Skip to Content

Published Standard · Version 1.0

The FounderGrade Verification Standard

Before a programme trusts a verification, it needs to know exactly what verification means. This is that document. It is versioned, it is public, and you are welcome to send it to your board, your council or your own funders.

Published pre-launch. This version governs pilot engagements. It has not yet been applied to a completed production engagement.

Pursuing DFSA Innovation Testing Licence · DIFC, Dubai

Section 1

Evidence Taxonomy

Five categories of evidence. Each one can demonstrate something specific and cannot demonstrate something else. A verifier states which categories were relied on for every finding.

Primary Artefact

Source repository at a named commit, a deployed address or endpoint, a signed document, a published paper.

Can Demonstrate
That a specific object exists, in a specific state, at a specific time.
Cannot Demonstrate
That the object is fit for purpose, secure, adopted, or produced by the party claiming it, without corroboration.

System of Record Export

Accounting exports, payroll records, bank statements, payment processor reports, contract management exports.

Can Demonstrate
What a system the party controls says happened, with dates and amounts.
Cannot Demonstrate
That the underlying transactions are genuine. Verification is not an audit and does not test the system's controls.

Third-Party Confirmation

A written confirmation from a counterparty, customer, employer or partner, obtained directly by the verifier.

Can Demonstrate
That an independent party, contactable and identified, agrees the stated fact occurred.
Cannot Demonstrate
Anything at all, when the confirmation is relayed by the assessed party rather than obtained directly.

Independently Observable Fact

A public deployment, a public filing, a company register entry, a live product the verifier can use.

Can Demonstrate
A fact the verifier confirmed personally, without relying on the assessed party at all. The strongest category.
Cannot Demonstrate
Intent, quality of the underlying work, or anything that happened before the observation window.

Self-Reported Statement

Progress notes, decks, narrative updates, screenshots supplied by the assessed party.

Can Demonstrate
What the assessed party claims. It defines the claim under test.
Cannot Demonstrate
That the claim is true. A self-report is never sufficient on its own to corroborate a milestone.

Section 2

Thresholds by Milestone Type

What is sufficient to corroborate each type of deliverable, and what is not. A milestone that cannot be mapped to one of these types is defined explicitly at scoping, in writing, before work starts.

Code Deliverable

Sufficient to Corroborate

  • Repository access at a named commit or tag, dated on or before the milestone deadline
  • The verifier builds or runs the stated functionality, or reviews it against the written milestone definition
  • Commit history consistent with the claimed authorship and timeline

Not Sufficient

Screenshots, a demo video, or a description of the code without access to the code itself.

Deployment

Sufficient to Corroborate

  • A public address or endpoint the verifier resolves independently
  • Deployment metadata showing the deployed artefact matches the reviewed source
  • The verifier exercises the deployed functionality named in the milestone

Not Sufficient

A transaction hash or URL supplied without the verifier confirming what is actually running there.

Revenue Claim

Sufficient to Corroborate

  • Payment processor or bank exports covering the full claimed period
  • At least one independent corroboration: a signed customer contract, an invoice matched to a settled payment, or direct customer confirmation
  • Reconciliation of the exported total against the figure claimed

Not Sufficient

A dashboard screenshot, a spreadsheet, or a founder-stated figure. Note that this is reconciliation, not an audit of financial statements.

Hiring Claim

Sufficient to Corroborate

  • A signed employment or contractor agreement with a start date
  • Payroll or payment evidence for at least one period, or direct confirmation from the named person

Not Sufficient

A profile page, an announcement post, or an offer letter that was never accepted.

Partnership Claim

Sufficient to Corroborate

  • An executed agreement, or written confirmation obtained by the verifier directly from the named counterparty
  • Evidence of activity under the agreement where the milestone claims activity rather than signature

Not Sufficient

A logo on a website, a joint announcement, or a forwarded email chain.

Research Output

Sufficient to Corroborate

  • The published output, with a resolvable identifier or public location
  • Evidence the output meets the scope written into the milestone definition
  • Where peer review is claimed, confirmation of the review status from the venue

Not Sufficient

A preprint described as peer reviewed, or an output whose scope was never defined at milestone definition.

