Free to copy · No attribution required
The Receipt Test
Six questions that tell you whether a system gives you a real evidence record, or just a process log.
Ask them of any vendor, any tool, or any internal process that helps produce a number someone will have to defend. They take five minutes, and no demonstration is required. That is the point.
The six questions
- 1
Can you export one complete evidence record?
A real yes produces the whole record as an export. Not a summary assembled later, and not a screenshot of the parts someone chose to show.
- 2
Can someone outside the vendor verify it without a vendor demo?
A real yes lets a third party locate the cited sources and repeat the check without special access, vendor personnel, or a demonstration.
- 3
Does it identify every source record and the date that source describes?
A real yes identifies each source well enough to find it, and separates the date the source describes from the date the system observed it.
- 4
Does it preserve conflicts, single-source claims, and missing coverage?
A real yes does not turn disagreement into false certainty. It keeps conflicting values, marks one-source evidence as one-source evidence, and says where coverage ends.
- 5
Can your team still inspect and use it after the vendor leaves?
A real yes gives you the exported record and the right to retain, inspect and use it outside the supplier interface, including after termination.
- 6
Does the record cite sources outside the vendor’s own system?
A real yes points at records the vendor does not own or control. A record whose only witness is the vendor’s own software can be perfectly tamper-evident and still tell you nothing about whether outside sources agree.
If the answers are vague, you may have a process log. You do not yet have a receipt.
A process log tells you that work happened. An evidence record shows what the answer rested on, and lets someone else check it.
What this test does not ask
It never asks whether the answer is true. No evidence record can settle that, and a vendor claiming otherwise is claiming something it cannot support. The test asks whether a record lets you go and look: whether the sources are named, dated, reachable, and reported honestly when they disagree or run out. What the sources mean, and whether to believe them, stays with the reader.
When two filings agree, that is what our record says. It does not say the world is that way. Two filings can agree and both be wrong, and that is exactly why the record hands you the filings instead of a verdict.
Put it in the paperwork
The questions work in a meeting. This works in an evaluation. Copy it into an RFP or a vendor questionnaire.
This is suggested requirement language, not reviewed contract language. It has not been checked against a vendor agreement by procurement counsel. Take it to your own procurement or legal team and adapt it. We publish it because the questions it forces are the point, not the drafting.
Evidence records. The supplier shall deliver, with its response, at least one complete evidence record produced by or with its system. The record shall:
(a) be exportable by the client in a documented, machine-readable format;
(b) identify each source record relied upon, with a reference sufficient to locate that source outside the supplier systems, including at least one source not generated or controlled by the supplier where the record makes any claim about facts outside the supplier’s own system activity;
(c) state, for each source, the date or period that source describes, and the date the supplier system observed it;
(d) state the verification status of each material value, including any conflict between sources and any value supported by only one source;
(e) state its known coverage limits, what the record does not cover, rather than omitting them; and
(f) be evaluable by the client or a third party using only the exported record and publicly available sources, without supplier systems, personnel, or demonstrations.
The agreement shall grant the client the right to retain, inspect, and use delivered evidence records outside the supplier interface, including after termination.
A response that offers any of the above on request, under NDA, in a demonstration, or as screenshots, rather than as the exported record itself, shall be deemed non-responsive to this requirement.
What a record actually says
One number, from one loan, in one deal. This is the whole thing in plain language.
The number: $69,000,000. The ending balance on loan 1 of Wells Fargo Commercial Mortgage Trust 2020-C58, as reported for the July 2026 filing period.
Who says so: two separate filings with the SEC, made by different parties for different purposes.
- The loan tape filed with the deal reports $69,000,000.
- The trustee’s monthly remittance report reports $69,000,000.
What the record concludes: the two agree, so it is marked confirmed. Had they disagreed, the record would say so and show both numbers rather than quietly picking one. That happens, and when it does we publish it.
What it admits: our dating is currently accurate to the filing period, not to the day. The record states that inside itself, not in a footnote somewhere else.
Both filings are linked from the record, so checking us means opening two documents on sec.gov and reading the same number. It takes about a minute and does not involve us.
The underlying record is machine-readable and will look technical. That is deliberate: it is written to be read by other software as well as by people. The plain-language version above says everything it says.
We take the test in public
A test we could not pass would not be worth publishing. Everything below is checkable right now, without an account and without talking to us.
1. Export one complete record
Yes. Every verification returns the whole record as JSON. The export is the response, not a report about it.
2. Verify without a demo
Yes. Each record cites the primary SEC filings by accession number with direct sec.gov links. Anyone can open those filings and check the value without touching us.
3. Sources and their dates
Yes. Each citation names the source, its accession, the period the source describes, and when we observed it. Where our dating is period-granular rather than exact, the record says so in the record.
4. Conflicts and coverage
Yes. Conflicting values come back as a conflict with both citations, never a silently chosen winner. One-source values are labelled and kept out of the agreement rate. Bond coverage is published as zero rather than omitted.
5. Keeps working without us
Yes. The record is plain JSON in your possession, the format documents are public files you can keep beside it, and the sources live in SEC EDGAR, which does not depend on us existing.
6. Sources we do not control
Yes. Every citation is a filing made by someone else and published by the SEC. Nothing in a receipt rests on a record we produced. We could stop existing tomorrow and the sources would still be there for anyone to open.
The verification service is free and needs no account. Check a receipt yourself.
What we do not claim: CMD+RVL holds no SOC 2 attestation today; it is a stated target for Q3 2026. Nothing on this page asserts what a regulator or examiner will ask. That would be an inference presented as a citation, and this page is about refusing exactly that move.
Three ways to use it
Evaluating something
Ask the five questions in the meeting. Then ask for one exported record rather than a walkthrough. What arrives, or does not, is the answer.
Writing the requirements
Put the clause in the RFP or the vendor questionnaire. Every supplier then answers it in writing, before anyone signs, and you can compare the answers side by side.
Already bought something
Run it on what you have. The answers are worth knowing either way, and a gap is cheaper to find now than when someone is asking you to produce the record.
If you advise other people on this kind of purchase, take the whole page. It is meant to be handed to a client with your name on it, not ours.