Trade Mirror

Uzbekistan Mirror Trade Evidence & Risk Screening Dashboard (v2) — a statistical reconciliation and risk-screening platform comparing Uzbekistan's import records with partner-reported exports. The central metric is the residual unexplained discrepancy; every result carries separate anomaly-strength and evidence-quality scores and an evidence-ladder position. Discrepancies are screening signals — never proof of wrongdoing.

Every import is reported twice: once by Uzbekistan, once by the partner that shipped it. When partners report exporting more to Uzbekistan than Uzbekistan records importing — beyond the normal CIF/FOB freight wedge (~10%) — a mirror trade gap appears. This project pulls both sides from UN Comtrade at HS2 (sectors) and HS6 (products), computes CIF/FOB-adjusted gaps and reliability-weighted screening scores, and presents them in an interactive dashboard with per-country investigation briefs and per-product profiles.

Scope: import side only, 2017–2024 (Uzbekistan began reporting to Comtrade in 2017; 2025 is annual-incomplete). Default view is the latest full year (2024). Only measured quantities are used — no assumed tax or tariff rates.

How it works

UN Comtrade API ──▶ scripts/fetch-comtrade.ts ──▶ data/raw/trade-rows.json
                                                          │
                                  scripts/build-analytics.ts
                                                          │
                                                          ▼
                                   src/data/*.json  ──▶  Next.js site (src/app)

• scripts/config.ts — methodology config: Uzbekistan (860), curated partners, high-risk HS chapters + 4-digit drill codes, the CIF/FOB band (6/10/15%), HS sections (categories), transit hubs, and the 2017–2024 analysis window. No assumed tax/tariff rates — the site reports measured trade values only.
• scripts/build-analytics.ts emits a granular dataset (cells.json per partner×commodity×year + meta.json); src/lib/dataset.ts aggregates it live so every filter (time, country, category, commodity, freight wedge) is exact.
• scripts/comtrade.ts — Comtrade API client. Uses the authenticated endpoint when COMTRADE_API_KEY is set, otherwise the public preview endpoint (capped at 500 rows/call). Caches every response on disk.
• scripts/fetch-comtrade.ts — pulls both sides (Uzbekistan-as-reporter and each partner-as-reporter) at HS 2-digit + 4-digit annually plus monthly for recent years. Reconciles partner codes against Comtrade's live reference (auto-corrects e.g. USA 842→841).
• scripts/build-analytics.ts — computes the mirror gaps, reliability tiers, revenue-at-risk, decomposition, concentration and unit-value spotlight, and writes the datasets the site reads.
• scripts/profile-data.ts — one-off profiler of raw-data coverage (weight availability, reporting years).

Setup

npm install

1. Get a free UN Comtrade key

https://comtradedeveloper.un.org → sign up → My Account → My API portal → Profile → Primary key. Free tier: 100k records/call, 500 calls/day.

cp .env.example .env
# then put your key in .env:  COMTRADE_API_KEY=xxxxxxxx

2. Pull data and build analytics

npm run data:all      # fetch from Comtrade, then compute analytics
# or individually:
npm run data:fetch
npm run data:build

3. Run the site

npm run dev           # http://localhost:3000
npm run build && npm run start

The site

View	What it shows
Overview	The four gap concepts (Gross / High-confidence / Core / Transit-sensitive) as clickable KPI cards, reform-annotated trend, top risk signals, "How to read this dashboard"
Risk Map	World choropleth; piecewise colour buckets; no-data & low-quality reporters greyed with explanation
Leaderboard	Every partner × commodity channel (HS2 or HS6), classified as Audit priority / Verify data / Transit-sensitive, with concentration Pareto
Products	HS6 product profiles: gap, top partners, yearly trend, unit-value check, interpretation & suggested priority
Sectors	Gap by HS2 chapter & category, sector × partner heatmap, price/kg under-valuation checks
Trends	Which goods and countries are getting riskier (or improving) over the full window
Partners	Investigation briefs: executive summary, key indicators, reporting quality, top sectors & products, trend, interpretation note
Methodology	The math, the four gap concepts, the risk-score formula table, interpretation guide, limitations

All analytical pages share a filter bar (view mode, years — default 2024, country, category, freight wedge) that recomputes everything client-side.

Methodology in one line

Import gap = partner_exports_to_UZB × (1 + CIF) − UZB_recorded_imports
Headline   = Σ max(0, gap) over partner × HS2 channels, reporting years only, 2017–2024

Positive = under-recorded on Uzbekistan's side. A gap is a signal of where to look, not proof — re-exports, transit hubs, lapsed reporters and timing differences all create legitimate gaps (see the Methodology page).

Data: UN Comtrade. Built with Next.js + ECharts. Deployable to Vercel.