A public standard · free to adopt
The Conferral Metrics Standard
How to measure attention that was granted rather than captured.
The entire analytics industry measures capture — impressions, sessions, watch time, clicks. Nothing standard measures conferral: whether attention arrived granted, whether trust is accumulating or depleting, whether asks land inside windows. “What gets measured gets managed” is why the engagement economy strip-mined its own asset. It managed what it measured, and it measured the wrong kind of attention. This is the instrument panel for the right kind.
Version v1.2 · published 2026-08-29 · revised 2026-09-03 · free to adopt with attribution
Cite as Conferral Metrics Standard v1.2 · the definitions are free and are never paywalled
Seven metrics, each with a proxy a real business can compute this week.
Two of these are computed for you. Granted Share is what the Granted Share Calculator works out from your last twenty customers, and the deposit/ask ratio is what the Overdraw Self-Scorer reads off your funnel. Run one of those first if you would rather see a number than a definition — the definitions keep.
One
The metrics
Definitions, formulas and proxies. Take them; they are free.
Deposit/Ask Ratio and Window Placement Rate
DAR: in any sequence (funnel, email stream, sales conversation, feed presence), the count of genuine deposits per ask. WPR: the share of asks placed inside an open window (immediately following a deposit or a satisfaction event) rather than cold, late, or stacked.
Proxy: tag the sequence stage-by-stage (the Overdraw Self-Scorer does this tagging); and for WPR an ask counts as in-window if it directly follows a deposit/satisfaction signal and is the only ask in that window.
Predicts: conversion efficiency and list decay. These are the two numbers the main paper says no conventional analytics captures; the Overdraw Score (existing audit spec) is their composite.
Trust Stock vs. Engagement Flow
Definition: the correction paper's balance-sheet reframe, made measurable. Engagement flow = the period's captured-attention volume (the numbers everyone already tracks). Trust stock = the standing reserve of conferral: measurable as retention/return rates, opt-in depth (subscribed, whitelisted, replies), willingness-to-recommend, and — negatively — withdrawal events (unsubscribes, mutes, unfollows-after-ask, refunds) treated as trust-stock drawdowns, not noise.
The discipline: report the two side by side. Flow rising while stock falls is the engagement economy's signature pathology in miniature — revenue this quarter, extracted from the asset.
Proxy for stock trend: (returning rate × recommend-willingness) tracked quarterly, with withdrawal-event rate as the drawdown line.
The Conferral Reproduction Number
Definition: per customer (or member, or user), the number of new customers their conferral eventually produces — referrals made × their close rate, plus attributable review/share influence. Already defined in the Economics Layer; standardized here as the growth metric of the granted layer. R ≥ ~1 with retention = self-sustaining granted growth (paid acquisition optional).
Proxy: "how did you find us" tally ÷ customer count over the period; even a memory-based estimate beats not tracking it.
Counterfeit Discount
Definition (Law 3, operationalized): the performance penalty attached to inauthentic conferral signals — the gap between genuine and manufactured versions of the same tactic (incentivized-undisclosed vs. organic reviews; manufactured vs. real deadlines' long-run conversion; bought vs. earned followers' engagement).
Predicts: where detection improves, CD deepens toward total (counterfeit deflation). Useful in audits as the argument-with-numbers against the fake version of every tactic.
Conferral Concentration
Definition (Law 2, operationalized): the distribution of granted attention across an ecosystem — a Gini-style coefficient over conferral (subscribers, trusted-source status) rather than raw reach. Primarily a research/platform metric; included for the benchmark and the political-economy line of the work.
Two
Measurement playbooks
The this-week version, by what kind of entity you are.
Small business
GS from the last-20-customers tally; R from the same tally; DAR/WPR from the Overdraw Self-Scorer or a manual sequence tag; CP from referred-vs-cold close rates in their own records; TS/EF as repeat rate + review sentiment vs. monthly sales.
Creator
GS = direct/returning consumption ÷ total; TS = list opens + reply rate + member-referral rate, EF = impressions; R = new subs attributable to member shares; CP = revenue per direct-audience member vs. per contested impression. (The Creator Engine's diagnostics, now named and comparable.)
Brand/company
GS by channel mapping; CP from channel-level conversion spreads; TS/EF as retention + NPS-style recommend-willingness vs. traffic; CD wherever both versions of a tactic run.
Platform
GS as chosen-source consumption share; TS as trust surveys + relational-content share, EF as engagement; CC across the creator base. (The Correction paper's program as a dashboard — this playbook is the artifact an advisory engagement would build.)
AI product
covered by <a href="/scorecard">the Conferral Design Scorecard</a>, which imports these metrics (GS of usage arriving by user intent vs. capture; TS as reliance-retention and revocation rates). ---
Three
Rules for this standard’s integrity
A metric that can be gamed will be. These four rules are part of the standard, not a preface to it.