Documentation Deliverable

Sufficient to Corroborate

  • The published documentation at a resolvable location, dated
  • The verifier checks coverage against the specific sections named in the milestone definition

Not Sufficient

A word count, a table of contents, or a private draft that is not accessible to the intended audience.

A Standard That Can Change Silently Is Worth Nothing.

Every change to this standard is published here with a version number and a date before it takes effect. An engagement is governed by the version in force on the day the milestone was defined, so a standard cannot change underneath work already in progress.

Section 3

The Three Outcomes

There are exactly three. There is no partial pass, no percentage and no score on a verification, because a milestone was either evidenced to the published threshold or it was not.

Corroborated

It Asserts
The evidence supplied meets the threshold published for this milestone type, and the verifier found nothing contradicting the claim.
It Does Not Assert
That the work is good, that the project will succeed, or that anything outside the milestone definition is true.
What Follows
The programme has a supported basis to authorise the next tranche. The decision remains the programme's.

Contradicted

It Asserts
The verifier found specific evidence inconsistent with the claim, and states what it is and why it is inconsistent.
It Does Not Assert
Fraud, bad faith or intent. Those are findings only a competent authority can make.
What Follows
The programme decides: withhold, request correction, or open a resubmission window. Appeals are available.

Unable To Verify

It Asserts
The evidence available did not reach the published threshold, and the verifier records exactly what was missing.
It Does Not Assert
That the claim is false, and it is not an accusation. It very often means a document was not supplied or a counterparty did not respond in time.
What Follows
The default path is a resubmission window, not a rejection. Programmes are asked to treat it that way.

What Verification Does Not Claim

This section matters more than the rest. A verification business that overstates what it can establish is worth less than no verification at all, because a funder would then rely on it.

  • Verification does not certify future success. A corroborated milestone says a past deliverable was evidenced, nothing more.
  • Verification is not a financial audit. We reconcile stated figures against records supplied. We do not test internal controls, do not express an audit opinion and are not auditors.
  • Verification does not replace legal, tax or regulatory compliance, and it is not a compliance attestation of any kind.
  • Unable To Verify is not an accusation and must not be reported as one.
  • A finding is not a release instruction. The programme authorises every payment.
  • Verification does not assess the quality, security or market prospects of the work, unless the milestone definition itself made that the deliverable and named how it would be judged.

Section 5

Who Verifies, and What Disqualifies Them

Independence is the whole product. A verifier who is conflicted produces a finding that is worth nothing, so the disqualifying conditions are published rather than described in general terms.

Who Verifies

A calibrated expert drawn from the verification pool, matched by sector and by the milestone type under review. Verifiers are named to the programme and to the grantee, and their calibration status is on the record.

Independent of the Grantee

A verifier is disqualified where they hold or have held equity, tokens, debt or an advisory position in the grantee, have worked for or with the grantee inside the last 24 months, are related to a founder or officer, or have an open commercial discussion with the grantee.

Independent of the Programme

A verifier is disqualified where they sit on the programme's selection committee, council or board, where they authored the grant application, or where their compensation varies with the outcome of the finding. Verifier pay is fixed by tier and is identical whatever the finding is.

Blind Where Blindness Is Possible

Assessment review is blind to grantee identity where the evidence permits it. Milestone verification frequently cannot be blind, because confirming a deployment or a counterparty requires identity. Where blindness is not possible we say so on the finding rather than pretend otherwise.

Declaration and Recusal

Every verifier signs a conflict declaration per engagement. Any conflict discovered mid-engagement triggers immediate recusal, the finding is voided and the milestone is reassigned at our cost.

How Disagreement Is Resolved

  1. 1Milestones above the programme's configured value threshold are verified by two independent experts who do not see each other's work.
  2. 2Where both reach the same outcome, that is the finding.
  3. 3Where they differ, a third verifier at a higher tier reviews the same evidence plus both written reasons, and their outcome stands as the finding.
  4. 4Both original findings remain attached to the record. A disagreement is never deleted, because the disagreement is itself information the programme should see.

Who funds the tie break: The tie break is funded from a reserve held back from the engagement fee, not billed to the programme and never billed to the grantee. Neither party pays more because experts disagreed, so nobody has a financial reason to avoid a second opinion.

