Civic technology

Most civic technology fails after delivery rather than during it. The build works, the pilot reports well, and then the person who understood it leaves. Method here is aimed at that failure specifically.

Maintenance cost is a design input

The question asked before a tool is chosen is who will run this in three years and what they will know. A stack that a small organization cannot staff is a stack that will be abandoned, regardless of how well it performs at handover. Reducing the number of moving parts is usually worth more than adding a feature.

Prefer the boring option

Established tools with long support horizons are chosen over newer ones with better ergonomics. Where a newer approach is worth the risk, the tradeoff is documented rather than assumed, which is the purpose of the technical research described under the capability list.

The exit is designed at the start

Data is exportable in an open format, content is stored as files rather than only in a database, and no step of the workflow depends on a service only this practice can access. If an organization decides to leave, it should be able to, and the handover documentation says how.

Accessibility is a build gate, not a review

Interfaces target WCAG 2.2 AA, checked automatically across routes and supplemented by keyboard and screen-reader review. Treating it as a gate means a regression fails before it ships rather than being logged for later. Accessibility sets out the layers of testing.

Data collection is bounded by the question

A civic tool tends to accumulate fields because collecting them is cheap. Each one is a retention obligation and a disclosure risk, so the schema is argued down to what the decision needs. Custody, consent, retention, and decision rights are settled under Governance before the first record exists.

What this practice does not claim

None of the above has been applied to a large multi-year public deployment. The approach comes from software and reporting work at smaller scale, and the pages that describe delivered work label what it was. Where a claim here has not been tested at scale, that is the reason.