A customer wrote to ask why a campaign had cost twice the preview of the previous, identical one. It was not identical. The text had been edited in a word processor, and the straight apostrophe in "don't" had come back curly.
The GSM-7 alphabet holds 160 characters (153 per part when concatenated). It does not contain the curly apostrophe. One character outside the alphabet switches the whole message to Unicode, at 70 characters (67 per part when concatenated), and a 150-character text that fitted one segment now needs three.
What the composer does now
The segment counter has always shown the encoding. It now also names the character that forced Unicode and offers to replace the usual suspects, curly quotes, long dashes, the ellipsis character, non-breaking spaces, with their GSM-7 equivalents in one click. The API does not rewrite your text; it reports encoding and segments in the response so your own code can decide.
Checklist for single-segment copy
- Write in a plain-text editor, or paste as plain text.
- Straight quotes, plain hyphen, three dots rather than the ellipsis character.
- No emoji in transactional messages.
- Preview with the longest real row, not the template: a long name can tip a message over the limit.
For languages written in non-Latin scripts none of this applies: Unicode is the cost of the script, and the right move is to write for two clean segments rather than one and a half.
