BIS Working Paper: Towards Verifiability of Total Value Locked DeFi — Complete Analysis

Why Total Value Locked DeFi Is the Most Important—and Least Understood—Metric

Total Value Locked has become the defining metric of decentralized finance. When investors evaluate DeFi protocols, when media outlets rank blockchain ecosystems, and when regulators assess the scale of decentralized financial activity, TVL is the number everyone cites. As of early 2025, aggregate DeFi TVL across all chains exceeds $100 billion—a figure that influences capital allocation decisions, regulatory postures, and public perception of the entire crypto industry.

Yet despite its ubiquity, total value locked DeFi as a metric has never been rigorously standardized, independently audited, or formally defined. The Bank for International Settlements (BIS) has now published the most comprehensive study to date examining how TVL is actually computed, how verifiable those computations are, and what structural problems exist in the ecosystem’s most-watched number. The findings are both illuminating and concerning.

For anyone who has ever cited a TVL figure in an investment thesis, a regulatory filing, or a market analysis, this BIS working paper is essential reading. It reveals that the metric most people treat as gospel is, in many cases, unverifiable, inconsistently computed, and potentially inflated by double counting. Understanding these dynamics is crucial for participants in the current state of DeFi and for the maturation of decentralized finance as a legitimate asset class.

The BIS Research Methodology: 939 Protocols Under the Microscope

The BIS study’s strength lies in its empirical rigor. Rather than theorizing about how TVL might be computed, the researchers instrumented the actual computation infrastructure—specifically, the DeFiLlama adapters repository—and recorded every HTTP and blockchain node call produced during TVL calculation.

The dataset is substantial: 939 DeFi projects deployed on Ethereum (from DeFiLlama’s total of 3,494 tracked protocols across all chains), analyzed across five repository commits spanning January 2023 to January 2024. The researchers recorded 197,775 proxied calls for the primary commit alone, capturing the complete data pipeline from raw blockchain queries to published TVL figures.

This methodology is significant because it examines TVL computation at the infrastructure level—not relying on API outputs or published numbers, but intercepting the actual queries that produce those numbers. By doing so, the researchers could identify precisely where data comes from (on-chain vs. off-chain), which functions are called (standard vs. custom), and where the same data is queried by multiple protocols (potential double counting). The token universe observed encompassed 12,246 distinct token addresses, of which 942 could be independently priced using Uniswap V2 liquidity pools.

How Total Value Locked DeFi Is Actually Computed: The On-Chain vs. Off-Chain Reality

One of the study’s most important findings challenges a fundamental assumption about DeFi transparency. While blockchain proponents emphasize that all data is on-chain and publicly verifiable, the reality of TVL computation is more nuanced.

The backbone of TVL calculation relies on two standard on-chain functions: balanceOf (for ERC-20 token balances) and eth_getBalance (for ETH balances). These are the queries that check how many tokens are held by a protocol’s smart contracts—the fundamental building blocks of any TVL computation. Additional standard calls include totalSupply, getReserves, token0/token1, symbol, totalAssets, and underlying.

However, 10.5% of protocols depend partially or entirely on external servers for their TVL computation. The study identified 64 distinct external servers queried during the process, with TheGraph being the most frequently accessed. This means that for over one in ten protocols, the published TVL figure cannot be reproduced using on-chain data alone—you need access to the same external data sources, which may not be publicly accessible, permanently available, or independently verifiable.

This finding strikes at the heart of DeFi’s value proposition. If the ecosystem’s most important metric requires off-chain data sources for its computation, then the transparency and verifiability claims that distinguish DeFi from traditional finance are partially undermined. Understanding blockchain fundamentals helps contextualize why this matters: the entire trust model depends on on-chain verifiability.

The Double Counting Problem: $16 Billion in Overlapping Total Value Locked DeFi

Perhaps the most striking finding in the BIS study is the quantification of potential double counting in DeFi TVL. The researchers identified 240 balance queries (230 balanceOf + 10 eth_getBalance) that were executed on the same smart contract addresses but associated with multiple different protocols.

