The metrics that separate noise from signal.

00Why this matters

Most price action is noise. The job is figuring out which numbers actually tell you something, and which ones look smart but mean nothing.

At @icmanalytics we bet on teams that ship and on tokens where the fundamentals quietly compound. Every dashboard on this site is built around the metrics below. Learn them once and the entire platform starts reading like a tape.

This page used to be a glossary. It is now the OS - the same frameworks we use on every project page, the same on-chain signals we surface on every dashboard, the same scoring system we apply to every Bittensor subnet, every launchpad, every Base token. If a metric appears here, it appears live somewhere on this site.

One ground rule before we start: every metric on this page can be gamed. We tell you what each one means AND where it lies. That second half is the actual edge.

What you will learn (TL;DR)

• Valuation - P/E, Buyback P/E, MC vs FDV, revenue vs volume, GMV vs revenue, APY vs ROI, vesting schedules, and the four jobs a token can do.
• On-chain signals - holder cohort flow across Retail / LP / CEX / Treasury / Whales, smart-money labeling and the cabal tracker, Jupiter limit-order reads, DCA In/Out as intent, and liquidity-adjusted exit value on the x*y=k curve.
• Sector walkthroughs - prediction markets with sport-by-sport edge breakdown, DeFAI vault yield (share-price-based) and settlement lag, win rate vs EV, and the Bittensor dTAO loop with our 35/40/25 master score.
• Intelligence layer - Trust Score and the four verification tiers, AI scout convergence as a leading indicator, user metrics + App Store signals, and the full risk roster.

01P/E Ratio (Price-to-Earnings)

The cleanest valuation question in crypto: how many years of earnings does the market cap cost?

Take the market cap, divide by annualized revenue. That is it.

How to read the number

• Lower P/E = better value. You pay less for each dollar of earnings.
• Higher P/E means the market is pricing growth, or the token is too expensive.
• A P/E of 10 means you are paying 10 years of current earnings for the asset.

What the bands mean

• Under 5 - Either undervalued, or revenue is not durable. Look harder.
• 5-15 - Reasonable price for the earnings. Most launchpads live here.
• 15-25 - Fair. The S&P 500 sits in this band (around P/E 20-25).
• 25-50 - The market is pricing growth. Demand a growth story to match.
• 50+ - Hyper-growth, or a vibes trade. Be honest about which one.

Why we care about this in crypto

On-chain protocols regularly trade at P/E under 10 with revenue that is GROWING, not shrinking. That is the bet we take. A launchpad at P/E 3 is priced like a small business and behaves like a software company - and the gap closes one direction.

Where it lies

Annualized revenue from one hot week is fiction. A token can show a P/E of 1 because last Tuesday's volume was 100x the trailing average. ALWAYS look at the 30-day revenue curve, not the 24h snapshot. We chart both.

02Market Cap vs FDV

Two prices, same project. If you only look at market cap, you are pricing today. If you only look at FDV, you are pricing the moment every unlock hits the open market.

Market Cap (the price now)

What every token in the market is worth right now, at the current price.

Fully Diluted Valuation (the price after every unlock)

What the project would be worth if every locked, vested, and to-be-minted token were trading today at the same price.

Why we read both

• Market Cap shows what the project is worth today.
• FDV shows what it would be worth if nothing else changes and every insider dumps.
• The GAP between them is the future supply schedule. Big gap = future selling pressure already baked in.

Where it lies

"Low circulating supply" headlines are a trap. A 5% float at $1 looks cheap until the 95% comes online. Read the emission schedule BEFORE you read the chart - a 5x cliff three months out is a calendar, not a fundamental.

03Revenue (24h)

What the protocol actually earns in the last 24 hours - trading fees, protocol fees, swap fees. Not volume. Revenue.

We multiply by 365 to annualize and compare against market cap. That is the P/E numerator.

