Architecture governance has a reputation problem, and the reputation was earned. Most technology leaders carry a memory of a review board that met every second Thursday, demanded documents nobody read again, and turned a two-week decision into a two-month one. When those leaders hear governance, they hear delay.
The tragedy is that the purpose of governance is the opposite of delay. Done well, governance is the mechanism by which an organization makes technology decisions faster, because the ground rules are clear, the decision-makers are known, and most choices never need a meeting at all. If your governance does not accelerate good decisions, it is not governance. It is ceremony.
If your governance does not accelerate good decisions, it is not governance. It is ceremony.
Start from decisions, not documents
The most useful question when designing governance is not what should we review? but which decisions deserve whose attention? Every organization has a small set of genuinely consequential technology decisions: commitments that are expensive to reverse, choices that touch regulated data or customer trust, dependencies that will outlive the people making them. Those deserve senior scrutiny. Everything else deserves a default answer and a fast path.
This is what proportionality means in practice. A three-tier structure serves most organizations: decisions that follow published standards proceed with no review at all; decisions that depart from standards in low-risk ways get a lightweight peer review; and the small number of high-consequence decisions get a proper forum with the authority to say no. The goal is for the overwhelming majority of decisions to travel the first path, and for engineers to know, without asking, which path they are on.
High-consequence decisions
Governance forum with authority to say no · Few
Low-risk deviations from standards
Lightweight peer review · Some
Decisions that follow standards
Proceed with no review at all · Most
Width reflects decision volume. Scrutiny rises toward the top; the overwhelming majority of decisions travel the base tier.
Standards as defaults, not gates
Standards and reference patterns are the highest-leverage governance instruments precisely because they remove the need for review. A good standard is a pre-made decision: use this integration pattern, this identity provider, this data classification scheme, and you never need to talk to us. Teams experience well-written standards as speed, not constraint: the safe path becomes the fast path, and the review queue shrinks to the cases that genuinely need judgment.
The corollary: every exception request is information. If teams repeatedly ask to deviate from a standard, the standard is wrong, outdated or badly explained. Healthy governance treats exception patterns as feedback and revises the default, rather than defending it.
Write decisions down
The cheapest governance practice with the highest return is the humble decision record: one page stating what was decided, why, what alternatives were considered and what would trigger revisiting it. Decision records turn governance from a gate into a memory. They stop the organization from relitigating the same choice annually, they make onboarding faster, and they create accountability without meetings; anyone can read why the landscape looks the way it does.
Keep the format small enough that people actually use it. Context, the options considered, the decision, the consequences accepted, and the condition that would cause a revisit. Five fields, one page, written at the time rather than reconstructed later. A decision record that takes an afternoon to produce will not be produced.
Reforming governance that has already gone wrong
Most organizations are not designing governance from scratch. They have a review board with a poor reputation and a queue, and the instinct is to add process to fix it. The more effective move is subtraction.
Start by looking at what the board actually reviewed over the last quarter and asking, for each item, whether a standard could have answered it. In most cases a large share could have. Publish those as standards, and the queue shortens immediately without anyone losing oversight of anything that mattered. Then separate the forum's two jobs, because reviewing routine work and deciding genuinely contested trade-offs are different activities that do not belong in the same meeting.
The signal that this is working is counter-intuitive: the forum should meet about the same amount but discuss far fewer things, each of them consequential. A governance body that reviews thirty items a month is a bottleneck. One that reviews three is making decisions.
A governance body reviewing thirty items a month is a bottleneck. One reviewing three is making decisions.
Measure the right thing
Finally, measure governance by its speed, not its activity. The health metric that matters is decision latency: how long does a team wait between raising a decision and being able to act? If the answer is weeks, your governance is consuming the value it exists to protect. Track it, publish it and treat increases the way you would treat a production incident.
Governance designed this way needs surprisingly little apparatus: clear decision rights, good standards, written decisions and one forum that meets with intent. It will not feel like the review board anyone remembers. That is the point.
Key takeaways
- Good governance makes decisions faster and clearer; if it only adds delay, it is ceremony.
- Right-size scrutiny to consequence: most decisions should follow standards and need no review at all.
- Standards and reusable patterns are pre-made decisions that remove the need for a meeting.
- Write decisions down, and measure governance by decision latency rather than activity.