Blog
The long versions of arguments we start on LinkedIn.
-
"In 1494 a Friar Wrote Down How Venetian Merchants Kept Their Books"
Luca Pacioli didn't invent double-entry bookkeeping. He printed it. The difference between a definition in one person's head and a definition anyone can reference is the same problem most companies have with their metrics today.
-
"Deming's Funnel, and Why Your Team Learns From Randomness"
W. Edwards Deming proved that adjusting a stable process after every miss makes performance worse. Most analytics teams do it monthly, and the cost is learning confident lessons from noise.
-
"\"We'll Clean the Data First\" Has No Stopping Condition"
Most data projects start with a cleanup phase that never ends, because cleaning without a downstream consumer produces an infinite list sorted by no criterion. The fix is to build the reporting first.
-
When GA4 Purchase Revenue Doesn't Match Your Payment Report
GA4 records purchase events; a processor records payment activity. Compare the same online orders on the same basis before using the gap as an alert.
-
"VisiCalc, 1979, and the Problem It Created"
Dan Bricklin's spreadsheet removed a bottleneck that was partly a bottleneck and partly a control. We've repeated the pattern three times since, and the scarce thing keeps being the same.
-
What a Wrong Number Costs, and Why Nobody Computes It
Companies can quote the cost of their reporting to the dollar. Ask what a wrong number cost last year and you get silence. The math is hard, and the incentives make sure it stays undone.
-
An Anomaly in the Numbers. Here's the Order to Check Things In.
When a number moves, most teams start with the most interesting explanation. The method that works starts at the database and branches from there.
-
Repeated Success With a Known Flaw Doesn't Raise the Alarm. It Lowers the Bar.
Diane Vaughan called it the normalization of deviance. NASA ran the mechanism twice, seventeen years apart. Data teams run it every week, and the tell is language.
-
An LLM Can Write the SQL. It Still Doesn't Know What Revenue Means.
Text-to-SQL got good fast. The bottleneck was never syntax. It was agreement about what your numbers mean, and that part has no vendor.