Volume is NOT revenue

An exchange that does $1B/day in volume but charges 0 bps is a $0 revenue business. Don't confuse the flow with the cut. Always ask: what does the protocol actually keep?

Tiered fee structures change the read

Some protocols charge different fees at different market-cap thresholds. Pump.fun's fee structure has 25 tiers - the creator/protocol/LP split shifts from 0.95% on tiny tokens down to 0.05% on a token that has graduated past the bonding curve. Same protocol, very different revenue per dollar of volume. We surface the tier breakdown on the Launchpad page so the revenue number means something.

Where it lies

One whale, one wash-trader, or one airdrop farm can blow up a 24h number by 10x. The 30-day median is the honest read. We also flag any day where revenue spikes more than 3x the trailing average so the spike does not lie.

04GMV vs Revenue

GMV is everything that gets transacted. Revenue is the cut the protocol keeps. Both matter, for different reasons.

Worked example

An e-commerce protocol facilitates $100M in sales at a 3% fee:

• GMV = $100M (the scoreboard for adoption)
• Revenue = $3M (the scoreboard for the business model)

High GMV with no revenue is a free service waiting to monetize. Strong revenue with low GMV is a small-but-profitable cult product. Healthy is both growing.

Real example

$DUPE (Dupe.photos) operates an AI-powered price-discovery marketplace - finding identical products at 60-80% less than retail. The token captures value from the marketplace's transaction flow. Current GMV, user count, and per-day fees are tracked on the DUPE Dashboard - we publish the live numbers there because they move.

Where it lies

GMV can be inflated by self-trades, internal swaps, and bot-on-bot traffic. Pair the GMV with the unique active users in the same window - real adoption shows up in both.

05Token Economics

Total supply is the cap. Circulating supply is what is trading. Burns destroy tokens. Buybacks take them off the market. The arithmetic decides whether holders compound or get diluted.

Burns

Tokens permanently removed from circulation. Cuts total supply, raises scarcity if demand stays constant. The interesting burns are funded by protocol revenue, not by promises.

Buybacks

Protocol revenue used to buy the token off the open market. The closest thing crypto has to a dividend - the better the protocol performs, the more buying pressure shows up on the chart. The full loop is in the next section.

Circulating vs Total supply

• Circulating - tokens trading in the open market right now.
• Total - the cap on tokens that will ever exist (including locked, vested, unminted).
• The ratio is the float. Low float + high FDV = early-stage, high-dilution-risk.

Token utility - the four jobs a token can do

• Governance - vote on protocol parameters. Useful in theory, often ignored in practice.
• Fee accrual - holders earn a share of protocol revenue. The only utility that maps to a P/E.
• Discount or access - holding the token reduces fees or unlocks features. Real demand driver if usage scales.
• Stake for security - the token secures the network. Demand floor scales with TVL.

A token with all four jobs is rare. A token with none is decoration on a brand.

06The Buyback Flywheel

The cleanest tokenomic loop in crypto is the buyback. Revenue comes in, gets converted to tokens on the open market, and either gets burned or redistributed. Holders compound. The token chart starts looking like an earnings chart.

How the loop runs

Real loops on this site

• $SIRE - the aVault's profits from on-chain sports betting fund buybacks plus burns. The loop is profit-gated: when the vault is in drawdown (as it currently is mid-2026), the buyback pauses. The SIRE Dashboard tracks the rolling buyback yield AND the drawdown periods - that honest framing is the point.
• $DUPE - marketplace GMV funds a buyback paired with a burn. We map the 8-step DUPE flywheel (Deals -> Sharing -> Earners -> Partners -> Selection -> Redemption -> Buybacks -> Burns) on the DUPE dashboard.
• Pump.fun, MetaDAO, Meteora - protocol fees fund buybacks. We track the 30-day rolling buyback yield AND the Buyback P/E.

