Guide · Architecture governance

Measuring architecture governance: the KPIs that matter

“Is our architecture governance actually working?” is a fair question — and one most practices can't answer. A small set of KPIs turns governance from a black box into something you can prove.

Why measure governance

"Is our architecture governance actually working?" is a fair question, and most practices cannot answer it. They can describe the process and list the meetings, but not demonstrate that anything is better as a result.

That gap is dangerous in a specific way: functions that cannot evidence their value are the ones cut when budgets tighten. Measurement is partly about improvement and substantially about survival.

What to measure

Governance KPIs fall into three families. A balanced set draws from all three, because measuring only one produces predictable distortion.

Throughput — is governance keeping up?

• Reviews completed per period against submissions received. Divergence signals a growing backlog before anyone complains.
• Average time from submission to decision. The single most useful number for your credibility with delivery teams.
• Percentage decided within target turnaround. More honest than the average, which hides outliers.

Quality — is governance changing outcomes?

• Decision mix. Proportion approved, conditioned, reworked, rejected. All-approve means the gate is theatre; mostly-rework means standards are unclear upstream.
• Conditions closed on time. The clearest indicator of whether decisions actually stick.
• Findings raised per review. Falling over time can mean improving submission quality — or fatigue. Read it alongside decision mix.
• Coverage. Proportion of qualifying projects that actually came to governance. Low coverage means you are being routed around.

Risk — is exposure visible and controlled?

• Open exceptions, with trend. See the exception process.
• Overdue exceptions as a proportion of open. The clearest single signal of governance decay.
• Average exception age. Rising age means remediation is not being resourced.
• Principle compliance rate across reviewed solutions.

Keep it to a handful

Eight to fourteen KPIs is the practical range. Fewer and you cannot see the picture; more and nobody maintains them and the pack becomes an end in itself.

Choose deliberately across the three families. A set weighted entirely to throughput will drive fast, shallow reviews. One weighted entirely to quality will produce thorough reviews that nobody can get a slot at. The balance is the point.

Watch what the measure does to behaviour

Every governance metric can be gamed, usually unintentionally.

KPI	Distortion if used alone	Balance with
Time to decision	Rushed reviews, waved-through solutions	Decision mix, findings per review
Number of reviews	Trivial items pulled in to inflate volume	Coverage of qualifying projects
Approval rate	Either rubber-stamping or obstruction, depending which way you push it	Conditions closed on time
Open exceptions	Discouraging people from raising them at all	Coverage; shadow architecture discovered

Report it regularly

Monthly for the architecture team, quarterly for leadership. The leadership version should fit on a page and lead with the trend, not the table — three or four numbers with direction of travel, then a short narrative on what changed and what you are doing about it.

Use RAG thresholds set in advance rather than judged after the fact. A KPI that turns amber because it crossed a line you defined last quarter is credible; one that turns amber because it felt amber is not.

And report the uncomfortable ones. A pack that is entirely green is not believed by anyone who has worked in a real organisation. Showing an overdue exception count that is rising, alongside what you are doing about it, buys far more trust than showing nothing.