- Definitions are free; never paywall them.The standard spreads by adoption; the business is measurement, benchmarking, interpretation — never access to the terms.
- Proxies are labeled as proxies.The standard's credibility rests on never overclaiming precision; every published number carries its method (measured/surveyed/inferred — the house evidence discipline, applied publicly).
- Version the standard(v1.2 now; changes logged). Metrics that get redefined silently die.
- Resist metric gaming in the definitions themselves:GS counts only genuinely conferred channels (a retargeted ad is not "warm"); incentivized referrals count toward R only if disclosed and would-survive-disclosure (the CD test). The standard must encode the sincerity constraint or it will be Goodharted into the very counterfeit it measures. ---
Four
Where these are already computed
The standard is not aspirational. Every instrument below scores against it.
- The Overdraw Self-Scorer — maps a sequence into deposits and asks, which is DAR and WPR
- The Granted Share Calculator — computes GS, and the Rent Index beneath it
- The Conferral Design Scorecard — imports GS and trust-stock measures for AI products
- A worked report on a coaching practice — GS, R, CP, DAR and WPR, scored
- A worked report on a creator business — the same panel, a different shape
- Conferral by Design — the methodology these metrics instrument
If a number in one of those reports is unfamiliar, it is defined above, with the proxy that produced it. That is the point of publishing this: a score you cannot audit is a score you should not trust.
Appendix
Proposed additions, not yet part of this standard
Published separately so they can be argued with before anything adopts them. One of them already has been.
Initiated Share v1.0 has been retired. It split sessions into prompted and initiated and read the split as evidence about wanted versus captured use. User-configured scheduled tasks defeat that reading: a person authorises future work, the product notifies them when it is ready, and the resulting session counts as prompted even though the prompt is delivering a grant the user made. Both ChatGPT and Gemini document the pattern in their own help material.
The replacement proposal is the session-origin distribution v1.1: eligible sessions reported across eight disclosed origin classes — user-configured scheduled delivery, user-authorised event trigger, expected transactional or safety notice, accepted product suggestion at task entry, product-authored workflow prompt, marketing or re-engagement prompt, unprompted user initiation, and unknown — reported separately, with authorisation, outcome, attribution and missingness fields. No weighted aggregate is proposed until evidence justifies both the construct and the weights.
Withdrawal Rate remains a separate proposal, reported by event type and alongside withdrawal opportunity, control exposure, friction, complaint behaviour, alternatives and re-granting. It is an exit diagnostic, not a direct measure of trust.
Neither is part of this standard, and this appendix does not redefine any metric above it. One note on names that still matters: Granted Share, defined above, is an acquisition measure — where customers came from. The session-origin distribution is a usage measure — how a visit began. A business can have a strong Granted Share and a session mix increasingly produced by its own prompting, and that combination is worth being able to name. The full correction is at the corrections record.
Revisions
v1.2 — 3 September 2026. A definition narrowed. Granted Share previously counted search-with-intent as conferred. That phrase reads two ways and one of them is wrong: a person searching for you by name chose you before they arrived, and a person searching a category or “near me” chose nothing while your listing competed for the click. The first is granted; the second is contested. The definition now reads a search for you by name, and the calculator splits the two into separate rows at its channel table v1.2.
Found by an external audit walking the live instrument. The vocabulary pass of 2 September renamed the labels without re-deriving the mapping underneath, which made a wrong classification read as corrected. Full entry in the corrections log.
v1.1 — 2 September 2026. Vocabulary only. No formula, threshold, proxy or predicted outcome changed.
Two metric definitions used the word owned for a creator’s own channel: the Granted Share proxy (“owned-channel opens”) and the Creator play (“GS = owned/returning consumption”, “revenue per owned-audience member”). They now read direct. Nobody owns people; what can be controlled is a direct, portable route back to them, and a standard that defines metrics in retired vocabulary teaches the retired model to everyone who adopts it.
Also resolved: the calculator had been stamping results “Conferral Metrics Standard v1.2” before a v1.1 existed — this page published v1.0. The corrected definitions are published as v1.1, which makes that stamp true rather than leaving a version nobody could read. Full entry in the corrections log.
v1.0 — 29 August 2026. First publication. Seven metrics, each with a proxy.
Cite as: Conferral Metrics Standard v1.2, Clint Miller, grantedattention.com/standard
Free to adopt with attribution. Use any of it without asking. The definitions are free and stay free — what is sold is measurement, benchmarking and interpretation, never access to the terms.
Every metric above carries its proxy, and every proxy is labelled as a proxy. A standard whose credibility rests on measuring honestly cannot itself overclaim precision — so where a number is inferred rather than measured, the standard requires you to say so, and the reports built on it do.