Predictive Intelligence
Built in Europe for the realities of modern information warfare, Repsense detects where a narrative is heading before it arrives — not by extrapolating volume trends, but by analysing the structural behaviour of how narratives form, coordinate, and spread. Early warning from the architecture of an operation, not from the spike it eventually produces.
90% narrative trajectory prediction accuracy across validated datasets · 3 days of lead time in the Kapciamiestis FIMI case - before mainstream media, before volume threshold
The Distinction
Structural Prediction vs. Volume Forecasting
Other platforms now claim predictive capabilities. It matters what they mean by that.
What competitors offer: volume forecasting
Meltwater’s 2025 year-end release introduced predictive analytics that "forecast whether a mention spike will grow or fade." Talkwalker claims 90-day volume forecasting. Signal AI offers "horizon scanning" for reputation risk. All of these analyse the shape of the volume curve - the rate at which mentions are accumulating - and extrapolate forward. If mentions are accelerating, the forecast says the spike will grow. If they’re decelerating, it says the spike will fade.
This is useful. It answers the question: "Is this trend getting bigger or smaller?" But it cannot answer the questions that matter more: Is this organic or manufactured? Is it being pushed by an organised operation? Who is behind it? What is its structural ceiling? And critically: what is developing below the volume threshold that will become the next crisis?
What Repsense offers: structural prediction
Repsense predicts narrative trajectory by analysing how narratives behave — not simply how fast they grow. Built in Europe for the realities of modern information operations, the platform detects the structural signals that appear before volume spikes: coordination patterns forming across platforms, content seeded simultaneously through multiple sources, amplification architectures activating around a narrative, and rising cross-platform duplication ratios. These indicators reveal when a narrative is not merely spreading organically, but being actively pushed — often long before conventional alert thresholds trigger.
The mechanism is different. Volume forecasting extrapolates the curve. Structural prediction analyses the architecture behind it. A manufactured operation with a coordinated amplification chain will spread faster and peak higher than an organic story of similar initial size — and Repsense identifies that divergence before it becomes visible in the data.
How It Works
Four Layers of Predictive Intelligence
Narrative Clustering & Trajectory Tracking
All monitored content is transformed into high-dimensional text embeddings that capture semantic meaning. Clustering algorithms detect groups of semantically similar content - narrative threads forming in real time. Each cluster is tracked over time: growth velocity, geographic spread, platform distribution, source diversity, and sentiment trajectory. This is the baseline: understanding what stories exist and how they are moving.
Coordination Detection
The platform analyses behavioural signals across content within and between clusters. Synchronised posting times across accounts. Duplicate or near-duplicate content appearing on multiple platforms within short windows. Amplification patterns that match known coordinated inauthentic behaviour (CIB) tactics. When a narrative cluster shows coordination signals above threshold, the platform flags it as potentially manufactured - regardless of its current volume.
Emerging Narrative Detection
Content that doesn’t fit existing narrative clusters - outliers in the embedding space - is continuously monitored for cluster formation. When outlier content begins to coalesce into a new cluster, the platform detects an emerging narrative before it has a name, a volume spike, or media attention. The earliest signal of a new story forming, surfaced automatically.
Trajectory Projection
Combining cluster growth dynamics, coordination intensity, platform spread patterns, source authority distribution, and historical pattern matching, the platform projects where a narrative is heading: predicted peak volume, estimated timeline to mainstream pickup, likely platform migration path, and structural ceiling. The projection is based on how the operation is built, how the operation is structured - which is why it provides lead time that volume extrapolation cannot.
What Predictive Intelligence Enables
From Detection to Decision
-
Detect narratives on a crisis trajectory before volume thresholds trigger. The response window between structural detection and mainstream pickup is measured in hours to days, not minutes. Proven: three days of early warning, two weeks before full escalation.
-
When negative coverage spikes, know immediately whether it is genuine public concern or a coordinated operation. The answer determines your entire response strategy. Coordination analysis provides this distinction at the moment of detection, not after a manual investigation.
-
For earned media campaigns, the same structural analysis projects whether coverage is building momentum organically or losing energy. Predict campaign trajectory before the reporting period closes.
-
Discover the conversation you should be in before it has a name. Outlier detection and cluster formation monitoring surface emerging themes from the information environment automatically - before competitors notice, before media picks it up.
-
Track how a narrative migrates across platforms - from Telegram to TikTok to mainstream news - and project its migration path based on structural patterns. Understand not just where the narrative is now, but where it will be tomorrow.
The Proof Point
Kapčiamiestis: Structural Detection in Action
In January 2025, Repsense detected a coordinated information operation targeting Lithuanian military infrastructure. The content volume was below any conventional monitoring threshold. No volume-based alert would have triggered. But the structural signals were clear: synchronised content seeding across Russian-language outlets, cross-platform duplication patterns, and amplification chains activating in a pattern consistent with known FIMI operations.
Repsense issued an early warning three days before the first mainstream media article appeared. Two weeks before full escalation. The same structural detection then traced the narrative from its origin through Italian, Croatian, Latvian, Bulgarian, and Belarusian outlets - reconstructing the full operation from a single initial detection.
This is what structural prediction delivers that volume forecasting cannot: detection at a point where the graph is flat, because the signal is in the behaviour, not the volume.
See What’s Coming Before It Arrives
Request a demo to see how Repsense’s predictive intelligence identifies narrative threats, emerging stories, and campaign trajectories from structural signals - before volume-based tools trigger.