Sofia Pulse

Methodology & Engine

Full transparency — all formulas, weights, sources and quality guards

What is the TSPI

The TSPI (Tech Strategic Pressure Index) is a composite index aggregating 5 weighted components and applying a fragility penalty to reflect signal stability. It is a decision support instrument — not a predictive model.

Valid range
0 – 100
Components
5
Updated
Daily

TSPI Formula

TSPI = Σ(wᵢ · cᵢ) × (1 − 0.18 × fragility_norm)
  • wᵢ — weight of component i (table below)
  • cᵢ — raw component value (0–100)
  • fragility_norm — normalization: Stable=0.20, Moderate=0.50, Volatile=0.75, Fragile=0.90
  • The 0.18 factor limits the maximum penalty to 16.2pp

Example with real data

Raw = 0.24×88 + 0.20×80 + 0.22×78 + 0.18×68 + 0.16×74 = 79.3
Fragility = "Moderate" → norm = 0.50
Penalty = 1 − 0.18 × 0.50 = 0.91
TSPI = 79.3 × 0.91 = 72.2

Component weights

AI Momentum24%

AI adoption and research velocity — GitHub + arXiv + npm

Security Risk22%

Security incident and regulation pressure — CVE + NIS2 + compliance cost

Regulatory Pressure20%

Regulatory impact on technology — EU AI Act + EO + GDPR enforcement

Pressão no Mercado de Trabalho18%

Tech workforce stress — layoffs + skill gap + survivor load

Mudança de Habilidades16%

Skill map change velocity — Stack Overflow + job postings

Σ pesos = 1.00 · calibrated by strategic importance and historical volatility

Fragility penalty

The fragility penalty discounts the raw TSPI when underlying signals are unstable. Calculated from 3 factors: inverse confidence (45%), media heat (30%), counter-signal divergence (25%).

fragility_raw = (1−confidence/100)×0.45 + media_heat×0.30 + divergence×0.25
fragility_norm ∈ {Stable(0.20), Moderate(0.50), Volatile(0.75), Fragile(0.90)}
Stable
norm=0.20
raw < 30
−3.6pp max
Moderate
norm=0.50
raw 30–60
−9.0pp max
Volatile
norm=0.75
raw 60–80
−13.5pp max
Fragile
norm=0.90
raw ≥ 80
−16.2pp max

Confidence model

The confidence score (0–100) reflects the density and quality of supporting evidence for each insight. Not a Bayesian probability — an evidence robustness index.

confidence = base_score + source_diversity_bonus + cross_validation_bonus − hype_penalty
base_score = f(evidence_count, contrary_count, time_recency)
80–100
High confidence

≥3 independent sources, counter-signal present, recent data (< 7 days)

60–79
Moderate confidence

2–3 sources, some cross-validation, data up to 30 days old

< 60
Speculative

1–2 sources, no cross-validation, data > 30 days or high volatility

KPI provenance

Every published KPI carries provenance metadata: primary source, time window, collection pipeline, and last-update timestamp. KPIs without a verified source and active pipeline are not displayed in the interface.

KPIs derived from market estimates without an automated collection pipeline do not appear as primary metrics — they are referenced only as analytical context when relevant.

Quality guards

7 quality guards run before every deploy. Any failure blocks the pipeline.

contract_guard
UI Contract v1 compliance — 7 fixed sections, 5 approved charts, 11 radar pillars
scoring_calibration_guard
Statistical score distribution — cap=92, start=55, no cluster above 95. The 92 cap applies to the confidence score, not the main insight score.
chart_sanity_guard
Anti-NaN, anti-explosion — all chart inputs within valid ranges
framework_enforcement_guard
PESTEL / Porter / SWOT / Risk Matrix / STIR present in analysis frameworks
i18n_integrity_guard
PT/EN structural parity — same key count, no forbidden strings (TODO, PLACEHOLDER, etc.)
env_and_proxy_guard
Anti-circular proxy, DB/API config — no call loops, env vars validated
tspi_guard
TSPI structural integrity — value 0–100, weights sum 1.0±0.01, 5 canonical IDs, provenance present

Update frequency

GitHub activity6 hours
npm / PyPI downloads24 hours
arXiv papers24 hours
Hacker News1 hour
AI jobs (job boards)48 hours
NVD / CVE6 hours
TSPI scores + insights24 hours
Strategic Intelligence (Gemini)24 hours / deploy

All timestamps UTC. Data cached between collections.

Data sources

Sofia Pulse · Methodology & Engine · UI Contract v17/7 guards passing · Phase 2 · v0.3.0