Trust

How We Test DeFi Protocols

Last updated: August 2, 2026

The testing pipeline

  1. 1

    Fund a live wallet

    An analyst-controlled wallet is funded with real capital sized to the protocol's realistic retail use case — enough to trigger real fees, funding, and slippage rather than dust-level rounding.

  2. 2

    Execute representative flows

    Deposits, trades (maker and taker where applicable), bridging, and withdrawals are executed on mainnet. Every transaction hash is retained internally as evidence for the figures we publish.

  3. 3

    Reconcile fees against documentation

    On-chain receipts are reconciled against the protocol's published fee schedule. Discrepancies — hidden markups, borrow drag, funding asymmetries — are what our reviews are built around.

  4. 4

    Assess security posture

    Audit reports, exploit history, admin-key configuration, oracle design, and upgrade mechanisms are reviewed and scored. This pillar carries the heaviest weight and caps the overall rating.

  5. 5

    Publish, then re-verify

    Fee snapshots and canonical links are re-verified monthly; the review's last-updated date changes only when the content itself is re-assessed.

Rating weights

Every /10 rating on Protocol Signal is a weighted composite of five pillars:

  • Security & Risk

    30%

    Audit coverage and auditor quality, historical exploits and how the team responded, admin-key and upgradeability posture, oracle design, and bridge/custody dependencies. A single unresolved critical finding caps the overall score.

  • Fees & Execution

    25%

    Published maker/taker or swap fee schedules verified by executing real trades and reconciling the on-chain receipts against the documented rate. Hidden costs — funding, borrow drag, in-wallet markups, slippage on realistic sizes — are measured, not assumed.

  • Liquidity & Reliability

    20%

    Order-book depth or pool TVL on the pairs traders actually use, uptime during volatility events, oracle behavior during spikes, and withdrawal reliability.

  • Transparency & Team

    15%

    Public documentation quality, on-chain verifiability of stated metrics, disclosure of tokenomics and insider allocations, and responsiveness to disclosed issues.

  • User Experience

    10%

    Onboarding friction, wallet support, mobile behavior, and how clearly risks are communicated inside the product itself.

Independence

Ratings are finalized before monetization is considered. Referral links are disclosed and marked rel="sponsored"; no protocol can pay for a score, a ranking position, or pre-publication review. Full details are in our editorial policy and affiliate disclosure.

Frequently Asked Questions

Do you actually use the protocols you review?

Yes. Before a rating is published, an analyst funds a wallet, deposits into the protocol, executes representative trades or transactions, and withdraws. Fee figures and execution notes in our reviews come from those on-chain receipts, cross-checked against the protocol's published schedule.

How is the /10 rating calculated?

Each protocol is scored against five weighted pillars — Security & Risk (30%), Fees & Execution (25%), Liquidity & Reliability (20%), Transparency & Team (15%), and User Experience (10%). Scores are set by the reviewing analyst, challenged in an editorial pass, and only then published. An unresolved critical security finding caps the overall score regardless of other pillars.

How often are reviews updated?

Fee schedules and canonical app links are re-verified on a monthly snapshot cycle; full reviews are re-assessed when a protocol ships a major version, changes its fee schedule, or experiences a security incident. Every page shows its last-updated date.

Do affiliate partnerships affect scores?

No. Ratings are finalized before any monetization decision, referral links are marked rel="sponsored", and several of our top-rated protocols have no affiliate relationship with us at all. The separation is documented in our editorial policy.

What happens when you can't verify a number?

We publish N/A instead of an estimate. Live market metrics (TVL, volume) that we cannot verify at review time are never fabricated to make a table look complete.