What is the architectural problem being solved? Describe the situation that makes a decision necessary: the constraints, the forces in tension, and what happens if no decision is made. Be specific. "We need to modernize" is not context. "The current batch process runs on a 7-year-old SQL Server cluster that cannot horizontally scale, and the growth forecast requires 5× throughput within 18 months" is context.
State the decision in a single, unambiguous sentence beginning with an active verb. "We will…", not "we should consider…" or "it is recommended that…"
For each alternative, state it clearly, explain why it was considered seriously, and state the specific constraint that ruled it out. "Not a good fit" is not a reason. "Exceeds the CFO's $2M capital ceiling for Year 1" is a reason.
| Alternative | Why it was considered | Why it was ruled out (specific constraint) |
|---|---|---|
List the specific, verifiable constraints that made this the correct decision. Distinguish between verified constraints (confirmed by evidence) and assumptions (not yet confirmed). Assumptions must be flagged: an ADR built on unverified assumptions is fragile.
| Decision Driver | Verified or Assumption? | Evidence / Source |
|---|---|---|
What does this decision commit the organization to? List both the positive consequences (what it enables) and the negative consequences (what it forecloses, what it will cost, what risks it introduces). Honest ADRs include both.
What signal would indicate that this decision should be revisited? Name the specific condition, not "if things change," but "if the transaction volume exceeds 50,000/sec" or "if the regulatory framework changes to require data residency in a specific jurisdiction." This is the mechanism that prevents ADRs from becoming permanent constraints that outlive their original rationale.
| Stakeholder | Role | Consulted or Informed? | Agreed? (Y/N/Conditional) | Condition (if applicable) |
|---|---|---|---|---|