What Provenance and Receipts Mean · Lesson 5 of 5
Honest refusal: when the record says it cannot say
One concept: honest refusal
Why it matters
Today the Gate Watch board lists 42 funds with a filed repurchase result, and beneath them a shorter table headed “Filings we could not read”. Its first entry is a filing from Bain Capital Private Credit: a SC TO-I/A filed on 2026-08-03, with the words "acceptance unreadable" where a percentage would otherwise sit. Did that fund honour none of its requests? Did the page break? Neither. The row is a refusal: a published decision to show no number, with the reason stated. Reading that row correctly is the whole of this lesson.
The concept
A refusal is a recorded decision to publish no figure. The source was read, a figure the result needed was missing or failed a check it had to pass, and the record says so in the place where the figure would have been. That description names three things a refusal always carries: that the source exists and was read, why no figure came out of it, and when. Those three things separate a refusal from the two states it is most often mistaken for.
The first is zero. A zero is a figure. If a fund’s filing states that it accepted none of the shares tendered, the board shows that as a percentage, read from a sentence in the filing, exactly as it shows any other percentage. A refusal states that no such sentence could be found, or that the sentences found failed an arithmetic cross-check. Treating the refusal as a zero puts a number the filing never stated into your reading of the board, and into any average or share you compute from it.
The second is an error. A page that fails to load carries no reason, no filing date and no actor, and it is expected to clear on the next visit. A refusal is the opposite in every respect: it has a stated reason, it names the document it was read from, and it stays on the page as part of the record, because the decision it records happened.
Kitalpha publishes refusals in two places, and they state slightly different things. On Gate Watch, the stated method is that figures are published only where extraction passes every check, and filings that fail are listed as refused rather than estimated. Each refused row names the fund, the form and the filing date, links the document, and prints the reason. The board’s own note explains why the list exists at all: a fund missing from the board is not the same as a fund with no gate, and silence would make a filing that could not be read look identical to a fund that honoured everything. On a brief, a draft that the automated pre-publication gate refuses is never published, but its edition page is kept. It states the status, lists each lifecycle event with its actor and a full timestamp, and links the corrected edition that passed the gate, where one exists. It does not print the gate’s own wording, because that wording can quote a sentence from a draft that never became public; the status and the actor are the public facts.
What a refusal tells you, then, is real information: the document exists, it was examined, a specific check stopped it, and on a stated date. What it withholds is equally definite: the figure itself, and any judgement about the fund or the week that the figure would have supported. A reader who fills the blank with a guess loses the first and invents the second. The table below places the two refusals from today’s records side by side; look for what each row still states even though neither carries a number.
| Item | Detail |
|---|---|
| Refused filing | Bain Capital Private Credit — SC TO-I/A filed 2026-08-03 |
| Reason the board states | acceptance unreadable |
| Figure the board shows | None: the row carries the reason in the place of a number |
| Refused edition | wpsr-2026-08-21 — status refused (not_published) |
| What replaced it | Successor edition wpsr-2026-08-21-3, published in its place |
| Events on its record | 2 dated lifecycle events, each with a named actor |
Worked example
Take the refused filing on the board today, a SC TO-I/A from Bain Capital Private Credit filed on 2026-08-03, and the board's own counts beside it. The board reports 42 tracked funds, meaning funds with a machine-readable result, of which 15 are shown as prorated. The refused list is a separate count. The steps below read both counts, size the refused list against the board, compute the prorated share correctly, and then compute it wrongly, the way the misconception would.
Record: Bain Capital Private Credit — refused (acceptance_unreadable) · as of · SC TO-I/A filed 2026-08-03
- Filings listed under 'Filings we could not read' on the board today 21
- Funds tracked on the board, each with a machine-readable result 42
- Refused filings for every 100 tracked funds 50.0% The refused list sits beside the board rather than inside it, so this compares the size of the two lists; it is a ratio between them, not a share of one total.
- Tracked funds shown as prorated, as a share of tracked funds 35.7% Every fund in the numerator and in the denominator has a filed figure. A refused filing sits in neither.
- The same share if every refused filing were wrongly counted as a prorated fund 57.1% This is the arithmetic the misconception performs: a blank read as a figure. The gap from the line above is the distortion the refusal prevents.
Read the last two lines together. The correct prorated share divides one count of funds with figures by another. The final line adds every refused filing to both the numerator and the denominator, as if each were a fund that honoured less than every request; the gap between the two lines is the distortion that reading a blank as zero introduces. Note also what the steps could not do: nothing here recovers the refused fund’s own percentage, because the filing did not state one that passed the checks. The refusal stays a refusal after the arithmetic.
Faded example
Now the second record, a brief rather than a filing: the WPSR edition wpsr-2026-08-21, whose status reads "refused" and whose successor is wpsr-2026-08-21-3. The first two steps read the record’s text fields for you. Complete the last one from the lifecycle list on the page, then reveal the answer and compare.
Second record: wpsr-2026-08-21 (refused) · as of
- Public brief text this edition carries, in paragraphs (the record reads published: false)0
- Later editions the record points to (one successor slug is stated)1
- events Tolerance ±0 events
Reveal the answer and the explanation
2 — Each row of the lifecycle list is one recorded event with a named actor and a full timestamp. The edition carries no public text, yet its record carries these events and names the edition that replaced it. That is the shape of an honest refusal: the number is absent, and the account of why it is absent is present.
Stored on this device only; not graded.
Retrieval check
Mark your confidence before each answer. Every option carries an explanation, and the ones you rejected are worth reading too.
-
1. A fund appears on the Gate Watch board under 'Filings we could not read', with a form, a filing date and a stated reason in the place of a percentage. The record is telling you that:
Choose your confidence first. -
2. Using the Gate Watch records, how many refused filings are listed for every 100 tracked funds? Divide the refused count by the tracked count, multiply by 100, and enter the result to one decimal place.
Source record: Bain Capital Private Credit — refused (acceptance_unreadable) (as of 2026-08-03)
Tolerance ±0.1 %Choose your confidence first. -
3. The record for the WPSR edition wpsr-2026-08-21 reads published: false, gives a status and a reason code, names a successor slug and lists dated lifecycle events with actors. Classify what this edition is:
Choose your confidence first. -
4. A reader says a refusal on Gate Watch shows the system failed, because the board exists to publish figures. Which reading matches the record?
Choose your confidence first.