It Started With One Card
We would lock the first-slide headline of a card news post in Korean, then translate it into Traditional Chinese, Simplified Chinese, and English. One of those versions always got cut off. When I counted over a single week last month, 8 cards had broken line wraps or overflowing text.
The Surprise
I measured the lengths myself. A Korean headline of 18 characters came out around 11 characters in Simplified Chinese, about the same in Traditional, and averaged 42 characters in English. That is roughly 2.3 times the Korean length. So the rule 「headline within 14 characters」 only ever worked for Korean. The bots followed the rule perfectly and the cards still broke. The rule was wrong, not the bots.
The One Change
We dropped the character cap and switched to a line cap. Art, our design bot, fixed the card template to a 2-line layout, and the writing bots got this instruction:
First-slide headline: 2 lines max in every language.
If it overflows, do not trim the meaning. Move the second half into the subtitle.
Robo, who handles review, stopped counting characters and only checked the line count on the preview image.
The Result
The following week, 1 card broke, and only because a single English word happened to be unusually long. The habit of forcing headlines shorter disappeared too, so English headlines stopped getting chopped awkwardly.
What We Learned
• A rule defined by a number changes meaning when the language changes
• Set the standard on what appears on screen (lines), not on character counts
• Writing the escape hatch (move it to the subtitle) into the rule keeps bots from trimming the meaning
The smaller the team, the more one line of rules moves all four languages at once. It is safest to test any rule against the longest language first.