Lenders deposit funds into Compound and receive interest payments from borrowers (supply-side fees).

Compound is a non-custodial lending protocol built on Ethereum. Compound was founded by Robert Leshner & Geoffrey Hayes and launched in 2017.

 COMP tokenholders own and govern the protocol. They currently take a cut of the total interest paid by borrowers (revenue). Compound’s investors include a16z, Paradigm, Polychain, etc.

governance

github

twitter

discord

etherscan

website

Lending

Compound

# Compound Active users (weekly) Methodology

## 1. Overview

Weekly Active Users (WAU) measures the number of unique users who interact with Compound's lending protocol within any 7-day rolling window. This metric answers the business question "How many distinct users actively used Compound's lending services in the past week?" and is calculated using probabilistic HyperLogLog data structures for efficient distinct counting across multi-dimensional breakdowns. The calculation aggregates user interactions from both Compound v2 and v3 protocols across Ethereum, Arbitrum, Base, and Polygon chains.

## 2. Data sources

The calculation uses the following underlying data sources:

- **Blockchain events**: Parsed smart contract events from Compound v2 and v3 protocols across multiple chains, including `Borrow`, `Mint`, `Redeem`, `RepayBorrow` events from CEther and CErc20 contracts on Ethereum, and `SupplyCollateral`, `Supply`, `WithdrawCollateral`, `Withdraw` events from Comet contracts on Ethereum, Arbitrum, Base, and Polygon
- **Transaction data**: Block timestamps from blockchain transactions to enable daily time aggregation
- **Protocol-specific contract data**: Smart contract addresses and event signatures for both Compound v2 (CEther, CErc20, CErc20Delegator) and v3 (Comet) implementations

## 3. Step-by-step calculation

### Component A: User interaction extraction
- **Purpose**: Identify all user interactions with Compound's lending protocol across supported chains and protocol versions
- **Logic**: Extract user addresses from blockchain events representing lending activities. For Compound v2 on Ethereum, this includes `Borrow`, `Mint`, `Redeem`, and `RepayBorrow` events from CEther, CErc20, and CErc20Delegator contracts. For Compound v3 across Ethereum, Arbitrum, Base, and Polygon, this includes `SupplyCollateral`, `Supply`, `WithdrawCollateral`, and `Withdraw` events from Comet contracts. Each event is classified by user type based on the specific blockchain event: borrowing events identify Borrowers, deposit events (Mint/SupplyCollateral) identify Depositors, withdrawal events (Redeem/WithdrawCollateral) identify Withdrawers, and repayment events (RepayBorrow/Supply) identify Repayers.
- **Key fields**: `block_timestamp` for time aggregation, `transaction_hash` for deduplication, `chain` for multi-chain analysis, `user_type` for behavioral segmentation, `user_address` for distinct user counting

### Component B: Daily active user aggregation  
- **Purpose**: Calculate daily active users by aggregating blockchain interactions to daily granularity
- **Logic**: Group user interactions by truncating `block_timestamp` to day boundaries and count distinct `user_address` values per day. Results are segmented by `chain`, `user_type`, and `product_id` dimensions to enable multi-dimensional WAU breakdowns. All interactions are classified under the 'lending' market sector and business segment. Chain availability filters are applied to ensure data quality and completeness.
- **Key fields**: `timestamp` (daily granularity), `timestamp_grain` (set to 'day'), `chain`, `user_type`, `product_id`, `user_dau` (distinct user address)

### Component C: HyperLogLog initialization for efficient counting
- **Purpose**: Convert daily active user counts into probabilistic data structures for efficient distinct counting in rolling windows
- **Logic**: Initialize HyperLogLog sketches with 15-bit precision for each daily user count, grouped by timestamp, chain, user type, and product dimensions. HyperLogLog enables approximate distinct counting with high accuracy and low memory usage, which is essential for calculating rolling window metrics efficiently.
- **Key fields**: `timestamp`, `timestamp_grain`, `chain_id`, `user_type`, `product_id`, `hll` (HyperLogLog sketch)

### Component D: Rolling window construction
- **Purpose**: Create 7-day rolling windows for WAU calculation using a datespine approach
- **Logic**: Generate a complete datespine covering all days from the earliest DAU record to yesterday to ensure no missing timestamps. For each date in the spine, create a 7-day lookback window by joining with the previous 6 days using timestamp arrays. This approach ensures consistent 7-day windows even when some days have no user activity. The datespine prevents gaps in the rolling window calculation that could undercount WAU.
- **Key fields**: `timestamp` (target date), `joinday` (dates within the 7-day window)

### Component E: Multi-dimensional WAU breakdowns
- **Purpose**: Calculate WAU across different analytical dimensions while maintaining distinct user counting accuracy
- **Logic**: Merge HyperLogLog sketches across 7-day windows for each dimension separately. Create four distinct breakdowns: project-level WAU (all chains and user types combined), chain-level WAU (distinct users per chain), user-type-level WAU (distinct users per behavioral category), and product-level WAU (distinct users per product). For each breakdown, unused dimension fields are set to null while maintaining the relevant grouping dimension. HyperLogLog merge operations ensure accurate distinct counting without double-counting users who appear in multiple days within the same window.
- **Key fields**: `timestamp`, `timestamp_grain`, `dimensions` (breakdown type), `chain_id`, `product_id`, `user_type`, `user_wau` (distinct weekly active users)

### Final aggregation
- **Combining components**: Union all dimensional breakdowns into a single result set with standardized schema, where each row represents WAU for a specific dimension combination on a given date
- **Time aggregation**: Daily rolling windows with 7-day lookback periods, where each day's WAU represents distinct users active in that day plus the previous 6 days
- **Final calculation**: Extract distinct user counts from merged HyperLogLog sketches using `hll_count.extract()` function, providing approximate but highly accurate WAU values. The `finalize_metric_user_wau` macro applies schema standardization and fills missing days with zeros to ensure complete time series data

*This methodology was automatically generated by Token Terminal's methodology analyzer.*

Unique revenue-generating users over a rolling 7-day window.

CompoundCOMP

Active users (weekly)

Active users (weekly)

Competitive landscape

Chain breakdown