Chapter 7 of 8 · From idea to working service~4 minutes

36 months, five stages, and at the end of each one an honest question: does it work or not?

I believe a design system is not made all at once, and it is not imposed before it exists. It is built a little at a time, tried out on real services, and extended only if the evidence shows it is worth it.

Prove first, then require. Institutions would be obliged to use it only after the system is stable and has worked in practice.


Stage 0 · Preparation (months 0–3)

It is decided who is accountable, and the money is secured. An inventory is made of the existing websites and services. Two pilot services are chosen, and how they work is measured before any change.
The question at the end: do we have the money, the people, the partners and the baseline data?

Stage 1 · First components (months 3–9)

The first building blocks appear: buttons, fields, error messages, the first “recipes” (CNP, the Romanian personal numeric code; address; “Before you start”). Everything is tested in public. The basic interface is available in Romanian, Hungarian and English.
The first thing the public would see, in months 5–6: the strip that helps you recognise an official website.
The question at the end: can at least two real services be built with these building blocks?

Stage 2 · Real-life test (months 9–18)

The two pilot services are rebuilt with the system and measured again: do people complete the application more often? More quickly? Do they need less help? At least 30 testing sessions are held with people with disabilities.
The question at the end: is it better, with no serious problems, at a verified cost?

Stage 3 · Expansion (months 18–30)

The first stable version (1.0) appears. A small group of institutions starts using the system. Based on the pilots, an estimate is made of how much adoption would cost town halls.
The question at the end: are the results repeated in other institutions too? Can the team support more users?

Stage 4 · Decision (months 30–36)

An independent evaluation says what worked and what did not. Only now is it decided whether the system becomes mandatory, through which legal act and with what funding.


What it would cost

Estimates. They need to be validated.

Period Team Estimated cost
The first 9 months 8–10 people €0.46–0.69 million
The next 9 months 10–13 people €0.56–0.86 million
Per year, once mature 13–18 people €1.08–1.80 million

As an order of magnitude: the annual cost of a mature system would be about 6–9 euro cents for every inhabitant of Romania (own calculation, based on a resident population of 19.04 million on 1 January 2026, provisional data from the National Institute of Statistics, INS).

What is not included here: the cost for each institution of adapting its own websites. For town halls, I propose a separate fund, from European money, paid only after an independent check confirms that the adaptation has been done. Italy did the same.


When we stop or change direction

An honest plan also says when it stops. I would stop or rethink the programme if:


Why not a law first, with a fixed deadline?

I analysed how the other 26 EU member states went about it. All the countries that made a design system mandatory built it first and only then imposed it. And the levers that worked were not fines, but the conditions for funding and approving IT projects: no compliance, no money.

Why I wrote all this →

For specialists
  • Review stages G0–G4, with public deliverables and criteria. A stage does not start if the previous one has not met its criteria.
  • Adoption regime: L0–L9 (months 0–9) voluntary (pilots, based on agreements); L9–L18 recommended for eligible new projects; after G2, a legal analysis of the funding and approval criteria; after G3, a model clause in procurement and the adoption fund; after G4, a possible regulatory framework (minimum requirements in an annex, deadlines running from the publication of each version, a right to feedback with a response deadline).
  • Cost assumption: €55,000–65,000 a year for one full-time equivalent post. The totals are the exact sum of the ranges per line (team, university programme, user testing, localisation, infrastructure, legal and communication).
  • Formula for the adoption fund: Σ (eligible institutions per category × standard cost validated in the pilots) + independent verification + risk reserve. The number of authorities funded is set after the calculation.
  • Main indicators: completion rate, time to completion (we do not accept shorter times if drop-out rises), need for assistance, real accessibility, cost per transaction, reuse.
  • Possible funding sources: the budget of the Romanian Authority for Digitalisation (ADR) (the core), the EU Technical Support Instrument, PoCIDIF (Romania’s operational programme for smart growth and digitalisation) and the 2028–2034 financial framework, Digital Europe.
  • Comparison: Slovakia, Hungary and Portugal have acts with the force of law; Italy tied €400 million from its national recovery plan (PNRR) to compliance checks; Germany keeps its KERN standard voluntary.

What I base this on

The links lead to the public versions of my research documents; the section number (§) takes you straight to the paragraph cited.

Claim Document Where
The stages, checkpoints, deliverables The full proposal §12
The criteria for stopping The full proposal §12.1
The budget, the team, the adoption fund The full proposal §14
The adoption regime The full proposal §13.1
The indicators The full proposal §15
All states built before imposing; the levers that work EU member states' laws §1 (points 5–7)
Italy: €400 million from the PNRR, paid after verification EU member states' laws §1, §3.4
6–9 euro cents per inhabitant Own calculation: €1.08–1.80 million / 19.04 million inhabitants INS, press release no. 212 of 28.08.2026 (provisional data)