Capstok — learn by doing

Why this matters

TVL (Total Value Locked) is the headline metric but the worst proxy for protocol health. A protocol can have $1B TVL of mercenary capital chasing emissions, or$ 100M of sticky users actually using the product. Knowing how to measure REAL engagement — active addresses, transaction count, revenue per user — separates real DeFi analysts from speculators.

Demo

Decompose protocol health.

Try it yourself

On Dune Analytics, query the daily active addresses for Uniswap V3, Aave V3, and one smaller protocol. Compare.
Compute retention cohorts for a protocol: of users who deposited in Jan, what % still have funds in July?
Identify two protocols with high TVL but low DAU — usually farming-heavy. List your concerns about their long-term health.

Prompt your AI

Use these three in order. Each builds on the one before.

1. Basics & terminology

In one paragraph, explain why TVL alone is a misleading metric.

2. Why it works (the mechanism)

Walk me through computing retention cohorts for a DeFi protocol.

3. Advanced — application & what's next

Given a protocol with $500M TVL but only 200 DAU, design the investigation that determines whether it's a viable product or a mercenary farm.

References

BAD METRICS (gameable, misleading):

  TVL alone: easily inflated via incentives; doesn't reflect engagement
  Total volume: dominated by wash trading
  Total users: usually sybils + mercenary actors
  "Active": vague — what's the timeframe?

GOOD METRICS (hard to game, predictive):

  Daily Active Users (DAU): unique addresses doing ≥1 tx in 24h
    Calculated: COUNT(DISTINCT from_address) for protocol's contracts
    Healthy: 1k-100k DAU for major protocols

  Median Position Size:
    Many small holders = healthy retail; few whales = concentration risk

  Average Position Duration:
    How long does the median position stay deposited?
    Sticky: months. Mercenary: hours-days.

  Revenue per User:
    Total monthly fees / monthly active users
    Real product-market fit signal

  Retention Cohorts:
    Of users who deposited in month X, what % still deposited month X+6?
    Best signal of real product-market fit
    Healthy: 30%+ retention at 6 months

  Fees vs Emissions Ratio:
    real_fees / emissions
    >1: protocol generates more than it pays out
    <0.5: dependent on emissions

  Stable Coin Inflow:
    DEX volumes spiking in USDC = real users
    DEX volumes spiking in protocol's own token = circular
    Healthy: 60%+ of volume in major stables

  TVL / Volume Ratio:
    Low (<5x): high capital efficiency
    High (>20x): inefficient; capital is idle

EXAMPLE COMPARISON:

  Protocol A:                     Protocol B:
    TVL: $1B                        TVL: $100M
    DAU: 500                        DAU: 50,000
    Median position: $2M            Median position: $200
    Fees: $500K/month               Fees: $5M/month
    Emissions: $5M/month            Emissions: $0
    Real APY: 0.6%                  Real APY: 60%

  Headline: A looks 10× bigger
  Reality: B has 100× more real engagement and 10× the revenue

WHERE TO GET THIS DATA:

  Dune Analytics (open-source SQL):
    Custom queries for any chain + protocol
    Free for non-commercial use

  Token Terminal:
    Pre-built dashboards for major protocols
    Subscription for advanced metrics

  Nansen / Arkham:
    Address-level tracking, behavior segmentation
    Commercial tools

  Etherscan / block explorers:
    Raw data; build your own metrics

TVL, Active Users, and Real Engagement Metrics