Teardown
Things You Should Never Do, Part I
Joel Spolsky, joelonsoftware.com, 2000. A classic — and still a draft with a lede problem. Read the way Friction reads any draft: structure first, commas later.
Friction predicts that 100 percent of readers start this post, that the share falls gently through the first 12 paragraphs to 72 percent, then drops sharply to 58 percent at paragraph 13 — the paragraph that assumes the reader remembers Borland and Quattro Pro — and then drifts down to 30 percent by the end. All numbers are predicted, not measured.
Note 1 of 3
The essay is hiding in paragraph 4
“They decided to rewrite the code from scratch” is the whole post in seven words. I’d open with that line and let the Netscape scoreboard follow as evidence, not as throat-clearing.
Note 2 of 3
The title promises a list. The list starts at paragraph 17.
“Things you should never do” sets up a listicle, and the four list items — knowledge, market leadership, shippable code, money — only land at paragraphs 17 to 20. A reader who came for the list quits before it starts.
Note 3 of 3
The cliff at paragraph 13 is context debt
“Philippe Kahn, the colorful founder of Borland” assumes the reader remembers a 1990s spreadsheet war. Anyone under 40 has no idea who that is, and this is where I predict they leave.
Reader proof
None here, and I won't invent a chart. This page carries no Beacon, so there's nothing measured to put next to the prediction. On a writer's own published post, the predicted map sits beside where readers actually quit.
Another one: read a teardown of Paul Graham's Maker's Schedule, Manager's Schedule.