Sort evidence before debating a team decision

Before a team debates what to do, check what its evidence actually says. Several confident accounts may come from one original message, while two apparently contradictory observations may describe different situations. An evidence-sorting session produces a traceable claim register that a later decision review can use.

Bound the question and collect permitted records

State the factual question without embedding the desired answer. Gather a manageable set of relevant records that participants are allowed to inspect. For each item, record its origin, date, context, and a reference through an existing document or system. Keep unrelated private conversations out of the collection.

Fictional practice example: notes disagree about why a document review was delayed. One message says an approver was unavailable; a task record shows a missing attachment; a colleague remembers that the brief changed. The exercise is to establish what each item supports, not to assign blame or invent a customer incident.

Classify the kind of support

Distinguish a direct observation, someone else's report, an inference, and a proposal. Keep the classification attached to the statement. A screenshot may directly show a missing field at a particular moment, but a comment explaining why it was missing can remain an inference.

Preserve exact wording where a small qualification matters, while minimizing sensitive content. If a source says a review might be delayed, do not rewrite it as a confirmed delay. Repeated references to the same note are useful for finding it, but they do not create independent confirmation. Trace copied accounts back to their original source when possible.

Group by claim, not by preferred option

Create one entry for each disputed claim and attach supporting and conflicting records. In the fictional example, separate whether an attachment was absent, whether approval was needed, and whether somebody was unavailable during the relevant interval. Those are different claims even when they concern the same review.

Retain context that could explain a contradiction: document version, date, viewer access, or the stage of work. Avoid averaging incompatible accounts into a vague compromise. A record from the first draft and a record from the revised draft can both be accurate. If their contexts cannot be recovered, mark that limit rather than choosing the more confident speaker.

Write the limits of the evidence

For each claim, state what is established, what remains uncertain, and what would distinguish the plausible explanations. Prefer a specific missing observation to a request for more evidence in general. For the example, checking which attachment was available at the review handoff may be more useful than collecting more opinions about who was responsible.

Do not rank the proposed solutions during this step. Evidence can constrain a later choice without determining it: a team may still need to consider cost, fairness, or risk. Keeping these judgments separate prevents an appealing option from changing the description of the facts behind it.

Test the register before handing it over

Ask a participant to select one important claim and follow it back to its source. If the origin cannot be found, mark the claim unverified and preserve its stated support type: an unverified report remains a report, not an assumption by default. Do not treat it as established. Check whether allegedly independent confirmation actually copies one account. Also check that unresolved contradictions survived the summary.

The completed artifact contains claim references, source context, support type, limits, and the next observation needed. It is ready for a decision review when another person can inspect those links and understand the uncertainty without hearing the whole original conversation.