Buyback P/E - the sharper metric

Standard P/E uses today's market cap divided by annual revenue. Buyback P/E uses the market cap on the days the protocol was ACTIVELY buying back. It tells you what the protocol thinks its own token is worth at the moment of capital deployment - and it is usually lower than the average price.

30-day rolling buyback yield

Tokens bought back over the last 30 days, divided by market cap, annualized. This is the compounded return from buybacks ALONE, before any price action. A protocol running a 12% rolling buyback yield is the equivalent of a 12% dividend stock - except the dividend is paid in compressed supply, not cash.

Where it lies

A one-time buyback funded by treasury is a marketing event. A sustained buyback funded by recurring revenue is the loop. Always ask: where does the money come from? If the answer is the treasury balance, the loop is finite. If the answer is monthly fees, the loop compounds.

07Vesting & Unlocks

The emissions schedule is the second price of any token. Spot is what most people see; the unlock calendar is what determines what spot WILL be in 90 days.

The shapes that matter

• Cliff unlock - a chunk releases all at once on a date. Brutal if not telegraphed in advance.
• Linear vest - tokens drip out over months or years. Gentler, but creates a constant headwind.
• Performance vest - tokens unlock only when KPIs are hit. Aligned, but rare.
• Fair launch - no team or VC unlocks, no allocations. The hardest to find, the easiest to bet on.

What we check before sizing in

• The 90-day forward unlock schedule. Anything over 5% of circulating supply dropping in one week is a calendar event, not a fundamental.
• The team + VC % of total supply. Over 30% allocated to insiders is a structural overhang.
• The float ratio (circulating / total). Under 0.2 with high FDV means most of the dilution has not arrived yet.

Buyback yield offsetting vesting

If a token has a 5% annual buyback yield AND a 10% annual emission schedule, net dilution is 5%, not 10%. We compute the net number on every research page so you can see whether the loop is winning the race against the emissions.

Where it lies

Vesting cliffs are public information, but the WALLET destinations are not always obvious. Teams sometimes split allocations across many sub-wallets to obscure the drop. We label the major insider wallets on every per-token dashboard and surface a concentration alert when a single one drops 30%+ in any 24h window.

08APY vs ROI

ROI tells you how much you made. APY tells you how fast you made it. Confuse the two and a 30-day pump looks like a permanent edge.

ROI (Return on Investment)

Invest $10,000, now worth $10,571. ROI = 5.71%.

APY (Annual Percentage Yield)

Same 5.71% ROI over 33 days = 63.2% APY. The same 5.71% over 365 days = 5.71% APY. Time is the multiplier.

Why APY matters for comparison

• Normalizes time - a 30-day vault and a 90-day vault become comparable.
• Assumes compounding - what reinvestment looks like, not a single payout.
• The DeFi standard - every protocol reports APY, so we do too.

Where it lies

An APY annualized off 7 days of unusual returns is a marketing number. Demand a long sample, or sit out.

09Holder Concentration & Cohort Flow

The top 100 holders of any token tell you most of what you need to know. WHO they are tells you the rest.

The five cohorts

The metrics cohort labeling produces

• Retail net flow - top-100 retail buy minus sell over 24h and 7d. The cleanest sentiment read on-chain.
• CEX hot wallet net flow - users moving tokens TO an exchange to sell, or pulling FROM the exchange to hold. Real flow, no airdrop noise.
• Net flow signal - aggregate cohort flow in USD over rolling windows. Direction = the read on whether smart capital is rotating in or out.
• Dropped holders - wallets that exit the top 100. We track them for 3 days after the drop. If they keep selling on day 4, it is a real exit. If they stabilize, it was rotation.

Where it lies

Wallet labels rot. A CEX address from 2024 may be reassigned by 2026. We refresh the labels on every dashboard reload and flag any wallet whose label confidence is below 80%. Don't make a call off a single-wallet flow; look at the cohort.

