We tested 300 versions of the same alert. The winning format is shorter than you think and starts with a verb.
An alert that nobody reads is worse than no alert at all. It teaches people to ignore the channel it arrives in.
What we tested
Over six months we sent design partners different versions of the same brief: long and short, with and without charts, written as a headline or as a question. Then we measured which ones led to an action within a day.
What won
- One sentence first. What changed, by how much, since when.
- The cause second. Even a likely cause beats no cause.
- A name third. Briefs with a suggested owner were acted on twice as often.
Charts helped only when they showed a single line against its normal range. Anything busier was skipped.
What we stopped doing
We removed severity colours, emoji and the word “urgent”. People trusted the briefs more when they sounded calm.
