An analysis of Bitcoin OP_RETURN metadata

1 Introduction

• Protocols use Bitcoin’s immutable, open blockchain to support applications such as document-existence certification and ownership tracking beyond currency transfers.Many save metadata using OP_RETURN, an instruction in Bitcoin’s scripting language.Source: §1 · p. 1
• The study asks how OP_RETURN metadata affects blockchain size and which blockchain-based applications use the instruction, and how.Source: §1 · pp. 1–2
• 1,887,708 OP_RETURN transactions were analyzed; 22 protocols were identified, associated with 51% of transactions, while 15% were empty.The analysis also tracks usage over time, application domains, and metadata space consumption.Source: §1 · p. 2

2 Background on Bitcoin

• Bitcoin transactions are recorded in a public, append-only blockchain, and their general form allows multiple inputs and outputs with scripts and values.Source: §2 · pp. 2–3
• A transaction spends redeemable outputs by supplying inputs that satisfy their scripts, and valid transactions must use UTXOs and preserve the input-output value constraint.The UTXO set is the set of redeemable outputs in the blockchain.Source: §2 · pp. 2–3
• OP_RETURN outputs always evaluate to false, making them provably unspendable and removable from the UTXO set.This avoids the UTXO-consumption issue associated with unspendable outputs that are difficult to distinguish from spendable ones.Source: §2 · p. 4
• OP_RETURN became standard in March 2014; later releases raised its data limit from 40 bytes to 80 and then 83 bytes.The initial 80-byte plan was reduced to 40 bytes in release 0.9.0; releases 0.11.0 and 0.12.0 extended the limit.Source: §2 · p. 4

3 Methodology for classifying OP RETURN transactions

• The corpus contains 1,887,708 OP_RETURN transactions collected through block 453,200, added on 2017/02/15.The stored fields include transaction and block hashes, block timestamp, and attached metadata.Source: §3 · p. 4
• Researchers identify 22 protocols associated with 33 identifiers, plus three protocols without identifiers, by searching the web for protocol-identifier bindings.They manually inspect search results; a protocol may have multiple identifiers or none.Source: §3 · p. 5
• The tool associates 970,374 transactions, about 51% of the total, with protocols; 15% are empty and 32% are unknown.Empty transactions have no data attached to OP_RETURN, while unknown transactions have no known identifier.Source: §3 · p. 5
• Frequency analysis of the first 1–12 bytes of unknown transactions finds roughly uniform strings and no relevant anomalies, yielding no new identifiers.The search uses D = 12, δ = 2, and N = 100.Source: §3 · p. 5

4 Qualitative analysis of OP RETURN transactions

• Protocols are grouped by intended application domain after researchers manually inspect each protocol’s website.Source: §4 · p. 5
• Assets cover ownership, exchange, or value of real-world assets; document notary certifies document ownership or timestamp, while digital arts concerns access and copyright.Other covers different goals; Empty has no OP_RETURN data, and Unknown lacks a detected identifier.Source: §4 · p. 6
• Because OP_RETURN space is limited and larger metadata requires more transactions and fees, asset protocols often use space-efficient representations and off-chain solutions.Document notary is distinguished from digital arts because it can work without the application remaining live, whereas digital-arts applications commonly require document gathering and user interaction.Source: §4 · p. 6

5 Quantitative analysis of OP RETURN transactions

• Table 1 reports protocol category, name and identifier, first-transaction date, transaction count, metadata bytes, and average metadata size.Dates are inferred from the timestamp of the containing block; the byte totals count metadata, not the OP_RETURN instruction or other transaction fields.Source: §5 · p. 6, Table 1 · p. 7

5.1 Overall statistics

• The scan detects 1,887,708 OP_RETURN transactions across 98,233 blocks through block 453,200; these constitute approximately 0.96% of all blockchain transactions over the measured seven years.Source: §5.1 · p. 7

5.3 Space consumption

• Metadata averages less than 40 bytes generally, despite extension to 80 bytes introduced on 2015/07/12; only Blockai metadata is close to 80 bytes. Document-notary protocols average 40 bytes, reflecting 16-byte identifiers and saved hashes.Source: §5.3 · p. 9
• As of 2017/02/15, OP RETURN transactions consumed ∼0.3% of total blockchain space; after subtracting ∼42 MB of headers across 453,200 extracted blocks, all transactions totaled ∼102 GB.Source: §5.3 · p. 9

5.2 Transaction peaks

• ∼96% of empty transactions fall within the peaks; July–August’s ∼100,000 spike seems spam-related, while September’s ∼300,000 peak coincides with a stress test.The weekly series covers March 2014 through February 2017.Source: §5.2 · pp. 7–8, Figure 2 · p. 8, §5.4 · p. 10

5.4 Distribution of protocols by category

• The unknown category accounts for ∼33% of OP_RETURN transactions, while assets lead transaction counts and document notary is the most numerous category.Empty and unknown transactions form a notable component of the category distribution.Source: §5.4 · p. 10, Figure 3 · p. 10

• Usage had risen steadily since March 2015, reaching an average of ∼25,000 new OP_RETURN transactions per week in February 2017.Source: §6 · pp. 10–11
• Verifying the conjecture about empty transactions required comparing their identifiers with those in, which were not available online.Source: §6 · p. 10
• OP_RETURN transactions accounted for ∼1% of blockchain transactions and used ∼0.3% of its space.Source: §6 · p. 11
• Sign-to-contract and pay-to-contract techniques can prove a value was known when a transaction was recorded without affecting transaction size.The authors identify comparing metadata-storage methods as future research.Source: §6 · p. 11
• The authors suggest that applications may choose OP_RETURN for the perceived security and persistence of Bitcoin, and that their needs could affect future protocol evolution if growth continues.Source: §6 · p. 11

• Figure 4 provides charts for assets.Source: Figure 4 · p. 14
• Figure 5 provides the first set of document notary charts.Source: Figure 5 · p. 15
• Figure 6 provides the second set of document notary charts.Source: Figure 6 · p. 16
• Figure 7 provides charts for digital arts.Source: Figure 7 · p. 17
• Figure 8 provides charts for other categories.Source: Figure 8 · p. 18