What we can prove. What we cannot. In the same breath.
A verification company that hides its limitations in the fine print has misunderstood its own product. This page is the fine print, promoted to a first-class page.
Evidence philosophy
IndeProof's founding rule: never certify success without sufficient independent evidence. Everything else follows from it.
- Independence is structural, not procedural. The verifier holds its own identity and its own least-privilege, read-only access. It never shares credentials with the system whose claims it checks. Independence you have to remember to maintain is not independence.
- Contracts freeze before counting. The Outcome Contract — outcome, proof source, matching rule, window, settlement, classification — is fixed before the first counted event. A rule that can bend after the fact proves nothing.
- Classification is deterministic. The same evidence always yields the same classification. There is no judgment call inside the window.
- Everything fails closed. Missing, late, contradictory, malformed, or duplicated proof never rounds up to success.
- Evidence outlives opinion. Every classification — including uncertain and excluded — is recorded as a sequenced, integrity-protected evidence artifact whose tampering is detectable on verification. If later evidence contradicts a classification, a linked superseding artifact is created — the original is never rewritten.
How uncertainty is handled
Uncertain is an outcome classification, not an apology. When the independent source cannot be observed — an evidence condition recorded as unavailable — or shows two records where the contract requires exactly one, the outcome is classified Uncertain and the ambiguity is preserved exactly as found. It is never resolved optimistically, never retried into success, and never quietly dropped from the totals.
Why exclusions are not failures
Some events are legitimately out of scope — but only when they say so themselves. The sole exclusion rule is an explicit internal-test flag set at dispatch. There is no discretionary exclusion after the fact, and every exclusion is itself recorded in the evidence chain. An exclusion you can audit is a boundary; one you cannot is a loophole.
Why a second source matters
Monitoring tells you what a system says about itself. Logs are written by the process being judged. Self-reported completion is the claim restated with more confidence. None of these can contradict the system, because they are the system. Verification requires an instrument that could disagree — a source the claiming system does not control, read by an identity the claiming system does not hold.
Current limitations
- Verification is per contract. A verdict speaks only to the specific outcome, source, rule, and window that were frozen — nothing generalizes automatically to other outcomes or systems.
- Evidence to date comes from a frozen internal pilot protocol on controlled data in an isolated test environment. There are no customer results, and none are claimed.
- A passing controlled run does not establish natural failure frequency, production security, customer value, willingness to pay, or generalization beyond its contract.
- IndeProof requires that a genuinely independent proof source exists and is readable. Where no such source exists, IndeProof honestly does not apply — and qualification will say so.
- IndeProof is not generic AI truth-checking, not monitoring or alerting, and not universal cross-system verification. It cannot prevent failures; it proves whether contracted outcomes occurred.
What you will never find here
No borrowed badges. No unearned certifications. No testimonials. No fabricated or extrapolated metrics. Every example on this site is synthetic or controlled and labeled as such — because the clear separation between controlled examples and real customer outcomes is itself part of the methodology.