Decisions need records, not just oral agreement

What it is

A decision reached orally in a meeting is, until written down, only "the attendees' memory": absentees don't know it, systems don't execute it, and weeks later even the attendees remember it differently. A decision record is the artifact written afterwards — conclusion plus rationale, retrievable by absentees and future selves: what was decided, why, what was discarded, who owns it, by when. A decision is not finished when spoken; it is finished when recorded. An orally-approved decision is in an unclosed state.

Why it happens

Oral decisions leak in three directions at once. Memory: collective recall loses the "why" faster than the "what" — weeks later people remember the rough conclusion but not the reasons a contender was rejected, so when its supporters raise it again the team has no ammunition to stop the entire argument being re-fought; re-litigation is the canonical cost of unrecorded decisions. Execution: oral consensus has no executable anchor — each person heard a slightly different version (especially those who withheld reservations), and only at merge time does it surface that agreement never truly existed. Accountability: oral commitments cannot be verified — "we agreed you'd do it" versus "I said I'd look into it first" is unadjudicable, so disputes resolve by volume or seniority. Writing closes all three leaks at once: it forces fuzzy consensus into a statable sentence (failure to write it is precisely how fake consensus gets exposed) and leaves a citable anchor for absentees and the future.

Studying it

Meeting science and organizational behavior offer two paths. One is action-item tracking: comparing recorded items against subsequent behavior, where documented action items show markedly higher completion rates than verbally assigned ones — a relationship observed repeatedly in field settings. The other is decision auditing: tracing decisions of one class (e.g., product direction changes) back to how completely they were recorded, then comparing well-recorded versus poorly-recorded groups on re-litigation counts and execution drift months later. Methodological caveat: teams that record well usually have mature processes anyway — "recording" and "executing well" may both be expressions of the same underlying variable, so the correlation does not read causally; intervention experiments (randomly requiring some meetings to record on the spot) come closer to causality but cannot blind participants, so Hawthorne effects must be dodged by using objective execution metrics rather than satisfaction ratings.

Where it stops holding

Recording granularity should flex with reversibility: irreversible, consequential decisions deserve full records (conclusion, rationale, rejected options); matters digested the same day incur pure bureaucratic drag if each is logged. Nor does a record substitute for consensus — writing down an unsettled item as "decided" weaponizes the record and destroys the log's credibility. Finally, audience determines the writing: records for the team can omit background; records for absentees and newcomers must be self-contained.

Applying it

• Assign a rotating scribe to every decision meeting, tasked with writing each decision within the meeting or an hour after, in a fixed structure: conclusion, rationale, rejected options, owner, deadline.
• Park decision records in one fixed, searchable place (decision log, document repository), and reference links from related tickets and code so executors never need to scroll meeting minutes.
• Spend two minutes at the next meeting reviewing the previous decisions and correcting them in place — an error-correction window while memory hasn't fully decayed.
• Unwritable means unreached: any "decided" that cannot be written as a complete sentence goes back for discussion rather than being recorded in bad faith.
• Verification: three months on, spot-check whether the record alone answers "why was option B rejected" and "who owns this"; count how often the same topic returned to the table — topics re-litigated more than once mark exactly where recording failed.

Related

• Same group: V5.09.1 Content that can be asynchronous should not occupy synchronous time · V5.09.3 Meeting cost grows linearly with headcount
• Nearby: V5.07.4 Whiteboard output must be distilled into conclusions or it dies with the session · V5.08.3 Searchable text records are the long-term value · V4.05 Handoff and Context Transfer
• Search terms: decision record · re-litigation · meeting minutes · action item follow-through