near_health
Long-run context
77 months of NEAR mainnet data and a composite Health Score computed without look-ahead (src/model.py).
What the data says
Generated by src/model.py from the numbers below. Rerun it and these sentences update.
Before vs after the break
Monthly charts
Monthly per-day averages. The dashed line marks the structural break the model found. Log scales where values span orders of magnitude. Hover for values.
Replication check
Our quarterly averages against the figures Nansen published in its NEAR quarterly reports. If these did not match, nothing else here should be trusted.
76% of NEAR's transactions vanished in 48 hours.
What happened between 31 March and 2 April 2026, whose traffic it was, and why it distorts every year-on-year number quoted about NEAR.
48 hours in April
Daily values, indexed to 100 = average of the 30 days before the break. Transactions collapse. Active accounts barely move.
Fingerprint of the traffic that left
Reading: whatever stopped was sending ~3M transactions a day, creating ~110k new accounts a day and calling ~1.9M different accounts a day, yet it accounted for at most about 3% of daily active accounts. That is the signature of a small set of very high-frequency senders. This data cannot name them.
Who it was
This part is not our finding. Credit where it is due.
A separate block-level analysis, "Usage vs Price: NEAR", sampled raw blocks on 25 March and 10 April 2026 and attributes most of the drop to one app: sub-accounts of users.kaiching (the KAI-CHING shopping-rewards app) went from about 48% of sampled transactions to 0%, roughly 63% of the total fall. Community posts at the time describe the project winding down. We found no official statement.
That fits the fingerprint above: an app relaying transactions to millions of user sub-accounts through a handful of signers produces huge transaction and new-account counts with almost no footprint in daily active signers. The remaining ~37% of the drop is unattributed.
The year-on-year trap
How to read it: logs add up exactly, so each year-on-year fall splits into drift before the break, the 48-hour step, and drift after it. ■ 48-hour step ■ everything else
Seven claims. Fixed thresholds. Five hold.
The investment case split into bull, bear and data-quality claims, each graded by code against thresholds stored in claims.json.
The scorecard:
NEAR against the chains making the same pitch.
Price and fees today, then five years forward: NEAR with and without the 1.6% issuance cut voted through in November, raced against every peer's minting and unlock schedule.
NEAR against AI, privacy and general-purpose chains
NEAR sells itself on AI and private execution, so we compare it with chains that make the same pitch, plus three general L1s. One method for all: price and market cap from CoinGecko, user-paid fees from DefiLlama. Fee yield = annualised 30-day fees ÷ market cap.
Read with care: Akash and Render fees are payments for compute, not gas. Zcash and Monero are payment chains with low fees by design, valued as money rather than on fees. NEAR Intents fees are gross: only part is converted into NEAR. Bittensor, Fetch.ai, Oasis, Secret and Aleo have no chain-fee series on DefiLlama, so they appear for price only. Daily transactions and active accounts are not compared because no single public source measures them the same way across these chains.
Every chain on the same six claims
Each bubble is a chain: further left means less new supply over five years, higher means a better composite score (average percentile on price, dilution, unlocks, value returned, fees and five-year supply). Bubble size is market cap, colour is the rating from the same rule NEAR's scorecard uses. Click a bubble or a row.
NEAR five years out: with and without the 1.6% cut
HSP-029 passed the House of Stake vote on 2 November 2026 with 99.97% of participating veNEAR. It lowers the issuance cap from 2.5% to 1.6% in a straight line over 24 months. It is not law yet: the Security Council can veto it until 16 November, validators must adopt it in a protocol upgrade, and a 90-day grace period follows ratification. The model assumes the ramp starts 14 February 2027.
Assumptions: issuance runs at the measured share of the cap (2.47% of a 2.5% cap); buybacks and burns start from the last 30 days ($1.80M and $101k) and grow at the rate you set; bought NEAR is held, not burned, so it does not reduce supply. Price is held fixed. Not a forecast of price.
Five years of new supply, chain by chain
Press play. Each bar fills with the tokens a chain is scheduled to add from November 2026: █ rewards (minted or paid from a pool), ▒ unlocks to investors, team and reserves. Rows re-sort as they go. Hover a row for its source.
For every $100 of new NEAR, about $16 comes back.
Issuance, gas burn and NEAR Intents buybacks over the last 30 days, and the same sum for nine peers, with each chain's supply growth split into minting and unlocks.
Minted vs returned: issuance, burn and buybacks
Where each chain's new supply comes from (12 months)
Each chain's supply change is split into new tokens minted as rewards, rewards paid from a pre-minted pool, unlocks of investor, team and treasury tokens, and change no schedule explains. Every row is checked against a second source: the chain itself, a fixed emission rule, or a published unlock schedule.
Rewards issuance vs value returned
Coverage = burns and buybacks (DefiLlama, last 30 days annualised) ÷ value of rewards issuance (minted + pool rewards) at today's price. Unlocks are left out of coverage: they dilute holders but no network pays for them.
Read with care: NEAR's 30-day coverage (16.4%) and its row here (23%) differ because this table values a full year of rewards at today's price, after a sharp fall this month. Render's 107% rests on DefiLlama's burn series ($747k in 30 days); a separate report puts cumulative Render burns at about 1.56M RENDER by October 2026, which would imply far lower annual burns and coverage nearer 15–20%. We could not settle which is right. Akash shows 0% because DefiLlama records no burns, although Akash's burn-mint upgrade (March 2026) burns some AKT. CoinGecko's year-ago supply was wrong for Monero and Zcash; both now use their emission rules. Per DefiLlama's note, Intents revenue has bought NEAR on the open market since 23 Feb 2026 and that NEAR is held, not burned, so buybacks do not reduce supply. Zcash, Monero and Akash show no burn or buyback in DefiLlama's data. Bittensor has supply data but no holders-revenue series; Fetch.ai, Oasis, Secret and Aleo are left out.