Run a decision review
An expectation written in advance is only worth something if someone eventually checks it. Left alone, open entries drift: the eval ran but nobody attached it, the deadline passed quietly, the decision got reversed in a meeting and the record never heard. The decision review is the ritual that prevents that — a short recurring pass over everything the workspace said would happen, settled against what did.
Fifteen minutes, every week or two, one person driving with the team in the room. Here's the pass:
- Open the stream, filter to Open
In Ledger, switch the entries filter to Open. Each entry shows its days-open counter — work oldest first. The counter is the agenda: an expectation open for 45 days is either waiting on real evidence or waiting on nobody.
- Settle what the evidence already answers
For each entry, ask: does evidence exist that answers this expectation? If the Caliper eval ran, attach the run under Add evidence and read it against the target. Then Settle — and spend the minute on the lesson, because that's the part the workspace keeps. The lesson is why it landed where it did, and what to do differently because of it.
Settle the misses with the same energy as the hits. A missed expectation with an honest lesson is the entry that saves the next quarter's version of this argument.
- Chase what's answerable but unanswered
Some entries are open because the evidence was never collected — the eval exists but hasn't run, the pilot numbers are in someone's spreadsheet. Assign the chase, right there. If the expectation was freeform ("participation up noticeably by Q4"), decide now what reading will settle it, and who brings it next review.
- Close out the overtaken
Some decisions were reversed or made irrelevant before the evidence ever arrived — the feature was cut, the segment pivoted. Don't leave them open forever and don't pretend they settled: record the reversal as a new entry that supersedes the old one. The original stays in the record with its history intact; the successor explains what changed. See what can change, and what can't.
- End on the Beliefs page
Read the newest lessons aloud — the ones this review just minted. This is the compounding step: the room hears what the workspace now knows, cited to the entries that earned it. If a belief sounds wrong to someone, that's the next decision entry, with its own expectation.
Keeping it honest
- Judge decisions, not deciders. The review reads what the workspace expected against what happened — it is never a scoreboard of who called it. The moment being wrong gets scored against a person, people stop writing checkable expectations, and the ledger degrades into hedged mush. Wrong-but-settled is a good outcome; open-and-forgotten is the only bad one.
- Write expectations the next review can settle. If an entry has sat open for three reviews because nobody can say what would settle it, that's feedback on the expectation — the fix is binding the next one to a Caliper eval or a concrete reading with a date.
- Record the room's decisions during the review itself. The review reliably produces new decisions ("we'll re-run the eval on the bigger dataset"). Capture them as entries with expectations before the meeting ends — the decision door takes a minute, or tell zv1 to draft them.
Run it for a quarter and the character of the beliefs page changes: from a handful of captured lessons to a settled, cited record of what this workspace tried, what it expected, and what turned out to be true.