Hearing about it from the client
At 7.05am the client's CEO texts the account director a link: a national site ran a story overnight about a former employee's claims, it has been shared widely on social media, and two other outlets have picked it up. The agency's monitoring caught the first piece at 11.20pm. The alert sat in an inbox as part of a digest with forty other mentions, most of them harmless.
The CEO's next text asks why he found out before the agency did. There is no good answer.
Why overnight monitoring does not work as set up
- Monitoring alerts are set to digest emails to avoid flooding inboxes.
- Real-time alerts would fire on every harmless mention, so they are switched off.
- Social platforms move faster than news monitoring, and are watched less.
- There is no rota for who reads alerts at night.
- Keyword rules cannot tell a product review from a serious allegation.
The information usually existed in the agency's systems. It was not routed to a person who could act, in a way that person would notice.
What a late start costs
Hours lost overnight are the hours when a story spreads and other outlets pick it up. Starting at 8am means responding to a story that has already been written by others. The relationship damage with the client is harder to fix than the story: an agency is paid in part to see things first.
How we build out-of-hours alerts
- We take mentions from your monitoring service's API or alert feed and any social listening tool you use, in real time rather than as digests.
- Each mention is scored for risk by rules plus a language model such as OpenAI or Anthropic Claude, using each client's own risk terms: legal claims, safety, executives, regulators, redundancies.
- Weight is added for the reach of the source and for the same story appearing in several places quickly.
- Low scores go into the morning digest as now. Medium scores go to a Slack or Teams channel. High scores call and text the duty manager through a service such as Twilio until someone acknowledges.
- Each alert shows the headline, the source, why it scored high, and links to the client's contact tree and holding statements.
- Duty managers can mark alerts as false alarms, and those notes are used to tune the client's rules.
| Risk level | Example | What happens |
|---|---|---|
| Low | Product mention in a round-up | Morning digest |
| Medium | Critical review on a large site | Team channel, no phone call |
| High | Allegation involving an executive | Duty manager called until acknowledged |
| High and spreading | Same story on several outlets within an hour | Duty manager and backup called |
Thresholds are set per client with the account lead. Some clients want to hear about everything, others only about the serious end.
An overnight alert that reaches someone
At 11.25pm the duty manager's phone rings with an automated alert: a client named in an allegation on a national site, scored high because it names the CEO and mentions a tribunal. She reads it, starts the client's contact tree, and the client's head of communications is on a call before midnight. A holding line is ready if the morning programmes call.
The account director learns about it from the incident record at 6.30am, not from the CEO.
Quieter nights look different. A critical but ordinary review lands in the team channel and is picked up in the morning. A harmless mention of the client's name in a sports report goes to the digest. The duty manager sleeps, and trusts that the phone will ring if it needs to.
Is your overnight cover like this?
- Monitoring alerts arrive as digests.
- Real-time alerts were switched off because of noise.
- Clients have told you about stories before you knew.
- There is no rota for reading alerts at night.
- Social mentions are checked only in office hours.