A team chooses to build an internal tool. Three months later, the tool works, but nobody agrees on why the team built it.
One person remembers a capacity problem. Another remembers a quality problem. The project board shows completed tasks. It does not show the assumptions behind the commitment.
A decision journal gives that choice a dated record. It holds the alternatives, available evidence, unresolved questions, and conditions for another look.
The template below is a proposed working method for business operators. The example is fictional. It makes no claim about a client's results or my own operating history.
Give One Decision Its Own Record
Start with a choice that commits attention, changes a workflow, or creates an ongoing obligation. Routine task completion usually does not need a separate journal entry.
Write the decision as a question with a boundary. “Should we automate?” leaves the scope open. “Should we build a checker for this intake worksheet?” identifies something the team can evaluate.
Meeting minutes record a discussion. A task list records work. A business case makes an argument for a proposed commitment. This journal preserves the reasoning at the point of choice, including reasons to reject the preferred option.
Atlassian's decision documentation template includes background, options, action items, and an outcome. The journal proposed here adds an explicit record of assumptions and later observations. Those additions are editorial choices, not Atlassian requirements.
Keep the journal beside the project material. Link to detailed notes instead of copying every discussion into it. The entry should remain readable when the original participants are unavailable.
Fill In the Decision Journal Template
Complete the initial entry before the decision's consequences become visible. If the decision already happened, label the entry as a reconstruction. State which records support it.
| Field | Fill in |
|---|---|
| Decision and boundary | [The choice, affected workflow, and exclusions] |
| Recorded on | [Date of the original entry] |
| Decision owner | [Person authorized to select an option] |
| Contributors | [People providing evidence or operational constraints] |
| Options | [Credible alternatives, including continuing the current process] |
| Known facts | [Checked observations, evidence locations, dates, and limits] |
| Assumptions | [Unverified conditions the choice depends on] |
| Selection and reason | [Chosen option and the tradeoff accepted] |
| Expected observations | [What the owner expects to see if assumptions hold] |
| Reconsideration conditions | [Evidence that would weaken the choice] |
| Commitments | [Authorized work, resources, and continuing ownership] |
| Review trigger | [Event or date that starts another review] |
| Dated updates | [Later evidence, revised decisions, and next actions] |
The decision owner needs actual authority. A journal cannot grant it. If approval belongs elsewhere, record the proposed choice and the remaining approval step.
Atlassian's DACI framework distinguishes a driver who coordinates the process from an approver who makes the decision. It also identifies contributors and people to inform. Borrow that distinction when one person gathers evidence and another authorizes work.
For a small team, roles can overlap. Write down the overlap. Avoid inventing a committee when one authorized person can make the bounded choice.
Separate Known Facts From Assumptions
A fact needs a source and a scope. “The worksheet contains inconsistent entries” is incomplete without the reviewed material and the rule used to identify inconsistency.
An assumption describes something the team has not established. “The same rules will cover future submissions” belongs in that category until evidence supports it.
Record the difference without turning every sentence into a confidence score. Useful labels include observed, reported, assumed, and unknown. Define them in the entry if the distinction matters.
For example, a reviewer may report that an intake step is difficult. That report establishes the reviewer's account. It does not establish the cause, frequency, or effect across every submission.
Attach the permitted source material where appropriate. Use the knowledge management structure to make the record retrievable. Keep confidential operational evidence within its authorized location.
Do not turn missing information into a zero. Unknown maintenance effort is still unknown. Give the question an owner and decide whether it blocks the commitment.
Work Through a Fictional Manual-Versus-Tool Choice
Consider a fictional service team reviewing project-intake worksheets. Staff compare submitted fields with an agreed checklist. The team is considering a small internal checker.
Everything in this example is hypothetical. No figures represent measured results, and no option carries a calculated return.
Decision boundary: Choose how to check required worksheet fields. Exclude document interpretation, exception approval, and messages to customers.
Decision owner: The fictional operations manager. The intake reviewer supplies examples. The proposed tool maintainer evaluates the build and maintenance requirements.
| Option | Work it commits | Uncertainty it leaves |
|---|---|---|
| Continue manual review | Maintain the checklist and reviewer assignment | Whether current exceptions can be described consistently |
| Improve the manual worksheet | Agree on field definitions and clarify the checklist | Whether better instructions address the observed disagreement |
| Build a small checker | Implement agreed field rules and assign maintenance | Whether stable rules cover enough of the permitted task |
The alternatives need comparable treatment. The tool requires ongoing ownership. Manual review requires an available reviewer. A clearer worksheet requires agreement on the underlying definitions.
