How to Read the 8-Factor Score: What a High Number Actually Means

The one-sentence take

Pledge's 8-factor score is a reproducible, evidence-bound composite — a weighted sum of eight 0–10 factor scores, computed the same way every time, with every load-bearing claim traced to a primary source or flagged as unverified. A high score means the product is structurally boring in the ways that matter. It does not mean "safe" — and it is not a buy rating. Read it as a map of where a product can hurt you, not as a verdict.

What the score actually is

Every scored product on Pledge carries a composite between 0 and 10. The composite is not a vibe and not a black box — it is arithmetic:

• Eight factors, each scored 0–10.
• Each factor carries a weight (the factors sum to 100, or to 1.00, depending on the rubric — the math is the same).
• The composite is the weighted sum, divided by 100, rounded to two decimals — computePledgeSafetyScore in the repo, verified by an automated data-integrity test. Every published score must be reproducible by running the product's factor scores through that one function. No machine learning, no hidden judgment call buried in a spreadsheet.
• Each factor carries structured evidence fields plus fallback evidence — the actual sources (lender docs, SEC filings, issuer pages, on-chain data) that support the score. A score with no attachable evidence is not a score; it's a claim.

That reproducibility is the entire point. Anyone can copy a rate sheet. Nobody can copy a scoring system where every number is checkable — and where the weakest factors are admitted in public.

The reference rubric: Bitcoin loans (weights sum to 100)

The flagship rubric — the one the whole site opens into — scores bitcoin-backed lenders on eight factors. Read the weights as the risk hierarchy:

(Weights verified from the Pledge repo src/lib/methodology.ts — METHODOLOGY_FACTORS, sum = 100.)

The 20-weight rehypothecation factor is the hierarchy in one number: the single most consequential thing a lender can do to you is use your collateral. Everything else is measured against that.

The other rubrics — same machinery, different risk frontiers

The 8-factor model is not one rubric; it's one method applied per category, each weighting toward its own risk frontier. This is the most important thing to understand about the scores, and the easiest to get wrong:

• Loans — the rubric above (rehypothecation 20 / custody 15 / regulatory 15 / reserves 12 / track record 10 / liquidation 10 / loss protection 8 / transparency 10).
• Custody providers — custody model 0.18 / key management 0.16 / insurance 0.14 / regulatory 0.12 / recovery 0.12 / transparency 0.10 / track record 0.10 / loss history 0.08 (weights sum to 1.00).
• ETFs — fee, liquidity, tracking, premium/discount, scale, issuer, transparency, custody (weights 25/20/12/10/13/8/7/5).
• Digital Credit / stablecoins — backing source 25 / mint-redemption path 15 / peg design 15 / yield source 10 / liquidity-exit depth 10 / smart-contract-bridge 10 / governance-counterparty 10 / transparency-proof 5.
• Treasuries — seven factors (NAV premium/discount 30 / BTC-per-share 15 / dilution 15 / governance 15 / liquidity 10 / fee 10 / track record 5).
• Wallets — deliberately not scored. Five review lenses (recovery safety, key control, ease of use, hardware support, privacy tooling) plus an evidence-quality state. "Fit over rank": a single number would hide the tradeoffs the reviews exist to surface.

Because each category weights toward its own risk frontier, a 7.5 loan and a 7.5 treasury are not the same 7.5. The method is comparable; the scores are not directly comparable across categories. A loan 8.0 measures borrower protection; an ETF 8.0 measures fee + liquidity + tracking. Same number, different property. Read every score with its category's one-line risk frontier.

The perfect-10 anchor: "nearly boring"

A 10/10 does not mean "zero risk." It means the best genuinely achievable real-world product in its category today — a mature, red-flag-free configuration a real best-in-class product can actually hit. The repo's own standard for loans:

"A 10/10 Bitcoin loan would be nearly boring." — the borrower keeps meaningful control of collateral; collateral cannot be reused; reserves or collateral handling are independently verifiable; legal status is clear; liquidation rules are conservative and explicit; the lender has survived multiple market cycles with clean public evidence.

The anchor rule does two jobs. It keeps headroom on the scale — a 10 is rare but landable, not an unreachable ideal. And it makes the honest claim explicit: most real products should not reach 10, because every borrowing path adds a tradeoff (custodial exposure, smart-contract risk, jurisdiction limits, quote-dependence, liquidation mechanics, thinner public proof). A 7–8 is strong, not weak. The top real products today sit at 7–9 — and the reasons they don't hit 10 are the reasons you should read the factor breakdown, not the headline.

What a high score means — and what it does NOT mean

What it means: the product is boring in the ways the rubric weights. For a loan: the collateral can't be reused, the custody structure can't be unilaterally moved, the liquidation rules are conservative and borrower-visible, the reserves are independently checkable, and the lender has a multi-cycle record. That is a genuinely useful signal — it's the difference between a rate sheet and a risk profile.

What it does not mean — the honest limits (the part nobody else publishes):

1. Scores are category-relative. A 7.5 loan ≠ a 7.5 treasury ≠ a 7.5 ETF. Never compare raw scores across categories.
2. Scores are lender-level, not product-level. Two lenders with equal safety scores can carry very different products — different APR, LTV, liquidation terms. The number tracks the lender; the agreement tracks the loan. Read the product, not just the score.
3. Scores are point-in-time snapshots. Rates, LTV tables, and live markets move. Every score on the site carries a data-layer date — treat anything older than a few weeks as needing a re-check at the moment you act on it.
4. The weakest-structured factors are the most safety-critical. Loss Protection has no structured evidence in the repo (prose-backed only); Transparency leans on proof-of-reserves and social links. The methodology says so — because the factors that can hurt you most are the ones with the thinnest public proof. A high score does not erase that gap.
5. Confidence is flagged, not assumed. On the custody desk, only two of eleven scores are verified; the rest are estimated (entity verified, some sub-factors inferred from public disclosure). An estimated score is a starting point, not a verdict.
6. A high score is not a buy rating. It is not a recommendation, not financial advice, and not a guarantee. It is a risk map. The map is not the territory — the agreement and the charter are.
7. The model has known inconsistencies, and it says so. SATA is double-listed on two methodologies (Digital Credit 8-factor and treasuries 7-factor) with slightly different scores — the one internal inconsistency a defensible rubric must resolve. It is flagged, not hidden, because the whole point of the transparency is that you can see the seams.

How to read it in practice

1. Read the category, not the number. What risk frontier is this score measuring? (Loans = borrower protection; custody = who can move the funds; ETF = fee + liquidity + tracking.)
2. Read the factor breakdown, not the composite. A 7.9 with a 6 on rehypothecation is a different product from a 7.9 with a 6 on track record. The composite hides the shape; the factors show it.
3. Check the confidence flag. Verified vs estimated changes how much weight the number should carry.
4. Check the evidence. Every factor has sources attached. If a claim that matters to you has no attachable source, treat it as unverified — even on a high-scoring product.
5. Read the actual agreement. The score is the map; the current terms are the territory. What's on the rate sheet today can change in the fine print tomorrow.
6. Timestamp everything. Any number you act on should carry its as-of date. The site's data layer is a snapshot; your decision is not.

The take

The 8-factor score is Pledge's answer to a market that sells rate sheets and hides custody structures: a number you can check, with the seams left visible. It is deliberately boring — reproducible arithmetic, attachable evidence, admitted weaknesses, flagged inconsistencies. A high score means the product is boring in the ways that matter. It does not mean safe, and it is not a buy rating. Read the factors, check the evidence, read the agreement — and treat the score as the map, never the territory. Risk is the product.