Every product team has a meeting where a settled question rises from the dead: why is onboarding three steps? Why did we drop annual billing? The tracker records what was decided; the reasoning lived in a call nobody can replay. So the team re-litigates—with worse attendance and a colder trail of data than the first time.
Trackers store outcomes, not arguments
Specs and tickets are final artifacts, scrubbed of the debate that produced them. The user quote that swung the room, the constraint that eliminated an option, the objection someone raised and got overruled on—that context evaporates with the meeting. A new teammate reading the ticket learns the what and inherits none of the why.
Record the decision minute
Twenty seconds of voice right after the discussion: what was decided, which alternative lost and to what argument, who dissented and on what grounds. Dissent especially—today's overruled objection is next quarter's early warning.
- The decision, the losing option, and the argument that settled it.
- The evidence behind it: attach the metrics screenshot, the whiteboard photo.
- The dissent and its reason, by name.
Link the reversal to the original
Decisions get reversed; that is health, not failure. What hurts is reversing them without noticing—or re-deciding a question because nobody remembered it was settled. When a decision changes, record the new fact that changed it. Chronology keeps the current state on top while preserving the trail, and the trail is what distinguishes ‘we learned something’ from ‘we forgot why.’
Answer ‘why is it like this?’ with sources
When the question arrives—from an executive, a support escalation, an engineer reading old code—ask the history instead of your memory. The answer comes back with dates, the original discussion note, and the attached evidence, with assumptions labeled as assumptions. ‘We chose this in April over X because of Y, and it survived the June review’ closes conversations that a vibes-based recollection would reopen.
Which decisions actually need this
A PM makes dozens of calls a week and recording all of them is how the habit dies by Thursday. The filter is whether the question will come back, and these five reliably do:
- Anything where an option was rejected that a stakeholder still likes.
- Anything justified by data that will look different in six months.
- Anything constrained by a temporary condition—a deadline, a missing hire, a dependency.
- Anything a named person disagreed with on the record.
- Anything that will be inherited: if you leave, the next PM will ask why this is the way it is.
The limits of a private decision log
Most importantly, a personal archive does not make the team's context durable. If the reasoning behind a roadmap decision lives only in your notes, the organization is one departure away from losing it—your notes have merely moved the single point of failure. Decisions with team-wide consequences still need a written record the team can access; the private layer is for details that never make it into that record.
Recording dissent is also politically loaded in a way that a method article should not gloss over. A dated note saying who objected and on what grounds is enormously valuable for learning and can read as score-keeping if it surfaces badly. Record the argument rather than the person's character, and be deliberate about what you would be comfortable showing the person it names.
And preserved reasoning does not settle every reopened question. Circumstances change, and a decision that was correct under April's constraints may genuinely deserve reversal in October. The record does not end the debate—it changes what the debate is about, from ‘what did we decide and why’ to ‘has the constraint changed’, which is a far shorter conversation.
Capture the argument in the minute after the decision, attach the evidence, and link every reversal to what it reversed. Settled questions stay settled—and reopened ones reopen for a reason.