When the same token balance is counted in the TVL of two or more protocols, the aggregate TVL figure is inflated. The study found that the combined value tied to these overlapping queries peaked at nearly $16 billion at the end of 2021—a period when DeFi TVL was at its all-time high and billions of investment dollars were flowing into the ecosystem based partly on TVL metrics.

To illustrate: if Protocol A holds 1,000 ETH in a smart contract, and Protocol B wraps or references the same 1,000 ETH through a composability layer, both protocols may report that ETH in their TVL. The aggregate TVL dashboard then shows 2,000 ETH locked, when only 1,000 ETH of real value exists. This is a structural consequence of DeFi’s composability—protocols building on top of other protocols—but it creates a measurement problem that the industry has not adequately addressed.

For investors and regulators using aggregate TVL as a gauge of DeFi’s scale and systemic importance, the double counting problem is material. It means that headline TVL figures may overstate the actual economic activity in DeFi by a non-trivial amount, with implications for risk assessment, regulatory responses, and investment sizing decisions.

Verifiable TVL (vTVL): A New Standard for DeFi Transparency

The BIS researchers don’t just identify problems—they propose a solution. The paper introduces the concept of verifiable TVL (vTVL): the portion of TVL that can be independently recomputed using only on-chain data and interpretable standard balance queries (balanceOf and eth_getBalance).

To measure how published TVL compares to what can be independently verified, they define a Discrepancy Ratio: the average ratio of (vTVL / published TVL) − 1. A Discrepancy Ratio of zero means perfect alignment; −1 means the verifiable TVL is zero (nothing could be independently confirmed).

Applied to a detailed case study of 400 protocols, the results are revealing:

• 94 protocols (23.5%) had negligible discrepancy (|Discrepancy Ratio| ≤ 0.05)—essentially perfect match between published and verifiable TVL
• 186 protocols (46.5%) had Discrepancy Ratio within ±0.5—broadly aligned but with meaningful differences
• The remaining protocols showed larger discrepancies, particularly smaller protocols with proportionally larger relative differences

These findings mean that for roughly three-quarters of DeFi protocols, the published TVL cannot be exactly reproduced from on-chain standard queries alone. The discrepancies arise from off-chain inputs used by aggregators, alternative balance functions that are difficult to interpret, missing price data for some tokens, and differences in how protocol versions or contracts are grouped.

Non-Standard Balance Functions and Protocol Heterogeneity

The study identified 68 non-standard, project-specific functions used for balance or TVL computation, employed by approximately 14.2% of protocols. These custom functions—like Lido’s getTotalPooledEther—represent protocol-specific implementations that may not be interpretable without deep knowledge of each project’s architecture.

There is encouraging news on this front: the use of non-standard balance queries has been declining over time. The ratio of alternative to standard balance queries dropped from 28.2% in January 2023 to 8.9% in January 2024—a significant improvement suggesting organic standardization within the ecosystem. As new protocols launch with standard ERC-20 patterns and as older protocols update their architectures, the share of TVL that can be computed using standard methods is growing.

However, the heterogeneity that remains is consequential. When 14% of protocols use custom functions for balance computation, any third party attempting to verify TVL must understand not just generic blockchain queries but the specific implementation details of dozens of individual protocols. This creates a knowledge barrier that effectively limits independent verification to a small number of specialized teams—hardly the open, permissionless transparency that DeFi promises.

Token Classification and Pricing Challenges for Total Value Locked DeFi

Computing TVL requires two things: knowing how many tokens are held (balance queries) and knowing what those tokens are worth (pricing). The BIS study reveals significant challenges on the pricing side. Of the 12,246 distinct token addresses observed in DeFiLlama balance queries, only 942 tokens (7.7%) could be successfully priced using Uniswap V2 liquidity pools.

The researchers classified tokens into seven categories: ETH/wETH, wBTC, non-crypto-backed stablecoins, crypto-backed stablecoins, governance tokens, derivative tokens, and others. This classification matters because different token types carry different risks and different relationships to “real” economic value. A dollar in USDC held by a lending protocol represents something fundamentally different from a dollar in a governance token or a wrapped derivative.