Read next: the cohort flow tells you WHO is moving, the Smart Money section tells you WHICH wallets are worth following, and Limit Orders & DCA tells you what those wallets PLAN to do at prices that have not yet printed.

10Smart Money & Wallet Labeling

Most wallets are noise. A handful of wallets are signal. The work is figuring out which is which BEFORE the trade, not after.

What makes a wallet smart

• Realized PnL above $1M+ across at least 6 months of activity.
• Win rate above 60% on positions held more than 24h (cuts the snipers).
• Holds across multiple successful tokens, not one lottery winner.
• Low correlation with broader retail flow. They get in before the herd, not with it.

How we find them

• @CieloFinance - per-wallet PnL tape, the cleanest filter for "is this address profitable".
• @nansen_ai - smart-money labels across chains, with sub-categories (whales, institutional, fund).
• @arkham - entity labeling. Find the OTC desk, the market maker, the fund treasury.
• Our own dashboards - cohort-labeled flows on Base L2 and Solana on every ICM Research page.

The cabal tracker (Solana)

On Solana we run a cabal tracker that maps the network of wallets funded by a handful of OG addresses. Same funding source + same trade timing = same hand. The label persists across token rotations - the bag is different, the playbook is the same. We mark these on every token's holder chart.

Where it lies

A smart wallet that suddenly becomes INACTIVE is more dangerous than one that keeps trading. Silence on a known-sharp address often means they sold the position and are waiting for the bid. Watch the cessation as carefully as the entry.

Cross-check: if you spot a smart wallet accumulating, jump to Limit Orders & DCA to see whether the broader book confirms - and to Liquidity Depth to see whether the position is even exitable at size.

11Limit Orders & DCA Intent

Spot price tells you where the trade clears. Limit orders and DCA programs tell you where the trade WANTS to clear. That gap is often the next move.

Jupiter limit orders (Solana)

Jupiter aggregates a public limit-order book across Solana DEXs. Each order is a wallet declaring "I will buy/sell at exactly this price". Aggregate them and you get a real-time supply/demand map below and above spot.

How we read the book

• A wall of buy limits 5% below spot = real support. If price falls into it, those orders fill.
• A wall of sell limits 5% above spot = real resistance. If price climbs into it, those orders fill and stall the move.
• A thin book on both sides = volatile range, easy to move price, hard to size in.

DCA In / DCA Out

A DCA (dollar-cost average) program is a recurring automated buy or sell. We surface the total being DCA'd in versus DCA'd out per token per day. Heavy DCA in = systematic accumulation that survives short-term price moves. Heavy DCA out = systematic distribution that signals a long-term exit.

Where it lies

Limit orders and DCA can be SPOOFED by sophisticated players. Always pair the orderbook signal with on-chain holder flow. If the limit book is bullish but the top-100 cohort is dumping, the orderbook is decoration. We cross-check both on every Solana token dashboard.

12Liquidity Depth & AMM Mechanics

Liquidity is not the same as price. A token can have a $0.10 spot price and a max sellable size of $5,000 - because the pool curve is steep. Liquidity-adjusted exit value is the only honest read of what your position is worth at the size you actually hold.

The constant-product curve (x * y = k)

AMMs like Uniswap V2, Raydium, and PumpSwap use a constant-product formula: pool A balance times pool B balance equals a constant. The curve is hyperbolic - the bigger your trade relative to the pool, the worse your fill.

What this means in practice

• A $1M position in a token with a $50K liquidity pool is NOT a $1M position. It is a $1M paper position with a maybe-$20K realized value.
• Liquidity-adjusted exit value = the actual USD you would receive if you sold the position into the existing pool, accounting for slippage along the curve.
• We compute liquidity-adjusted exit value on every Solana wallet PnL dashboard. The paper number flatters; the realized number is the truth.

