Assessment Methodology

This page is the authoritative method behind every business case — how benefits and costs are calculated, the assumptions, and the sources. It is self-contained: the personalisation prompts rely on this page (and the per-stakeholder pages), not on any repository file. Detailed per-country metrics and derivations live in the /research folder of the project repository; the authoritative external sources are listed at the end.

How benefits are calculated — the attribution chain

Each research category sizes a total addressable pool, and most cite a digital-addressability fraction. But that fraction is a broad basket — single windows, e-customs, AI risk engines, e-payment. Crediting all of it to verifiable documents would over-claim. So every lever decomposes:

pool  ×  digital-addressable %  ×  UNVTD/GRID share  ×  adoption ramp  =  attributable value
        (research-given)          (the haircut)        (uptake)

• UNVTD/GRID share — the portion of the digital basket realised specifically by verifiable documents + verifiable identity (not the whole basket). Central 45%, range 30–60%, applied uniformly. This is the single largest source of uncertainty, so every figure is presented as a range. (Lever ④ is an exception — its share is data-derived; see below.)
• Adoption ramp — realistic uptake at a 10-year horizon: Conservative 30% · Central 50% · Ambitious 70%.

The value levers

share = UNVTD/GRID share 0.45 (range 0.30–0.60). adoption = 0.30 / 0.50 / 0.70.

Lever-specific notes:

• ② documentary share (60%) — verifiable documents attack the valuation / misinvoicing portion of leakage, not physical smuggling.

• ② effective vs nominal tariff — size the pool on the effective collected tariff (duty collected ÷ imports) where the authority’s own figures give it; the published nominal / income-band tariff is only a fallback. In FTA-heavy economies the effective rate can be far below nominal (e.g. Thailand ~1% collected vs ~10% nominal), so prefer the effective rate or apply an FTA/exemption haircut, and state which you used.

• ③ compliance share (60%) — UNVTD attacks the compliance gap (under-reporting, missing trader), not the policy gap (exemptions, reduced rates). Value splits within the lever: invoice issuance (Buy-Sell) vs invoice↔settlement matching (Payment & Reconciliation) — counted once.

• ②③ pool rate — use the central estimate, not the floor. Leakage (②) and the VAT/CIT compliance gap (③) are measured from random samples that themselves understate the recoverable pool: audit applies non-detection multipliers (it misses some), and selection bias means measured leakage never sees what audit never reaches. This is the “strong, specific evidence” the conservatism rule allows — so for these two pool rates take the central value of the band (② leakage at the mid of the income-band range; ③ VAT gap at/above 9.5% and CIT toward the middle), not the bottom, even for high-compliance economies, which retain a real pool (Australia’s random sample: ~32% error-lines; ~$0.25bn/yr recovered via voluntary disclosure alone). Leave the documentary/compliance 60% share and the UNVTD/GRID-share + adoption haircuts unchanged — this changes only where in the leakage/gap band you start. See research/compliance-assurance/coverage-recoverable-pool-findings.md.

• ④ data-derived 58% — the finance research already decomposes the gap into UNVTD/GRID mechanisms: digital ID 27% + verifiable documents 9% + asset collateral (warehouse receipt / eBL) 22%. So ④ uses 58% in place of the default 45%. The headline is finance enabled; the economic benefit is that volume × (cost-of-capital saving + trade margin).

• ⑤ derived — national productivity is the macro roll-up of ①–④, not an independent pool; the administrative-burden research is used only as a top-down cross-check.

• ⑦ algorithmic assurance — the shift from sampled manual audits to algorithmic auditing of every verifiable transaction. ⑦’s own value is the assurance cost saved (audit / due-diligence / KYC labour) plus risk reduction; the revenue and finance that better assurance enables are counted under ②③④ (mechanism narrated here, not re-counted). Two haircuts: the automatable fraction of assurance labour, and the algorithmically-clearable share — the proportion of transactions/applications where verifiable evidence is sufficient to auto-assess (excludes sanctioned/blacklisted and novel-risk cases). The clearable share is a defended assumption (no authoritative figure; anchored on the 95–98% AML false-positive rate and trusted-trader coverage), used with a sensitivity range. Tax ceiling (option A): the move from self-assessed to evidence-based assessment is treated as a qualitative upside to ③, not yet a number — the evidence (e.g. Mexico CFDI → +14% declared revenue over three years; Peru → +8% reported VAT liabilities, chiefly via deterrence; Italy SdI → €2–4bn/yr) is VAT-strong but CIT-thin.

• Coverage & deterrence — why verifiable documents realise more of ②③ (and fund the incentive pool). Sampled audit reaches under 1% of transactions with heavy selection bias; verifiable documents flip coverage toward ~100% at ≈ zero marginal cost and add a deterrence / voluntary-compliance effect audit cannot (the probability of any one transaction being audited is ~zero). Realised outcomes: Mexico CFDI +14% declared revenue (3yr); Peru +8% VAT liabilities (yr 1, deterrence-led); Italy SdI €2–4bn/yr. Even high-compliance economies retain a recoverable pool (Australia: ~32% random error-line rate, ~$0.25bn/yr voluntary disclosure). Guardrails: size the pool from random non-compliance rates, never selection-biased targeted hit rates; an error-line rate is not a revenue rate (dollar-weighted leakage is far smaller); separate the compliance gap (addressable) from the policy gap (not). This larger realised ②③ is what funds the deferred-duty / expedited-clearance incentives customs and tax share with adopters — making the adoption ramp partly endogenous (richer incentives → faster uptake → more verifiable transactions → more recovery). Detail: research/compliance-assurance/coverage-recoverable-pool-findings.md.

