✳   A closer view of Venezuela. Our lead research programme is taking shape.Explore the programme
Our method / Evidence to judgment

Political science,
run as data science

Six analytical layers connect reporting to events, actors, incentives, and possible outcomes. The reasoning should be visible enough to challenge, and clear enough to use in a decision.

Six analytical layers

Different lenses
One evidence record

Each layer asks a different question of the same material. The value lies in seeing how the interpretation was reached and where it could be wrong.

Layer 01

NLP & entity recognition

Turn reporting into events you can examine

Identify the people, organisations, places, and actions in a report. Keep the original source beside the extracted event, so an analyst can check what the text actually establishes.

See how to use it

A statement that a terminal is open becomes a claim about a named place, activity, and time. It does not become proof that operations are normal.

Layer 02

Network analysis

Follow the relationships behind the decision

Connect documented authority, ownership, sanctions, and operating relationships. A useful graph shows the evidence for a connection and distinguishes a formal role from inferred influence.

See how to use it

An approval may involve a national ministry, a local authority, and a service provider. Map the dependencies before deciding which relationship matters most.

Layer 03

Drama theory

Ask whether a threat or promise is credible

Examine the positions in a confrontation: what each actor wants, what they threaten, and what would make another actor comply. Compare the declared position with the options actually available.

See how to use it

A threatened withdrawal matters differently when the actor has an alternative buyer, a financial constraint, or allies unwilling to follow.

Layer 04

Prospect theory

Examine how pressure changes behaviour

Consider whether an actor sees the situation as a gain to protect or a loss to reverse. Use that as a hypothesis about risk-taking, then test it against observed decisions.

See how to use it

A costly escalation may reflect an attempt to recover a deteriorating position. The explanation needs behavioural evidence, not an assumption about personality.

Layer 05

Bayesian inference

Let evidence change the scenario assessment

Start with explicit alternatives and prior judgments. Ask how much more likely a new observation would be under one scenario than another, then record the reason for any revision.

See how to use it

Several reports repeating one announcement are one evidentiary chain. Treating them as independent updates would create false confidence.

Layer 06

Revealed preference

Set the bar from decisions, not slogans

With the client, examine previous entries, declines, and stated constraints. Use that record to propose decision criteria, then confirm which requirements apply to the opportunity at hand.

See how to use it

Past decisions may suggest a minimum requirement for contract protection or access. They do not establish a universal threshold, or replace the client’s current mandate.

From source to judgment

How evidence moves the assessment

01 / ObserveReports become
reviewable events

Source, actor, place, time, and competing accounts.

02 / InterpretRelationships and
behaviour add context

Authority, incentives, dependencies, and alternatives.

03 / ReassessScenarios meet
the client’s criteria

What changed, what it means, and what needs review.

Analytical workflow illustration. Arrows describe the reasoning sequence; they do not represent data volumes or live feeds.
Reading the event record

A label needs
an explanation

Explore the intended treatment of evidence states. This is an editorial demonstration, not an automated assessment of live sources.

Evidence state

One evidentiary chain, however many headlines

Reports trace back to one source or source family. Repetition does not make the underlying account independently confirmed.

METHOD DEMONSTRATION / NO PROBABILITY ASSIGNED
What to establish

Identify the original account, its time and place, and how the source knows.

Analyst’s next step

Seek an independently obtained account or a different collection method.

Freshness is separate

A record can be corroborated or contradicted and also be stale. Review windows depend on event type; freshness does not settle the disagreement.

The hard ask

No soft answers

Ask what supports the judgment, what challenges it, and what would change the answer. A useful briefing should be able to show its reasoning.

What would make you change your mind?

Name the observation, the relevant scenario, and the reason it would change the assessment. Set a review date. If the evidence is insufficient, say what is missing.

Does lower country risk mean the opportunity is viable?

Not necessarily. Country direction and commercial readiness are separate judgments. Test the particular contract, operating dependencies, and institutional requirements against the client’s criteria.

How do we know the method adds value?

Agree what a first briefing must demonstrate. For forecast performance, inspect dated forecasts, revisions, outcomes, and a consistent benchmark across the record, including unsuccessful calls.

Put intelligence in context

What would better evidence
change for you?

Tell us your sector, the decision you face, and what a useful first briefing would need to demonstrate.

See how we can help you