Whitepaper · v1.0 · November 2026
Larefe: wallet-level product analytics
A one-page description of what we measure, how, and what we will not claim.
1. Problem
Onchain products are measured in wallets, but decisions are about people. Industry-wide there are roughly 181M monthly active addresses and an estimated 40–70M active users (a16z, 2025). In incentive programs the gap widens: one L2 airdrop review flagged about 40% of eligible addresses as sybils. Web analytics, meanwhile, stops at the wallet connect, so the step from visitor to depositor is invisible to both stacks.
2. Event model
One stream holds three kinds of events: frontend events (page views, clicks, wallet connects) from a web SDK; decoded program activity on Solana (instructions, transfers, failures with reasons); and offchain tables the team brings. Intent, submission and success are separate events, so failed transactions never count as conversions.
3. Clustering
Wallets are grouped into likely-same-owner clusters from funding graphs (common funder within two hops), shared deposit addresses, device continuity observed by the SDK and timing. Every metric can be counted by raw wallets or by estimated users (clusters), and both are always shown together.
4. Farming flags
Clusters whose behaviour matches farming — synchronized claims inside a short window, a common funder, large size (one airdrop cut clusters above 20 wallets) — are flagged with a confidence level. The heuristic and threshold are visible and editable. Flags change what is excluded only where the user asks.
5. Measurement
- Funnels mix web and onchain steps in order, inside a conversion window.
- Retention cohorts by first connect or first deposit, daily or weekly.
- Attribution is first- or last-touch inside a lookback window; lift needs a holdout.
- Monitors open alerts with the query, wallets and events attached.
6. $LAREFE
Usage is priced per event: 1M free each month, then from $0.00005 per event, falling with volume. $LAREFE pays invoices at 20% off; staking 25,000 $LAREFE adds 1M free events a month; holders vote on chain and decoder coverage. It confers access and discounts, not yield or any claim on revenue.
7. Limits
- Clusters are estimates of common control, not identities. We do not deanonymize.
- No farming filter catches every farmer; stricter filters also exclude real users.
- User counts derived from wallets are model estimates and are reported as such.
- Attribution shows correlation; causal lift requires a control group.
- Coverage is per chain with stated freshness; we do not claim every chain.