Known facts within this fictional scenario: The team has a checklist, sample worksheets, and conflicting interpretations of one required field. The team has not agreed on an exception rule.
Assumptions: The input format will remain stable. A maintainer will remain available. The unresolved field can stay under manual review.
Decision: Clarify the worksheet first. Defer the checker until the team agrees on the rules it would implement.
Reason: The unresolved question concerns the rule itself. Building a checker now would encode a disputed interpretation.
This choice does not declare manual processing universally better. It states why one team would defer one proposed tool under the example's conditions.
Describe What You Expect to Observe
An expectation should identify evidence that can support or challenge an assumption. “The process will improve” gives the reviewer no concrete comparison.
In the fictional example, the owner expects reviewers to apply the clarified field definition consistently to a permitted sample. The owner also expects unresolved cases to reach the named exception owner.
Record how the team will examine those expectations. Identify the sample boundary, reviewer, evidence location, and limits. Do not present a convenient sample as proof about every future case.
The expectation can be qualitative. Reviewers might explain why each flagged field needs attention. That observation differs from a target for revenue or a claim about saved time.
If a measure is available, record its definition before collecting results. State what counts as an exception and which cases remain outside the comparison.
An operator dashboard can point to relevant observations. Keep the decision record linked to the underlying evidence. A changed dashboard number alone does not explain why it changed.
Set a Review Trigger That Can Be Recognized
A calendar date is useful when somebody owns the review. An event can be more relevant when the decision depends on a specific development.
For the fictional team, the trigger could be completion of the checklist review. Another trigger could be a change to the intake format. Both connect the next decision to the original uncertainty.
Write what the owner must decide at that point: continue manual review, revise the worksheet again, or prepare a bounded checker proposal. Reviewing does not imply expansion.
Set an additional trigger for a failed prerequisite. If the proposed maintainer becomes unavailable, reopen the tool option before anyone commits to it.
Place the review task in the team's active work system. Link back to the journal. A date buried in a note has no assigned action.
Preserve the Original Entry and Append Updates
Keep the original reasoning intact. Append later evidence under a dated heading. If an original fact was wrong, add a correction identifying the error and its source.
Do not silently replace “maintenance owner unknown” with a person's name after assignment. The later assignment matters because it changes what the team can responsibly consider.
A fictional update might read: “2026.11.12: Reviewers accepted the revised field definition. One exception remains unresolved. Continue manual review for that field. Prepare a separate checker proposal for the agreed fields.”
The date and events in that update are invented for illustration. The useful structure is observation, interpretation, decision, and next action.
When an update changes the scope, record the new authorization separately. The original decision to clarify a worksheet does not authorize a tool build or a live workflow change.
If the team changes its documentation system, retain a readable copy of the original entry and updates. Preserve access restrictions with the evidence.
Review the Decision Process and the Outcome Separately
At the review, ask two sets of questions.
First, examine the process. Did the owner consider credible alternatives? Did the entry separate evidence from assumptions? Were the chosen scope and approval conditions clear?
Then examine events. What happened? Which expectations held? Which assumptions failed? Which observations remain unavailable? Record external changes without using them to erase earlier gaps.
A working tool can coexist with a weak decision process. A rejected tool proposal can reflect a careful comparison. The outcome alone does not answer the process questions.
In the fictional example, the clarified worksheet might still produce disagreements. That finding would challenge the expected observation. It would not establish that building the original checker would have solved the problem.
Avoid scoring the decision with evidence the owner could not have known. Identify what was available at commitment and what became available later.
Use the weekly review framework to surface entries whose triggers have arrived. Reserve the detailed comparison for the decision record itself.
Close With the Next Decision
End each review with an action the owner can carry out. Continue the choice, change it, stop the work, or gather named evidence before deciding again.
Write the unresolved question if the evidence remains incomplete. Assign its owner. Set its next trigger. Do not manufacture a conclusion to close the document.
Keep the original reasoning visible.
Sources
- Atlassian: DACI Decision Documentation Template. Background, options, action items, and recording a decision. Checked 2026.10.28.
- Atlassian Team Playbook: DACI Decision-Making Framework. Coordination, approval, contributor, and informed roles. Checked 2026.10.28.