Commercial terms, version 1.0

VALUE ATTRIBUTION RULE

Version 1.0. Published because a fee a customer cannot reproduce is a fee a customer will eventually dispute.

WHY THIS IS PUBLISHED RATHER THAN HELD

EOS is priced as a share of delivered value. That makes attribution a commercial term, not a finance feature, and it makes one question load-bearing: when something improves, how much of the improvement is ours?

Every vendor in a value-share arrangement faces the same temptation, and most of them lose year two to it. A vendor who claims full credit for an ambiguous outcome wins the first invoice and loses the renewal, because the customer's finance team eventually does the arithmetic and finds it generous in one direction only.

So the rule is written down, published, conservative by construction, and reproducible by the customer without our involvement. If we cannot show our work, we do not invoice.

THE RULE IN ONE SENTENCE

The attributable share is the lower bound of what the evidence supports, computed by the method below, from a pinned evidence pack, against a baseline captured before the intervention.

Four constraints, each doing real work:

Lower bound, not best estimate. Where a range is defensible, we bill the bottom of it. Pinned pack. A pack that re-derives against live data cannot support a measurement, enforced in code rather than in a policy. Baseline before. A baseline computed after the fact is a negotiation, not a measurement. Reproducible. Every input is in the pack the customer holds.

THE SIX GATES

A finding must pass all six before any value is attributed. A finding that fails any one is worth recording and is worth nothing on an invoice.

Gate 1: The finding was surfaced by EOS

The discrepancy was raised by the platform, not already open in the customer's own queue. A finding the customer had already logged is not ours to bill, and the exception loop records which is which at the moment it fires.

Gate 2: The baseline predates the intervention

Captured before any action was taken, from source records, and pinned. If no baseline exists from before, the attributable share is zero. Not estimated, not negotiated. Zero.

Gate 3: The evidence pack is PINNED and intact

The pack carries the source values it was built from, the rule version and the calculation version, and its integrity digest verifies. A rederived pack is ineligible.

Gate 4: The outcome is measured at settlement

Not at recording, not at decision. Value is what actually settled: the invoice that was corrected, the terms that changed, the stock that was found. A predicted value is not a measured one.

Gate 5: The customer has confirmed the outcome

A named person on the customer side confirms the measured figure. Our measurement alone never supports an invoice. This gate exists because the alternative is asking a customer to accept a number produced entirely by the party being paid.

Gate 6: Confounders have been assessed and deducted

See the next section. This is the gate that does the most work and the one most vendors skip.

CONFOUNDER DEDUCTION

An outcome rarely has one cause. The method is to name the plausible alternative causes, deduct the share attributable to each, and bill what remains.

The deduction rule, and it is deliberately unfavourable to us:

Situation	Attributable share
No plausible alternative cause identified	100%
One plausible alternative cause, not quantifiable	50%
One plausible alternative cause, quantifiable	100% minus its quantified share
Two or more plausible alternative causes, not quantifiable	0%
The customer names an alternative cause we did not	Their assessment governs

Note the asymmetry, which is intentional. An unquantifiable confounder costs us half. Two cost us everything. A customer-identified confounder always wins over ours. In every ambiguous case the rule resolves against the vendor, because the vendor is the party with the incentive and the customer is the party who has to trust the result.

Worked example. A price variance of $480,000 is surfaced, traced and corrected.

Measured settlement value                          $480,000
Confounder A: a contract renegotiation was already
  underway, quantified at 30% of the correction    -$144,000
Confounder B: a commodity index moved in the same
  period, not quantifiable at line level           -50% of remainder
                                                   ----------
Attributable value                                 $168,000
Fee at the applicable band                         computed on $168,000, not on $480,000

The customer can reproduce every line from the pack. So can their auditor.

WHAT IS NEVER ATTRIBUTED

Value from a finding the customer had already identified.
Value from a decision the customer would demonstrably have made anyway, where they say so.
Avoided cost with no settlement event. A discrepancy found and not acted on produces no value,

however real it was.

Compounding or run-rate projections. Only the measured period is billed. A saving that

genuinely recurs will be measured again next period, on its own evidence.

Value in a period where the pack cannot be reconstructed. If the evidence has degraded, the

fee goes with it.

THE DISPUTE PATH

The customer challenges a measurement, in writing, with their reasoning.
We reproduce the calculation from the pinned pack and show every input.
If their reasoning identifies a confounder we did not deduct, the deduction is applied without argument and the invoice is corrected.
If we disagree on a quantification, the customer's figure governs for that period and the question is escalated for the next one.
Nothing is billed while a dispute is open.

Step 3 and step 4 are unusual and they are the point. A dispute path where the vendor adjudicates its own fee is not a dispute path. Conceding the ambiguous cases costs less than one renewal, and it is the only version of this rule a finance team will sign.

THE COMMITMENT

We will not invoice value we cannot show. We will not bill a period whose evidence has degraded. We will deduct a confounder a customer raises. And where the method leaves a range, we will bill the bottom of it.

That is a worse rule for us in every individual case and a better one across a relationship, which is the trade a value-share model exists to make.

Web3 Ventures Enterprise. This rule forms part of the commercial terms and is published so it can be read before it is relied on.