Skip to content
Software Architecture Guide

Architectural Decision-Making Process

Explanation

Section titled “Explanation”

Teams do not make all architectural decisions in one session. Making one decision usually triggers further decisions right away. The decisions depend on each other, so the work is never finished at once.

Which decisions matter most. Architecture can be described as the decisions you wish you could get right early in a project. Only decisions that are hard and costly to change deserve to be recorded. Cost, effort, and risk help teams decide which decisions to tackle first.

When to decide. Teams aim for the most responsible moment. That is when they are best informed, not as early as possible. See Last Responsible Moment.

Comparing options. A decision counts as done only after the team has compared at least two options, at least semi-systematically. Listing fake alternatives to fill a template has no value. The team weighs the trade-offs between options. The decision record should state the problem and those trade-offs.

Who decides. An empirical study found that architectural decision-making is a very challenging thinking task. Most teams decide together, in a consultative way. In an advice process, anyone may decide but must consult affected people and experts. The goal is broad input, not consensus.

Uncertainty and revisiting. Teams decide under uncertainty. They record how confident they are and which context changes should trigger a new look. They also plan when to review and possibly revise a decision, so decisions keep iterating over the project.

Example

Section titled “Example”

A team building an online shop must choose how to store orders. They compare two real options and write down the trade-offs. They pick one and note their confidence and a review date. The choice immediately raises new questions, such as how to back up the data and how other parts of the system read orders. Each of those becomes its own decision.

Trade-offs

Section titled “Trade-offs”

Deciding late means the team is better informed, but waiting too long can leave other work blocked. Consulting many people brings broader input, but it takes more time than one person deciding. Recording confidence and review triggers adds effort, but it makes later revision easier.

Checks

Section titled “Checks”
  • Each recorded decision shows at least two options that were compared.
  • Each decision record states the problem and the trade-offs considered.
  • Decisions state when they will be reviewed or which context changes trigger re-evaluation.
  • Recorded decisions are the ones that are hard and costly to change.
  • Follow-up decisions triggered by a decision are identified.
Section titled “Related”

Sources

Section titled “Sources”