The derivative token category is particularly problematic. Derivative tokens—such as liquid staking tokens, LP tokens, and wrapped versions of underlying assets—represent claims on other assets rather than independent value. Including derivative tokens in TVL calculations at face value can lead to inflation of the metric, as the same underlying value is counted once in the base asset and again in its derivative. The evolution of DeFi has created increasingly complex token relationships that make accurate TVL measurement progressively more challenging.

Five Recommendations for Improving Total Value Locked DeFi Verifiability

The BIS paper concludes with five concrete, practical recommendations for improving TVL standardization and verifiability across the DeFi ecosystem:

1. Prioritize On-Chain Data Sources

TVL should be computed from on-chain sources whenever possible, avoiding opaque off-chain feeds. When off-chain data is necessary, the sources should be documented and independently accessible.

2. Publish Protocol Contract and Token Lists

Protocols should publish explicit, up-to-date lists of all contract addresses and token identifiers used in their TVL computation. This allows third parties to reproduce and check TVL figures independently—a basic requirement for transparency that many protocols currently fail to meet.

3. Standardize Balance Methods

Protocols should prefer standard balance methods (balanceOf, eth_getBalance) over custom on-chain functions. Where custom functions are necessary, they should be clearly documented with their purpose, inputs, outputs, and relationship to underlying asset balances.

4. Publish Token Categorizations

Aggregators should publish their token categorizations—especially distinguishing derivative tokens from non-derivative tokens—and allow users to select which asset types to include or exclude from TVL calculations. This would enable users to compute TVL metrics that exclude derivative tokens, significantly reducing double counting risk.

5. Standardize Protocol Selection and Version Management

Clear standards for which protocol versions and contract deployments are included in TVL calculations, with explicit version management metadata. This ensures TVL is clearly attributable to specific protocol deployments and prevents confusion between different versions of the same protocol.

These recommendations align with broader transparency initiatives in blockchain finance, including Bitcoin’s foundational principles of verifiable, permissionless systems. The BIS paper argues that implementing these standards would not only improve TVL accuracy but strengthen the foundational promise of DeFi: that financial systems built on public blockchains should be inherently more transparent and verifiable than traditional alternatives.

Implications for DeFi Investors, Regulators, and Builders

The BIS working paper on total value locked DeFi has different implications for different stakeholders in the ecosystem.

For Investors

The study should prompt investors to treat TVL figures with more nuance. Rather than comparing raw TVL numbers across protocols, sophisticated investors should examine: whether TVL is computed using standard on-chain methods, what percentage of TVL comes from derivative vs. non-derivative tokens, whether the protocol’s TVL has been independently verified, and whether double counting with related protocols is possible. The growing emphasis on transparency in financial markets will increasingly demand this level of scrutiny.

For Regulators

Regulators worldwide are grappling with how to assess the systemic importance of DeFi. TVL is often cited in regulatory discussions as evidence of DeFi’s scale. The BIS study demonstrates that these figures are unreliable without standardization—providing a research foundation for regulators to demand better reporting standards from DeFi platforms and aggregators. The BIS Innovation Hub has been actively developing frameworks for digital asset supervision.

For Builders

Protocol developers and aggregator platforms should view the BIS recommendations as a roadmap for improving ecosystem credibility. Implementing standard balance queries, publishing contract lists, and documenting custom functions are relatively low-cost changes that significantly improve verifiability. As institutional capital considers DeFi exposure, the ability to provide independently verifiable metrics will become a competitive advantage. Projects referenced by regulatory bodies like the SEC will face particular scrutiny on metric accuracy.

The broader lesson is that DeFi’s promise of transparency is not automatic—it requires deliberate design, standardization, and commitment to independent verifiability. The BIS working paper provides both the diagnosis and the prescription. The question is whether the ecosystem will implement the changes needed to make total value locked DeFi a metric worthy of the trust placed in it.