15.3%, then under 7%. Only the last one was
about customers.
Two things I would tell anyone building on top of user data.
Your logs are full of machines, and machines are indistinguishable
from insight until you filter them out.
And check your own pipeline before you conclude anything about your
customer. The most confident wrong answer I produced this year came
from data I trusted because it was mine.
Full write-up:
