Know when to nudge, and when not to.
The hard part was never hitting send. It was sitting with a story and not knowing whether it deserved anyone's evening.
Talk to usThree possible outcomes
NUDGE TARGETED
The story is relevant to a defined segment. Send to people with strong affinity who are not overexposed or fatigued.
NUDGE BROAD
The story matters to a wide share of the audience. Send broadly, but still respect fatigue rules and quiet hours.
DO NOT NUDGE
The story is weak, mistimed, or the audience is too tired. Skip the send and protect the channel for something that earns the interruption.
What a recommendation looks like
Take a football story. You start with the whole list. Then you cut people who are tired, people who just got three sports pushes, people who have not opened anything in months. What is left is the send. The cuts need reasons, or the meeting goes nowhere.
Illustrative example
- Total browser audience
- 180,000
- Strong sports affinity
- 52,400
- Dormant readers protected
- −17,400
- High fatigue protected
- −9,200
- Recommended audience
- 33,800
- Decision
- NUDGE TARGETED
Decision inputs
- Topic and entities
- Geography
- Urgency and shelf life
- Topic and entity affinity
- Recent engagement
- Recent notification exposure
- Ignored streak
- Quiet hours and cooldowns
- Editorial policy
What editors see
Every recommendation shows the audience included, the audience protected, and the reasons behind both. Editors can review the message, send a test, override the decision, or skip with a logged reason. Nothing ships silently.
How the intelligence works
Language models help structure the story. Transparent policies and audience signals produce the recommendation. Editors remain responsible for the final send.
Know when to nudge.
Know when a notification is worth sending.
The useful question is who should not get this one.