Skip to main content
NEAR Intelligence
Sign in

← All models

Intents · Fee activity

Published with limits

Do gross fees move with trading volume?

A model of daily changes in Intents activity.

Provider-reported gross fees, measured against reported trading volume.

gross_fee_activity_v1 · version 1 · Published 14 Nov 2026

Fitted observations
89
Fitted sample
6 Aug – 3 Nov 2026
Method
OLS / HAC · 7 lags
Source
SRC-LLAMA

Activity measure

Trading volume

Reported Intents volume in USD.

The model uses its daily log change.

Fee measure

Provider gross fees

Reported fees in USD, before any claim about protocol revenue.

The model uses their daily log change.

The important limit

Gross fees are calculated from volume through a fee rate. Their relationship is therefore partly built into the data. Both series share Dune lineage. This short sample cannot establish causality, demand elasticity, sustainable revenue, protocol revenue or token value.

What the model found

Published snapshot

Fees and volume move together in this sample.

The model estimates a positive relationship between their daily log changes. Because fees are constructed from volume, this mostly describes the fee mechanism. It does not answer whether Intents creates sustainable value for NEAR.

Estimated relationship · slope

0.8542

95% confidence interval · 0.5719–1.1366

00.30.50.81
Point estimate and 95% confidence interval

Dot: the point estimate. Line: the 95% confidence interval around it. A positive slope means the two daily log changes tend to move in the same direction in the fitted sample.

What this supports

A short-run association between the two reported series over the fitted observations.

What remains unanswered

Protocol revenue, sustainable fee income and value captured by the NEAR token.

The recorded interpretation (verbatim)
The recorded interpretation block
{
  "sentences": [
    "The fitted slope on 'dlog(provider_intents_volume_usd|SRC-LLAMA)' is 0.8542493147420771 (HAC standard error 0.14203451836578282, 95% CI [0.5719403297667323, 1.136558299717422]): a short-run association between contemporaneous transformed values of the declared series; it is not a causal effect and not an elasticity of demand.",
    "The estimate uses 89 observations between 2026-08-07 and 2026-11-03; a window of 90 calendar days cannot establish a long-term stable mechanism.",
    "The ADF unit-root null is not rejected for at least one level series; the level is treated as nonstationary and the preregistered difference baseline stays the published form.",
    "The KPSS stationarity null is not rejected for any level series; short samples make this weak evidence, not proof of stationarity.",
    "Ljung-Box does not reject the no-autocorrelation null at the declared lags; with this sample size that is weak evidence, not proof of a correct specification.",
    "Breusch-Pagan indicates heteroskedasticity; the preregistered response is the already-declared HAC covariance — no reweighting, no outlier removal and no specification change was performed.",
    "Influence check: 4 observation(s) exceed the preregistered Cook's distance threshold (0.0449438202247191); they are reported with dates and remain in the sample — removal requires an operational reason and a with/without report.",
    "No protocol-revenue, sustainable-revenue, token-valuation, causal or demand-elasticity claim follows from this output: the fitted object is the named proxy association only, and the mechanical fee-on-volume construction means the slope largely restates the effective fee rate."
  ],
  "template_id": "bounded_association_v1",
  "authoring_note": "this is the bounded template default rendered by the diagnostics run; an authored operator interpretation, if any, is added at publication time (NI-054) and never replaces the attached limitations.",
  "gate_limit_codes": [
    "MECHANICAL_FEE_VOLUME_RELATIONSHIP",
    "PROXY_NOT_PROTOCOL_REVENUE",
    "OQ02_JUDGING_ACCEPTANCE_RISK",
    "SOURCE_REVISIONS_POSSIBLE",
    "SINGLE_PROVIDER_LINEAGE_NOT_INDEPENDENT",
    "SHORT_SAMPLE_NO_LONG_RUN_CLAIM",
    "NO_CAUSAL_IDENTIFICATION"
  ],
  "rendering_policy": "every sentence below is rendered from a fixed template in research/models/diagnostics.py with recorded numbers and recorded test decisions substituted; there is no free-form generation, no causal wording and no sustainability wording anywhere in this module.",
  "prohibited_labels": [
    "protocol_revenue",
    "token_valuation",
    "causal_effect",
    "demand_elasticity",
    "sustainable_revenue"
  ],
  "spec_limitation_notes": [
    "Mechanical construction warning (docs/ECONOMETRIC_METHODOLOGY.md): provider gross fees are reported as a fee applied to the same provider's reported volume, so a fee-on-volume association largely restates the effective fee rate and must be displayed before any coefficient interpretation.",
    "Both variables share one provider lineage (Dune); the association is not independent confirmation of anything.",
    "The output is the named proxy association only; it is never protocol revenue (OQ-01) and H1/H4 stay unresolved (OQ-02 judging risk)."
  ],
  "mechanical_construction_warning": "Mechanical construction warning (docs/ECONOMETRIC_METHODOLOGY.md): provider gross fees are reported as a fee applied to the same provider's reported volume, so a fee-on-volume association largely restates the effective fee rate and must be displayed before any coefficient interpretation."
}
Exact coefficients and inference

The summary above is rounded for display. The recorded values are preserved below, verbatim. No specification was selected using p-values.

