The Gap Nobody Talks About
Every FIMI detection framework in use today — EEAS, EU DisinfoLab, DISARM — correctly focuses on detecting coordinated inauthentic behaviour. They identify narratives, map amplification networks, tag tactics with DISARM codes. This is necessary work. But it answers only half the question.
The other half: is the operation actually working?
An influence operation that reaches 10 million accounts but changes nobody's mind is a fundamentally different threat than one that reaches 50,000 accounts and shifts measurable opinion in a swing constituency. Current tooling treats both identically. Retelnist does not.
Introducing V(x,t): The Belief-Shift Velocity Score
Our core effect-measurement metric, V(x,t), combines four components into a single operational effectiveness index:
- Frequency (F): Raw narrative volume across monitored platforms (Telegram, X, Facebook, YouTube, Reddit, RSS, LinkedIn)
- Sentiment Polarity (S): Direction and intensity of emotional valence, tracked per narrative cluster
- Identity Coupling (I): Degree to which a narrative has been attached to in-group/out-group identity markers — the strongest predictor of belief entrenchment
- Amplification Asymmetry (A): Ratio of organic reach to coordinated amplification, flagging artificial velocity
The formula is weighted and normalised against a 30-day rolling baseline, reported with 95% confidence intervals derived from bootstrap resampling across 1,000 iterations.
V(x,t) is not a sentiment score. It is a measurement of how fast a given narrative is embedding itself into the belief structures of a target population.
Why Confidence Intervals Matter
Intelligence products without uncertainty quantification are not intelligence products — they are opinion pieces formatted to look like data. This is a problem endemic to the disinformation research space.
When Retelnist reports that a narrative has a V(x,t) of 0.73 ± 0.08 (95% CI), that interval is load-bearing. It tells the analyst how much trust to place in the figure, whether the signal is above noise, and whether an observed shift is statistically distinguishable from random variation.
We apply permutation testing to rule out autocorrelation artifacts in time-series data — a methodological step that most commercial platforms skip entirely.
Operationalising Measurement for StratCom Units
The practical implication of effect measurement is resource allocation. A national StratCom unit with limited counter-narrative bandwidth cannot respond to every flagged operation. V(x,t) provides a triage signal: address operations with high velocity and rising identity coupling first; monitor low-velocity operations without immediate response.
This is the difference between a reactive posture (respond to everything detected) and a strategic posture (respond to what is actually moving the needle).
What This Looks Like in Practice
In our deployments, clients typically discover that 20–30% of detected narratives account for over 80% of measurable belief-shift effect. The remaining 70–80% of detected operations are noise: real coordination, real effort by adversarial actors, but zero measurable population-level impact.
Responding to that noise wastes counter-narrative resources and — more dangerously — can amplify low-reach narratives by drawing attention to them.
Measurement discipline is not a nice-to-have. It is the foundation of proportionate, effective response.
The CWPI: Combining Detection and Effect
Our Cognitive Warfare Presence Index (CWPI) combines Layer A (DISARM-aligned detection) with Layer B (V(x,t) effect scoring) into a single composite metric. It answers the operational question directly: How much cognitive warfare pressure is this environment currently under, and where is it concentrated?
CWPI reports are generated weekly for standing clients and on-demand for crisis situations, with statistical lag analysis identifying whether pressure is increasing, stable, or declining.
We believe this methodology represents the current state of the art in operationalised cognitive security measurement. We also believe it can be improved — and we publish our methodological notes for peer review precisely because reproducibility and auditability are non-negotiable properties of serious intelligence work.