By Amit Shuster, VP of Product, Vetrio.io
A camera crew is forty minutes from a checkpoint that was clear this morning. Someone standing at that checkpoint has already posted about it. Everything that keeps journalists alive in the field happens in the gap between those two facts.
A real-time risk intelligence platform closes that gap for a global news agency. Vetric is what that platform sees with: 10B signals analyzed every month, 99.9% coverage that holds when a story breaks, and the places where the first account often appears. The platform builds the alert. The safety team makes the call. We made sure nothing that mattered was missing.
What are we really protecting?
This is not brand monitoring or a news feed. The platform’s customer is a global safety function responsible for the physical safety of reporters, photographers, and camera crews working in active conflict zones and contested elections. A handful of workflows carry the weight, and they all run on social intel.
Route risk assessment, before anyone deploys
Before a crew is sent, the safety team assesses the route: what has happened along it, how recently, how severe, and how close to where their people will physically be. The platform serves this as a geospatial view of incidents, past, ongoing, and emerging, plotted against staff positions and planned travel routes.
That view is only as good as its inputs. In the places where a crew most needs the assessment, across Africa, Asia, and parts of Europe, the incident record is not in an archive anyone can search. It is in posts from people who were standing there, in the local language, minutes after it happened. Vetric is what makes that visible to the platform continuously, instead of as a research project that finishes too late to matter.
Live awareness while the crew is moving
Once the assessment is approved and the crew is in the field, the requirement inverts: no longer historical trend, now the next forty minutes. Developments in the towns ahead, a checkpoint, a closure, a strike, a crowd forming, have to surface while the route can still be changed.
This is the workflow where latency stops being a spec-sheet number. An alert that arrives after the wire confirms the event has no operational value: the crew has already driven through. The entire value sits in the gap between the first post from the ground and the first confirmed report, and that gap is measured in minutes.
Elections, and threats aimed at the reporter
Around contested elections the threat model changes shape. Journalists are targeted for who they are, race, gender, outlet, as much as for what they cover, and an agency covering an election at scale has a wide surface to defend.
Coordinated targeting, harassment campaigns, and mobilization form in the open on social platforms days before anything physical happens. Detecting that formation is a coverage problem before it is a modeling problem: the platform’s classifiers can only score what reaches them.
The consolidation problem underneath all of it
Before the platform, the agency’s security operations ran differently in every region: different tools, different databases, different definitions. Their security leadership wanted a single pane of glass: one place to pull metrics, run trend analysis, and push credible, actionable information to teams. The platform sells that consolidation. It can only deliver it if the layer underneath is itself consolidated and complete.
The requirement
Risk alerting for people in conflict zones sets a harder bar than most intelligence use cases. The platform’s engineering and intelligence teams were working against four constraints at once:
- Latency. An alert that lands after the wire story is worth nothing. The value is entirely in the minutes between the first post and the first confirmed report.
- Breadth. Journalists work everywhere. Coverage limited to the most obvious platforms leaves the highest-risk regions dark, and those are precisely the regions the risk assessment exists to cover.
- Reliability under load. Demand spikes exactly when a major event breaks. Degrading at that moment is the one failure mode the customer cannot tolerate, because that moment is when a crew is closest to it.
- Durability. Platforms change constantly. Every change absorbed in-house is engineering time not spent on the product the customer actually buys.
Building and maintaining that layer internally is possible. It is also a permanent engineering commitment that grows with every platform added and every change shipped upstream, and it competes directly with the roadmap.
The solution
- Coverage went live across the social intel that mattered operationally, including the places that carry ground truth where mainstream reporting thins out.
- The gap between something happening and the platform being able to see it came down to seconds.
- Availability held at 99.9%, including through high-volume breaking events.
- 10B signals a month, a scale already proven rather than projected.
- What happened before sits alongside what is happening now, so the same visibility serves both the pre-deployment assessment and the live alert.
- Platform-side change became our maintenance burden, not the partner’s incident queue.
- Compliance was built in from the start: sanctions screening on every customer, a defined list of prohibited use cases, and founder-level review before an account goes live. That mattered when the end customer is a news organization and procurement review is strict.
The results
Engineering attention returned to the product. The platform’s differentiation is its detection models, its mapping, and the workflow its customers live in but not the layer underneath. Vetric removes that layer from the roadmap, permanently.
Coverage that expands without a project. New platforms and regions arrive as part of the relationship rather than as a quarter of engineering work.
A defensible answer to the coverage question. When a safety director asks what the platform sees in a district where no wire service has a correspondent, there is an answer that holds up in a procurement review.
What it looks like when it works
None of this shows up on a dashboard. It shows up as a crew that took the other road, a deployment postponed by a day, a reporter briefed before she landed. The platform earns that. The safety team earns that. Our part is smaller and it is not optional: when the first person on the scene says something, it has to reach the people whose job is getting everyone home