TermEstimateHAC standard error
const-0.027697492071932340.03560372368015006
dlog(provider_intents_volume_usd|SRC-LLAMA)0.85424931474207710.14203451836578282
const · 95% CI-0.0984637481140075 to 0.04306876397014282
dlog(provider_intents_volume_usd|SRC-LLAMA) · 95% CI0.5719403297667323 to 1.136558299717422
const · t / p-0.77793806964563 / 0.43871601689821427
dlog(provider_intents_volume_usd|SRC-LLAMA) · t / p6.014378227003388 / 4.146222767879358e-8

Daily log differences · HAC covariance · R² 0.4344 (rounded)

The fitted observations cover 6 Aug – 3 Nov 2026. The input window is [6 Aug 2026, 4 Nov 2026) UTC; one observation is lost to the daily difference.

All recorded limitations · 7

Prohibited labels — this result must never be presented as:

  • protocol_revenue
  • token_valuation
  • causal_effect
  • demand_elasticity
  • sustainable_revenue

Recorded limitation notes (verbatim)

  • Mechanical construction warning (docs/ECONOMETRIC_METHODOLOGY.md): provider gross fees are reported as a fee applied to the same provider's reported volume, so a fee-on-volume association largely restates the effective fee rate and must be displayed before any coefficient interpretation.
  • Both variables share one provider lineage (Dune); the association is not independent confirmation of anything.
  • The output is the named proxy association only; it is never protocol revenue (OQ-01) and H1/H4 stay unresolved (OQ-02 judging risk).

Gate limit codes (all recorded codes visible)

MECHANICAL_FEE_VOLUME_RELATIONSHIP
Provider gross fees are constructed from reported volume through an effective fee rate; a fee-on-volume association largely restates that construction and must be displayed before any coefficient interpretation. It is not demand elasticity, not sustainable revenue and not token value.
PROXY_NOT_PROTOCOL_REVENUE
The fitted object is the explicitly named proxy only; it must never be labeled protocol revenue, and H1/H4 stay unresolved.
OQ02_JUDGING_ACCEPTANCE_RISK
External acceptance of a transparent reproducible proxy association while H1/H4 remain unresolved is an open judging risk (OQ-02).
SOURCE_REVISIONS_POSSIBLE
Provider-reported series are revisable; the saved snapshot is not declared final and a revision is a new dataset version, never an overwrite.
SINGLE_PROVIDER_LINEAGE_NOT_INDEPENDENT
All model variables come from one provider lineage; their association is not independent confirmation of anything.
SHORT_SAMPLE_NO_LONG_RUN_CLAIM
90 calendar day(s) cannot establish a long-term stable mechanism
NO_CAUSAL_IDENTIFICATION
Observational association only: endogeneity, unobserved incentives, common market trends and provider construction remain limitations even with robust standard errors.

Recorded admission reasons

MECHANICAL_FEE_VOLUME_RELATIONSHIP · limit
Provider gross fees are constructed from reported volume through an effective fee rate; a fee-on-volume association largely restates that construction and must be displayed before any coefficient interpretation. It is not demand elasticity, not sustainable revenue and not token value.
PROXY_NOT_PROTOCOL_REVENUE · limit
The fitted object is the explicitly named proxy only; it must never be labeled protocol revenue, and H1/H4 stay unresolved.
OQ02_JUDGING_ACCEPTANCE_RISK · limit
External acceptance of a transparent reproducible proxy association while H1/H4 remain unresolved is an open judging risk (OQ-02).
SOURCE_REVISIONS_POSSIBLE · limit
Provider-reported series are revisable; the saved snapshot is not declared final and a revision is a new dataset version, never an overwrite.
SINGLE_PROVIDER_LINEAGE_NOT_INDEPENDENT · limit
All model variables come from one provider lineage; their association is not independent confirmation of anything.
SHORT_SAMPLE_NO_LONG_RUN_CLAIM · limit
90 calendar day(s) cannot establish a long-term stable mechanism
NO_CAUSAL_IDENTIFICATION · limit
Observational association only: endogeneity, unobserved incentives, common market trends and provider construction remain limitations even with robust standard errors.

Every number above is subject to these limits. Nothing on this page is a causal effect, a demand elasticity or protocol revenue.