Concentrated liquidity (Uniswap V3, V4, Orca, Trident)

Newer AMMs let LPs concentrate their liquidity in a price range, which gives better depth at the active price - but creates DARK ZONES above and below where price moves through with almost no resistance. Watch the active-range LP positions; if the trade leaves the band, the chart can fall fast.

Where it lies

FDV-to-liquidity ratio is the honest size of a market. An $80M FDV with $30K of liquidity is not an $80M market - it is a paper print that will reprice the moment $200K hits the bid.

Loop it back: Market Cap vs FDV is the headline; liquidity-adjusted exit value is the truth. Any time a chart looks "cheap," check FDV/liquidity before you size in.

13Prediction Markets

Markets where the wisdom of the crowd has a price tag. The price IS the probability.

You buy shares in an outcome. If it happens, your share pays $1. If not, $0. A $0.60 price is the market saying 60% odds, no more, no less.

How they work

• $0.60 share = market believes 60% probability
• $0.30 share = market believes 30% probability
• Profit = Payout - Price Paid (if you are right)

Why prediction markets are a different asset class

• The book never bans winners. Sportsbooks limit winning bettors. Prediction markets need liquidity, not losers.
• The capital scales. $500K today, $50M tomorrow - same rails, no bookie phone call.
• No hidden vig. No spread baked into the line - the price is the probability.
• Decentralized, censorship-resistant. No single point of failure or political kill switch.

Sport-by-sport edge breakdown

Not all sports are equally exploitable. We track win rate AND P&L per sport on the SIRE dashboard. Basketball books are tighter than baseball books - basketball edges are smaller but more consistent. Soccer has the widest dispersion and the rarest edges. NFL is the deepest liquidity but the sharpest book. Knowing WHICH sport carries the edge in a given window matters as much as the headline win rate.

Probability vs price divergence

Liquidity-driven repricing can push a Polymarket line to 70% well before the underlying probability changes. Sharp players watch for divergence between the model probability and the market price - that gap is the trade. A model showing 55% odds on a line trading at 70% is a short signal; same model at 55% on a line at 40% is a long.

The two anchors

• Kalshi - first CFTC-regulated event-contract exchange. Valued at $11B in the latest round.
• Polymarket - decentralized prediction market on Polygon. Valued at $10B+.

Where it lies

• Volume can be bot-driven near close - liquidity spikes do not always mean conviction.
• Low-liquidity country-specific markets can be one-sided pumps, not real consensus.
• Resolution risk - some markets resolve on a single source (a government tweet, an obscure scoreboard) that can be late or disputed. Read the resolution criteria BEFORE you size.

14DeFAI & On-Chain Vaults

The protocol holds the funds. The smart contract enforces the rules. The AI makes the decisions. No human at the keyboard, no fund manager skimming 2-and-20.

The shape

• AI models trained on history pick the trades or bets.
• Smart contracts execute them on-chain - every trade visible, every settlement final.
• Token holders share the upside through buybacks, burns, or revenue splits.
• 24/7 operation - the strategy runs whether the team sleeps or not.

Vault yield vs cumulative P&L

There are two ways to report a vault's return. Cumulative P&L is the sum of all wins and losses since inception. Vault yield is the change in share price (NAV per unit) over a window. Yield accounts for new deposits and withdrawals diluting or concentrating ownership. ALWAYS use share-price yield for comparing across vaults - cumulative P&L flatters big-AUM vaults that grew by deposit, not by edge.

Settlement lag

On-chain vaults with off-chain settlement (bets, options expiries, scheduled payouts) often have a 24-72 hour withdrawal lag after profit lands. The protocol has to wait for the bet to clear before crediting your balance. It is structural, not a red flag - but it is a real constraint on your liquidity that the headline APY does not show.

Staking tier retention

Smart vaults segment depositors by staked amount and report retention per tier. A vault where the 100K+ tier retains at 80%+ and the under-1K tier retains at 30% is fine - the conviction tail is sticky. A vault where the BIG tiers churn is a different story. Look past the headline TVL to the tier breakdown. The SIRE dashboard surfaces staking tier retention on every refresh.

