← Four-layer screen

Methodology · DFI Mandate-Fit & Four-Layer Screen v1.0

How the DFI screen is built

From a 94-asset roster to a defensible, per-institution shortlist — and where the honest limits are. Version v2.137.23

The screen answers the questions a development-finance officer asks, in the order they ask them: is this in our mandate, are the economics real, does it survive downstream risk, and is it financeable now? It composes existing feeds — it originates no figures of its own. Every cell propagates its source feed's three-state badge (Sourced / Derived / Pending / Absent) and carries a machine-readable data-source deep-link to its origin.

The four layers

For a selected institution the roster is passed through four gates, each a strict subset of the one above (a monotonic funnel):

  1. Mandate — geography gate from the mandate-fit layer (below).
  2. Bankability-complete — assets with Sourced/Derived published economics, using predicates identical to the cross-asset bankability screener and the flagship report (after-tax / pre-tax / PEA).
  3. Deep risk-assessed — assets carrying a downstream verdict (operating-flagship coverage) and/or a per-asset E&S safeguards screen (development-pipeline coverage) — two assessment layers over different universes, counted together (community / environmental / jurisdictional-security context).
  4. Financeable shortlist — the diligence-grade financeability set (sponsor, capital need, funding status, offtake), ranked by DFI capital need then NPV.

An asset that fails a gate drops at that gate and is not carried forward. Where an institution's bankable set exceeds the deep-assessed set, the funnel narrows sharply between layers 2 and 3 — that narrowing is shown, not hidden.

Mandate-fit method

Each institution's criteria are curated from its published mandate and applied to the roster across four dimensions:

The shortlist

The shortlist for an institution is the intersection of all four layers. It is ranked by DFI capital need (High→Low) then by common-basis NPV — the ranking basis is stated on the surface so the order is auditable. Each shortlisted NPV carries its study vintage and a Derived age chip (≤3y / 4–6y / >6y); where the citation carries no year, the cell shows the citation or Absent rather than an invented date. The shortlist exports to a clean print/PDF committee brief that retains provenance and caveats.

Stated limits

Cross-references: mandate-fit screen · four-layer screen · dfi-mandate.json feed · bankability-depth method · downstream method