Section 7

Appeals

The verified party may appeal any finding. So may the programme.

Notice

Within 10 business days of the finding being issued

The appellant files the specific ground: a factual error, evidence that was supplied but not considered, a threshold applied incorrectly, or an undeclared conflict.

Acknowledgement

2 business days

We acknowledge, confirm the ground is within scope and freeze any pending release the programme has linked to the finding.

Review

10 business days

A verifier who took no part in the original engagement reviews the original evidence, the written reasons and the appeal ground.

Outcome

Issued at the end of review

Upheld, amended or overturned, with written reasons. An overturned finding is marked as overturned on the permanent record and is never erased.

Escalation

Within 10 business days of the outcome

Where a conflict was found, or the ground was procedural, the matter escalates to the platform review panel and the panel's decision is published in redacted form.

Disagreeing with a correctly reached finding is not a ground of appeal. Neither is a preference for a different verifier.

The Record and Attestation Model

Attestation-based record keeping is an established pattern in grant tooling, so programme operators reasonably ask how a verification is anchored. Here is exactly what is recorded, what is public, what never leaves the platform, and how a third party checks a finding without asking us.

That a verification engagement was opened

Anchored · Public

A milestone identifier, the programme identifier and a timestamp. No evidence and no narrative.

The finding: Corroborated, Contradicted or Unable To Verify

Anchored · Public

The outcome, the standard version applied, the verifier tier and a timestamp.

A cryptographic digest of the written reasons

Anchored · Public

A hash only. Anyone holding the reasons document can confirm it is the one that was recorded, and nobody can read it from the hash.

The written reasons themselves

Off Chain · Programme and Parties

Held in the platform record, released to the programme and the verified party, and exportable by both.

The evidence the grantee submitted

Off Chain · Parties Only

Never anchored, never published. Confidential material stays confidential by design, not by policy alone.

Verifier identity

Off Chain · Programme and Parties

Disclosed to the programme and the grantee for conflict checking. The public record carries the tier and calibration status, not a name.

Appeals and overturned findings

Anchored · Public

An overturn is recorded as a new entry referencing the original. Nothing is deleted or rewritten.

How a Third Party Confirms a Verification Happened

  1. 1Take the milestone identifier from the programme's own record or from the finding document.
  2. 2Resolve it against the published record to see whether an engagement was opened and what the outcome was.
  3. 3Hash the reasons document you were given and compare it to the digest on the record. A match proves the document is the one that was recorded and has not been altered.
  4. 4Where the programme publishes a transparency page, the same identifiers appear there, so a community member can perform the same check without our help.

The Network Is Not Yet Chosen, and We Will Not Pretend Otherwise

  • The specific EVM-compatible network has not been selected. We will not name one before the decision is made, because naming a network we have not committed to would be misleading.
  • No contracts are deployed today. The platform is pre-launch and nothing described here is running in production.
  • Contract addresses, the network choice and the record schema will be published at launch, on this page, before any production record is written.
  • The record model is deliberately portable: identifiers, outcomes and digests, with no dependency on any single network's features. If the network choice changes, historic digests can be re-anchored on the new network with the original timestamps retained off chain and the migration published.
  • Anchoring is a durability and tamper-evidence mechanism. It is not what makes a finding credible. The verifier's independence and the published standard are what make a finding credible.

The rules a finding is produced under are published in the verification standard, and the division of labour between tooling and people is published in the methodology.

Version History

Every change to this standard is published here with a version number and a date before it takes effect. An engagement is governed by the version in force on the day the milestone was defined, so a standard cannot change underneath work already in progress.

  • Version 1.0

    31 August 2026

    First published standard: evidence taxonomy, thresholds for seven milestone types, three defined outcomes, independence and conflict rules, tie-break funding, and the appeals timetable.

Questions on the standard, or a correction to it, go to hello@foundergrade.ai. Related reading: the methodology, the rubric and data handling for programmes.

For Programmes

Test the Standard on One Cohort.

A pilot applies this document to milestones you already fund, and tells you where your own definitions need tightening.

Pursuing DFSA Innovation Testing Licence · DIFC, Dubai