METHODOLOGY

How a BDS-1000 score is made

The BDS-1000 is a consumer-awareness tool, not a legal liability framework. It asks one question: is a company's relationship with Israel's occupation material and significant enough that a consumer or investor would reasonably want to know about it? Every claim that feeds a score is cited to public evidence - 23,819 sources across 257 companies.

1 · Four evidence-only audits

Every company is investigated across four independent domains. The audits contain evidence only - no scores, no conclusions - and every factual claim carries an end-noted citation. Where we searched and found nothing, the audit says "No public evidence identified": gaps are shown, not padded.

MilitaryDefence contracting and procurement, dual-use and tactical products, supply to defence primes, base sustainment, munitions, export licences.
DigitalCloud and AI infrastructure serving state or military bodies, surveillance technology, data exposure and sovereignty, Israeli tech acquisitions and R&D.
EconomicOperations and investment in Israel, settlement sourcing and supply chains, financing the state, retail presence, revenue significance.
PoliticalPublic statements and lobbying, donations - including to military-welfare bodies, executive and board affiliations, suppression of employee or consumer dissent.

2 · Scoring - deterministic, then human-vetted

Each domain receives three component scores from the audit evidence: Impact (what the company does), Magnitude (how big it is - to the company or to Israel, whichever is higher), and Proximity (how direct the relationship is). They combine as:

V = I × min(M / 7, 1) × min(P / 7, 1)

The four domain scores then produce the composite BDS-1000 score. The dominant domain carries the score; the remaining three enter as their average, at a 0.2 weight:

BRS = ((VMAX + 0.2 × AvgOTHERS) ÷ 12) × 1000

Formula revised August 2026. The previous formula summed the non-dominant domains against a larger divisor, which had the effect of discounting a company for being clean in three domains while deeply complicit in one. The revision weights the single worst domain more heavily — a company's score is now primarily its most serious documented conduct, adjusted for the breadth of the rest. A fully complicit company still scores 1000, and a company with equal complicity across all four domains scores identically under both formulas; concentrated profiles score higher than before, and no company scores lower.

Scores are computed deterministically from vetted component values - the arithmetic is checked by validation scripts, not trusted to judgment. Every listed company has passed the evidence gates, independent examiner review required by its method, and final human vetting.

Published scoring methods

V4 79 companies
Human-reviewed Impact, Magnitude and Proximity values anchored to a vetted precedent library.
V5.1 5 companies
Published rubric band tables, framing enumeration, earned-reduction rules and red-team review.
V6.4 68 companies
Evidence-packet frame enumeration, multiple blind fixed-three banding draws, deterministic reconciliation and independent triage.
V6.5 105 companies
The current default: deterministic eligibility screening, strongest-two comparison when needed, one blind banding draw and independent triage.

Method labels describe how the evidence was converted into component values. The public formula and tier thresholds above are shared across the current index. Each company page shows its method beside the review date.

3 · Tiers

TierRangeReadingCompanies listed
A Extreme 800–1000 Core, systemic involvement in the machinery of occupation 18
B Severe 600–799 Severe, sustained material support in at least one domain 55
C High 400–599 Significant documented involvement 63
D Moderate 200–399 Moderate or indirect involvement 67
E Limited 0–199 Limited documented nexus 54

4 · Evidence standards

5 · Updates and corrections

Every dossier shows its last-updated date. The index is regenerated from the underlying research corpus, so corrections propagate to every affected page. If you believe a cited fact is wrong or out of date, contact us with the claim and the source - corrections are reviewed against the same evidence standards above.