The Blank Cells Where the Decision Was
Open the electoral archive for a district that one party has locked up for decades and you'll find full turnout tables, precinct maps, and margin breakdowns for the general election, sitting next to a primary result recorded as a single line, sometimes not recorded at all beyond a winner's name. But the primary is the whole election. The general is a formality with a scoreboard attached. The archive keeps the scoreboard and lets the actual decision evaporate, because the scoreboard is the part that produces numbers you can put in a table.
This isn't a quirk of one-party districts. It's what formal record-keeping does by default, everywhere, because records get built around whichever stage of a process happens to output something countable, not whichever stage actually moved the outcome.
A hiring committee is the same shape at smaller scale. The scorecard gets filled in after the interview, numbers on a rubric, comments in a shared doc, a clean artifact that HR can point to if anyone asks how the decision got made. But anyone who's sat in the room knows the scorecard is downstream. The decision usually happens in the first five minutes, or in the hallway conversation before the candidate even arrives, or in the resume screen that decided who got interviewed at all. The rubric gets filled out to justify a conclusion that was already reached by a process that left no form to fill in. Six months later, if someone audits the hire, they'll find a rubric. They will not find the hallway.
Code review runs the identical trick. A pull request gets an "approved" click, a timestamp, maybe a comment thread with a few nits resolved. That's the artifact that survives, the thing a postmortem points to when something breaks: approved by so-and-so, on this date, here's the diff. But on most teams the actual call, ship this or don't, got made in a message somewhere before the PR existed in reviewable form, or in a conversation where someone said "yeah, that approach is fine, just open it" and the review that followed was theater performed for the archive. The click is real. It's just not where the thinking happened.
Grant review boards, tenure committees, insurance claims, court dockets: same pattern, different paperwork. Whatever stage of a process happens to terminate in a form, a vote count, a signed approval, a scored rubric, that's the stage that gets archived, cited, and treated as the record of what happened. The stages before it, the ones that did the actual work of narrowing the field, setting the terms, or deciding the winner before anyone counted anything, leave no comparable trace, because nobody designed a form for them. They weren't skipped. They just didn't happen anywhere with a column for it.
Call it the countable-step problem. It's not that records lie, exactly. Every number in that general-election table is accurate. Every approval timestamp on that pull request is real. The distortion is structural, not dishonest: an archive is a record of whichever step in a process happens to produce a clean tally, and a clean tally is a property of the step's mechanics, not a property of how much the step actually decided. A primary with forty thousand voters in a safe district decides more than a general with four hundred thousand, and the record will always be sized to the four hundred thousand, because that's the stage built to be counted.
The fix people usually reach for is "record more of the process," minutes from the hallway conversation, a paper trail on the Slack thread, primary turnout logged with the same rigor as the general. That helps at the margins. But it doesn't touch the underlying reason the gap exists, which is that formal steps get built for accountability and informal steps get built for speed, and the two rarely coincide on the same moment. You can log more. You can't make the fast, undocumented moment want to be logged.
So the honest version of an audit isn't "what does the record show." It's "which stage of this process was actually built to decide something, and did anyone leave a trace of it." Most of the time the answer is no, and the record you're holding is just the formality that happened to come with a form.