Signal layer

Metrics & methodology

Every country score is built from public data, assembled into four axes. This page is the full index: what we measure, where it comes from, and why.

Normalization: All raw values are normalized to 0–100 using 5th–95th percentile clipping across the full 40-country watch list, preventing outliers from compressing the range. Direction is flipped where low coverage = high gap.

Fuse

mean(unemployment_youth, unemployment_male_total, violence, events_per_capita?)

Combustibility. Where idle military-age cohorts are concentrating and where things blow up.

Metric	Code	Source	Unit	Cadence	Direction	Notes
Youth unemployment	SL.UEM.1524.MA.ZS	World Bank	% of male labor force age 15–24	Annual	High = worse	Primary ranking metric. Leading indicator of instability.
Male unemployment (15+)	SL.UEM.TOTL.MA.ZS	World Bank	% of male labor force	Annual	High = worse	Broader base; corroborates youth figure.
Intentional homicides	VC.IHR.PSRC.P5	World Bank / UNODC	Per 100k population	Annual	High = worse	Proxy for structural violence and state fragility.
Recent conflict events	ACLED	ACLED API	Protest + violence events, last 90 days	Weekly sidecar	High = worse	Real-time unrest signal. Added to fuse when available.

Access gap

mean(electricity_gap, mobile_gap, ownership_gap, ai_gap)

How much of the pipe is missing: power, internet, devices, and inference. Low coverage = high gap.

Metric	Code	Source	Unit	Cadence	Direction	Notes
Electricity access	EG.ELC.ACCS.ZS	World Bank	% of population	Annual	Low = worse (higher gap)	The gate. No power, no pipe.
Mobile subscriptions	IT.CEL.SETS.P2	World Bank / ITU	Subscriptions per 100 people	Annual	Low = worse	Connectivity proxy. Can exceed 100 (multi-SIM).
Mobile ownership	ownership	ITU / Our World in Data	% of population owning a phone	Annual sidecar	Low = worse	Actual device access, not just subscriptions.
Internet users	IT.NET.USER.ZS	World Bank	% of population	Annual	Low = worse	Covers mobile and fixed broadband.
AI tool usage	ai_usage_est	DataReportal / GWI survey	% of population	Annual sidecar	Low = worse	Survey-based where available; internet × survey% proxy otherwise.

Impact

√(fuse × access)

Geometric mean of fuse × access. Requires both a burning fuse AND a real access gap. Punishes imbalance.

Metric	Code	Source	Unit	Cadence	Direction	Notes
Impact score	—	Derived	0–100	Each build	High = more urgent	Primary ranking for intervention priority. Top countries: Somalia 90, Yemen 87, Congo 85.

Reach

100 − freedom_score_priority

Intervention probability. How reachable the region is, shaped by political freedom.

Metric	Code	Source	Unit	Cadence	Direction	Notes
Freedom score	FH	Freedom House	0–100 (100 = most free)	Annual sidecar	High freedom = high reach	Higher freedom = more likely intervention lands. Not a moral filter, a logistics filter.

Context

—

Supporting indicators used for display and context. Not scored.

Metric	Code	Source	Unit	Cadence	Direction	Notes
Population	SP.POP.TOTL	World Bank	Total headcount	Annual	—	Used to compute per-capita event rates and scale reach.
Working-age share	SP.POP.1564.TO.ZS	World Bank	% ages 15–64	Annual	—	Demographic context. High share + high unemployment = maximum fuse.

Watch list

The 40-country watch list is the top 40 countries by male youth unemployment (World Bank, latest available year). Rankings are recomputed on each data refresh. The list is not editable. It is purely data-driven.