What you actually own

For the first time you can back a strategy team AND own the productive asset. The model improves, the buybacks compound, the tokenholder is the LP. No prime broker, no minimum investment, no allocator gate.

15Win Rate & Expected Value

Win rate is how often you are right. Expected value is whether being right pays. Confuse the two and you blow up a profitable strategy chasing a higher win rate.

Win Rate

A 58% win rate means winning 58 of every 100 bets placed. That number alone tells you almost nothing about whether you are making money.

Why win rate alone is a trap

• Odds matter more. Winning 70% of -200 bets is WORSE than winning 40% of +300 bets.
• Stake sizing matters. One big loss can wipe out fifty small wins.
• Edge vs variance. Short-term win rate is noisy. Long-term EV is the signal.

Expected Value (EV)

Positive EV means the bet is profitable across a large sample, regardless of any single outcome. This is the only number that survives the long run.

What we want from a betting strategy

• Win rate above 55% - real edge.
• Positive EV across the full sample - real strategy.
• Tight stake sizing - survives the drawdowns that always come.
• Long sample, not 200 bets - the law of large numbers is the only honest test.
• Rolling 7d/30d performance, not just all-time - regime shifts show up in the recent windows first. The all-time number can mask a model that has stopped working.

16Bittensor: dTAO, Subnets, and Scoring

Bittensor is the on-chain market for machine intelligence. Each subnet is a vertical (sports analytics, text-to-image, chat, audio) where miners produce work and validators score it. Tokens emit on a market-driven curve called dTAO. Understand the structure and the rest of the alpha lives in subnet-by-subnet ratings.

The pieces

• $TAO - the network token. Backs every subnet's alpha token via the dTAO mechanic.
• Subnet - a vertical with its own miners, validators, and use case. Subnet 11 (@dippy_ai) does roleplay AI. Subnet 44 (CrunchDAO) does sports analytics. Over 100 subnets are live.
• Alpha tokens - each subnet has its own token, priced against $TAO via the dTAO bonding curve. Bid up the alpha = bid up the subnet's emissions.
• Validators - score miner output, earn TAO emissions, and decide which alpha tokens pull harder on emissions.
• Miners - do the actual work (run the models, serve the requests). Get emissions in alpha.

The dTAO feedback loop

More buyers of an alpha token leads to higher alpha price, leads to more emissions to that subnet, leads to better miners joining, leads to better product, leads to more demand for the alpha, leads to more buyers. The loop runs both ways. A subnet with a bad model and a falling alpha sees emissions drain and miners leave. Bittensor is the cleanest market in crypto for the question "is this AI product getting better?"

How we score subnets

Every Bittensor subnet on this site gets a 0-100 master score, weighted three ways:

• Market (35%) - alpha token mcap, liquidity, recent flow, dTAO bonding position.
• GitHub (40%) - commit frequency, PR activity, code quality, contributor count. The single highest-weighted dimension because shipping is everything.
• AI fundamentals (25%) - BRAID-cascade LLM analysis of whether the subnet's work is actually useful, well-built, and shipping.

The grade ladder

Rebuilt nightly. A subnet at 85+ is S-tier. A subnet under 40 will get emitted out of the rankings.

Where it lies

Alpha token prices can pump on emissions speculation WITHOUT the underlying product getting better. Always cross-check the GitHub line and the AI fundamentals line - if the score is 75 but GitHub is 20, you are looking at a market-only pump and the loop will fail. Other risk: $TAO halvings cut absolute emissions, so a subnet that depends purely on emission flow has structural headwind on every halving cycle.

Cross-check: the master score is the verifiable layer; pair it with the Trust Score to see what is on-chain vs claimed, and with Scout Convergence to see whether the narrative is converging on the same subnet.