Organisation-level benefits (bottom-up)

The levers above are country-pool / top-down — correct for national personas (country, customs, tax, registry), fed by the per-country Evidence tables. Organisation personas (trader, logistics, bank, cargo-terminal operator) are modelled bottom-up from the firm’s own activity, using the Organisation Benchmarks page for the numbers.

• Driver — dual mode: the firm’s activity counts (documents, invoices, shipments, applications) come bottom-up from the firm where available, else top-down from its size band × sector ratios. Confidence: bottom-up → High; one ratio hop → Medium; stacked hops → Low.
• Benefit categories per persona fall in four groups — Cost · Revenue / growth · Risk · Working-capital — each computed as driver × unit value × UNVTD/GRID share × adoption (mapping to levers ⑥①④⑦).
• Benefit mix is segment-dependent — two axes, not one size band. Trade volume (size band) drives the cost-of-trade, clearance & document benefits (①/⑥/⑦-enabled), which scale with shipment/document count — the large-firm lever (clearance value includes inventory-in-transit working capital, which dwarfs demurrage for high-value cargo). Finance-gap exposure (credit standing × country banking depth × trade profile — not size alone) drives the trade-finance benefit (④): a level-shift for constrained firms, near-zero for well-banked ones. Weight the levers per the Organisation Benchmarks segmentation table — never a flat mix.
• Re-keying calibration: the per-document saving is the verifiability premium (authenticity, no fraud/disputes/exceptions), not full manual re-keying — modern AI/OCR accounting already automates the keystrokes. Default to the lower end.
• Incentives received (deferred duty, pre-arrival clearance, cheaper finance) are often the largest, adoption-driving benefit. They are value transfers — the adopter counts them as received; the gaining counterparty (customs ②, bank ⑦) counts its gain net of the incentive offered; the system counts the value once, at creation (do not double-count). Incentive strength drives the adoption ramp.

Full design and the remaining research gaps: research/methodology/org-level-benefits-scoping.md.

How costs are calculated — the decentralised model

Cost has the opposite shape to benefit: a small, finite set of one-time uplifts, amortised across the whole economy, with marginal cost per business / document approaching zero.

Rejected benchmarks: per-entity identity intermediaries (e.g. LEI) and network hubs with transaction fees (Peppol 4/5-corner) — both price a centralised architecture UNVTD does not use, and so overstate cost.

From benefit and cost to a result

Each business case nets the benefit ramp − cost ramp over the same adoption curve → payback period, NPV, and simple ROI. Payback is the headline.

Defaults so NPV and payback are computed, not asserted:

• Horizon: 10 years.
• Adoption ramp: benefits scale linearly from zero to the chosen mature share (Conservative 30% / Central 50% / Ambitious 70%) by year 5, then hold flat to year 10. The one-time cost falls in year 0–1; recurring opex applies each year.
• Discount rate: 5% for national / public-sector cases, 8% for organisation cases (state which; vary it in sensitivity).
• Payback = the first year the cumulative discounted net benefit turns positive. NPV = discounted (benefit − cost) summed over the 10-year horizon.

Counting rules

• Value is counted once, at the lever. Scenarios narrate mechanisms (primary or secondary); secondary contributions are counted only under the owning lever.
• ⑤ is a roll-up, never added on top of ①–④.
• Transfer pricing — a Border-compliance mechanism whose value lands in ③ domestic revenue; counted once, under ③.
• ⑦ counts assurance, not the revenue it enables — ⑦ books the cost of assurance saved plus risk reduced; the extra revenue (②③) and finance (④) that algorithmic auditing unlocks are counted once, under those levers.
• Wrong-entity check — a lever’s value belongs to the entity that owns it. If the subject owns no lever (e.g. a navigation/safety port authority that handles no cargo and issues no documents), the honest output is to say so and re-target the case to the entities that do own the levers (cargo-terminal operators, logistics providers, customs) — never fabricate figures for an entity that cannot capture them.

Conservatism

• Default to the lower end of every range unless there is strong, specific evidence for a higher figure. Standing exception: the ②③ leakage / gap pool rates — use the central estimate, because their random-sample basis already understates the pool (see the ②③ lever note).
• Present figures as ranges, assign a confidence level (High / Medium / Low), and give conservative / base / optimistic scenarios.

Authoritative sources

• Cost of trade — ESCAP–World Bank Trade Cost Database; UNESCAP paperless-trade studies.
• Revenue — border — Global Financial Integrity (trade misinvoicing); World Customs Organization.
• Revenue — domestic — OECD / European Commission VAT Gap reports; IMF RA-GAP; CIAT.
• Trade finance — ADB Trade Finance Gaps; IFC–World Bank MSME Finance Gap; World Bank secured-transactions work.
• National productivity — OECD administrative-burden studies; national e-invoicing programme results.
• Operational efficiency — FIT Alliance eBL survey; McKinsey trade-documentation estimates.
• Compliance & assurance — OECD Tax Administration / ISORA (cost of collection, audit staffing); WCO post-clearance audit; Wolfsberg/ICC/BAFT trade-finance due diligence; IMF (e-invoicing compliance effects).
• Identity — GLEIF (as a counter-example to the decentralised register-anchored model).

Detailed per-country tables and the derivation scripts behind these figures are in the /research folder of the project repository.