AI CATEGORIZATION
THAT EXPLAINS ITSELF
Most crypto tax platforms use AI but never tell you why they categorized a transaction. We do the opposite. Every categorization comes with a confidence score and plain-English explanation. Green badge means we're 99%+ confident. Orange means review recommended. Red means we need your input. You stay in control.
WHY CRYPTO TAX
CATEGORIZATION IS BROKEN
Traditional crypto tax platforms import transaction history from exchanges and wallets, then guess at categorization based on limited data. The result? Up to 30% of DeFi transactions are miscategorized or missed entirely.
Here's what goes wrong:
LIQUIDITY POOL TRANSACTIONS
Platforms see 'Token Out' and 'Token In' but miss the LP token mint and fee accrual events. Result: Incorrect cost basis and missing income.
MULTI-HOP SWAPS
Aggregators like 1inch execute trades across multiple DEXes in one transaction. Platforms categorize it as a single swap. Result: Inaccurate purchase price for cost basis.
STAKING REWARDS
Protocols distribute rewards through various methods—direct transfers, rebasing tokens, or wrapped derivatives. Platforms inconsistently track these. Result: Underreported income or double-counting.
NFT ROYALTIES
Secondary sales trigger royalty payments via smart contract events, not visible in wallet history. Platforms miss them. Result: Unreported income.
The common thread? These platforms only see what's in your wallet history. They don't analyze the underlying smart contract interactions where the accounting truth actually lives.
HOW WE SOLVE IT:
EVENT LOG ANALYSIS
CryptoForms doesn't just import your transaction history. We run full Ethereum nodes and analyze Event Logs—the data emitted by smart contracts during execution. This is where the accounting truth lives.
TRANSACTION DISCOVERY
We sync with 500+ exchanges, wallets, and DeFi protocols to build a complete transaction history. Unlike other platforms, we also query blockchain nodes directly to catch on-chain activity your wallet provider might not show.
EVENT LOG PARSING
For every on-chain transaction, we extract and decode Event Logs emitted by smart contracts. These logs contain the accounting truth:
- • Swap amounts and fees
- • LP token mints and burns
- • Staking deposits and withdrawals
- • Reward distributions
- • NFT transfers and royalty payments
AI CATEGORIZATION
Our AI models—trained on 10M+ crypto transactions—analyze the event logs and assign categories:
Each categorization includes a confidence score (0-100%) and reasoning.
VERIFICATION
Every categorization is cross-checked against blockchain state: token balances before/after, USD values at block timestamp, gas fees deducted, and protocol-specific logic. If anything doesn't add up, we flag it for manual review.
THE CONFIDENCE
SCORING SYSTEM
We don't just categorize transactions—we tell you how confident we are in each decision. This transparency lets you focus manual review time where it matters most.
90-100%
Confidence
AUTO-APPROVED
High-confidence categorizations based on clear event log patterns and verified blockchain state. Typically 85-90% of all transactions.
Examples:
- • Simple swaps on Uniswap, Curve, Balancer
- • Standard ERC-20 transfers
- • Common staking deposits/withdrawals
- • NFT mints with clear pricing
70-89%
Confidence
REVIEW RECOMMENDED
Medium-confidence categorizations where our AI found multiple possible interpretations. Quick manual review recommended. Typically 8-12% of transactions.
Examples:
- • Complex multi-hop swaps
- • First-time protocol interactions
- • Unusual token movements
- • Partial LP withdrawals
<70%
Confidence
MANUAL REVIEW
Low-confidence categorizations that require your input. We show you what we found but need you to make the final call. Typically 2-5% of transactions.
Examples:
- • Unknown contract interactions
- • Failed transactions with unclear intent
- • New DeFi protocols (pre-training)
- • Custom smart contract calls
Real-World Accuracy Statistics
99.7%
Overall accuracy rate
40+
Hours saved per tax season
87%
Average green confidence rate