17Trust Score & Verification Tiers

Every metric on this page can be on-chain or claimed. The Trust Score is the discipline of saying which is which BEFORE the analysis, not after the fact.

The four tiers

The Trust Score read

Each ICM Research project page carries a Trust Score from 0 to 100, broken into 5 buckets:

• On-chain data (verifiable)
• Product / traction (some verifiable, some claimed)
• Team (background check + history)
• Tokenomics (vesting, supply, control wallets)
• Community (engagement, growth, sentiment)

A 60+ project with mostly Verified entries is fine to size. A 60+ project with mostly Self-Reported entries is a different bet - the score is the team's claims about themselves. Read both columns, not just the headline.

Where it lies

Verified data can still be misleading (one whale wash-trading themselves is verified volume), and Self-Reported claims can still be true. The Trust Score is not "this is real" - it is "here is what we could prove and here is what we could not." Always read both halves.

18AI Mindshare & Scout Convergence

Narrative is a leading indicator. Price catches up to mindshare, not the other way around. The trick is measuring mindshare BEFORE the chart confirms it.

What we track

• $TICKER mention frequency across X over rolling 24h, 7d, 30d windows.
• Sentiment polarity per project (bull, bear, neutral).
• AI-scout convergence - whether multiple AI analytics agents (@aixbt_agent, @pineanalytics, @0xJeff, @gkisokay, others) are independently mentioning the same project in the same window.

Scout convergence tiers

How we use it

Scout convergence is our radar, not our buy signal. When 3+ scouts pick up a project in the same week, we open the dashboard and check whether the FUNDAMENTALS justify the attention. Most of the time they do not. The ~20% of the time the chart has not moved yet AND the fundamentals match - that is the trade.

Where it lies

AI scouts can be gamed. A team that pays for coverage can manufacture a convergence event. Cross-check: does the on-chain holder flow show conviction buying, or is it all wash and bot-traded volume? Narrative without flow is a pump-and-fade.

19User Metrics & Retention

Past the financial numbers, the product question: do people actually use it, and do they come back? Teams that ship show it in the user numbers.

Monthly Active Users (MAU)

Unique users who touch the product at least once a month. The reach number.

Daily Active Users / MAU (the stickiness ratio)

The ratio that tells you whether the product is a habit or a one-time visit.

• DAU/MAU above 50% - daily habit (your phone is the model).
• DAU/MAU 20-50% - good engagement, regular use.
• DAU/MAU under 20% - occasional, on-demand. Fine for some products, fatal for others.

Retention

The cleanest product-market-fit signal there is. Anything above 70% monthly retention is a product people NEED. Below 30% is a product they tried.

Off-chain signals matter too

For consumer-facing products we also track App Store ratings, review counts, and review sentiment as on-chain-adjacent signals. A wallet with a 4.8 iOS rating and 50K reviews is doing something right that the on-chain wallet count does not capture. The AVICI Dashboard integrates App Store metrics next to the on-chain holder data, refreshed in line with the rest of the site.

Why this matters for valuation

User growth drives transaction count, which drives fees, which drives token value. But quality > quantity:

• Retention > acquisition - 1,000 retained users beat 10,000 churned users.
• Revenue per user - the higher this gets, the deeper the moat.
• Organic growth - users referring users is the only durable acquisition channel.

20The tools we actually use

No metric on this page is useful without a place to pull it from. Here is the working stack - what each one is for, in one line.

21Risks to Monitor

Every metric on this page can be gamed. The honest read of fundamental analysis is that you are not eliminating risk - you are just choosing which risks to take.

22Our thesis

In the AI era, agency wins. Teams that ship - and protocols that earn real revenue on-chain - are wildly underpriced versus the web2 comparables the market still uses to value them.

This page is the cheat sheet. The dashboards are the receipts. Read both, then make a call.

Follow @icmanalytics for daily fundamentals on the projects worth tracking. Let's build and connect for a decentralized future.