◆Writing
Arguments we have already had.
On invalid traffic, what a verdict is worth, and why so much of this category is sold as a number nobody can check.
23 October 2026
Who Grades the Grader?
The vendor's report says the traffic is clean. Now ask who checked the vendor.
20 October 2026
Your Fraud Report Is a Horoscope
If the vendor cannot tell you what was wrong with the click, the verdict is not detection — it is a mood.
15 October 2026
Show Me the Rule
A model's verdict is not evidence; the rule it found is the only thing I can take to the seller.
8 October 2026
Models Find Patterns, Not Verdicts
The moment a model makes the call, you can no longer ask what was wrong with the click.
2 October 2026
Block the fraud is not an instruction
An instruction without a definition of acceptable loss is not an instruction; it is a prayer.
28 September 2026
The Customers You Never Knew You Lost
You pay for traffic, but the anti-fraud vendor only tells you about the fraud it caught, never the customers it threw away.
23 September 2026
A Fraud Score Is a Dodge
A probability tells you how sure the vendor is; it does not tell you what the click did.
16 September 2026
Your Fraud Threshold Is a Budget
A threshold is not a technical setting; it is a statement about what a customer is worth to you.
11 September 2026
A Slider Is Not a Fact
A confidence threshold tells the vendor how much uncertainty you will tolerate. It does not tell you what happened.
8 September 2026
Who Profits From Ambiguity?
When one party controls the definition, the evidence, the classification and the explanation, uncertainty flows in only one direction.
2 September 2026
What a Bot Actually Looks Like in 2026
The modern bot executes JavaScript, holds a session, and waits three seconds before clicking. You catch it on the contradiction it cannot avoid.
28 August 2026
"We block 99.9%" is a meaningless claim
A percentage nobody can audit is marketing, not measurement — and it never answers the only question that matters: 99.9% of what?