Reputation is an asset you notice only when it drops. Noogle's system read billions of posts a year, so the question was never whether people were talking. It was which voices could move policy, the press or customers, and how much time the team had before a post spread.
The brief asked for a global monitor of key influencers on Twitter: catch the spikes in a post's virality, a topic's momentum or its press coverage, and keep a sentiment score for every person of interest towards Noogle.
The problemLoud is not the same as dangerous.
Counting mentions tells you a topic is loud. It does not tell you whether to worry. A politician with a low opinion of Noogle posting about a fine is a different event from the same words posted by an account nobody follows. So we turned the product around and kept a list of hundreds of thousands of key opinion formers instead, and watched what they posted.
The peopleThree readers, one alert.
Executives needed to know in a line whether to act, often on a phone. Four hundred stakeholders in about 20 countries read in the languages they chose. And the Newsroom, the team of data scientists, engineers, sociologists and journalists working with Noogle's Social Media Lab, had to check every alert before it went out.
A mention count tells you a topic is loud. A person of interest tells you who to call.
The decisionsAn alarm people would trust at 3 AM.
An alarm that fires too often gets ignored, and one that fires late is useless. My research into the human factors of alarm design came down to three decisions. Every alert carries one of three levels, set by potential impressions: low under 3 million, medium up to 10 million, high above that. Every alert says in a line why it fired. And every alert sets itself against past crises, so an executive can see at a glance whether this is a the leaked memo or a Cambridge Analytica.
Machine learning sorted each alert by how exposed Noogle could be, measured against hundreds of past scandals. Then a person read it. I wrote the policy they followed, six questions long, and nothing went out until all six were answered.
How big is 13.6 million?
Potential impressions in a crisis's first day, the yardstick every alert carried. The October 2019 alert is its first 48 hours.
The resultIt reached the room in time.
The whole run, from the post to the inbox, fit inside 30 minutes, and validated alerts reached C-level executives and 400 stakeholders in about 20 countries. We went from concept to MVP in three months, and the research behind the alarm design helped win the contract's renewal.
The system also found things nobody was looking for. Four alerts from the record show what the time bought.
The record
Four alerts that paid off.
The Forde alert
- Forde · February 2019A Grammy ad over a cover of the Rolling Stones drew mockery within minutes. Forde narrowed its reach and relaunched with another song.Self-reported
- Noogle · October 2019The publishers' open letter: the alert named the politicians and journalists, and the team avoided needless exposure.
- Noogle MapsA local campaign read as patronising. The team answered and relaunched with real people's testimonies.Self-reported
- Allied Leaf TobaccoVaping tricks trended at an Xbox launch. ALT knew in real time, before Microsoft addressed it.Self-reported
The alert family, redrawn in vector from my design files.