Nobody answers a preparedness survey honestly.

Decision 01

Study the serious games that already exist before designing one

We surveyed serious games on the market and took apart the methods they used, rather than starting from a blank page. World Rescue was the primary reference — a published serious game solving an adjacent problem well enough to learn its structure from.

What came out of that survey was the shape of our own solution: branching narratives over realistic scenarios, immediate feedback with the correct procedure explained, and assessment analytics running underneath the whole thing rather than a score at the end.

Rejected: designing from first principles. In a four-month project with two designers who had not built a game before, originality was the expensive option and the least useful one.

Decision 02

Let users write the game's action list, via card sorting

The game is only a measurement instrument if the actions it offers are the actions people would really consider. So we ran a card sorting activity in MIRO: candidate crisis actions on cards, ranked by importance, with participants free to add their own cards. Alongside it we ran interventions with varied scenarios to watch what actions people reached for under different framings.

We went in with a hypothesis: that international residents and Dutch residents would have different plans of action. It held. The output was a genuinely diverse list of choices to build into the game — one that neither pair of us would have written alone.

Traded away: a tighter, more designable action set. A list assembled from two populations is harder to balance than one we invented.

Decision 03

No paper prototypes

Standard process says sketch, then paper prototype, then digital. We sketched — the game's look and the UX cycle it moves through were both settled in that phase — and then skipped paper entirely.

The reason is specific to what we were making. This game's instrument is pressure: it measures decisions taken while a crisis feels real. A paper prototype cannot convey the severity of a crisis without visual cues and graphics, so testing on paper would have measured how people behave when nothing is at stake — which is the exact failure mode of the questionnaire we were replacing.

Rejected: the orthodox fidelity ladder. Testing a severity-dependent artefact at a fidelity that cannot express severity produces clean data about the wrong thing.

Decision 04

Illustration-style 2D, built from the card sort

We chose an illustrated 2D style, and derived the game screens directly from the card-sorting activities rather than inventing a new interaction language for them — refined, but structurally the same thing participants had already handled.

This decision has a cost we measured later, and it is in the results below.

Decision 05

Settle the information structure in one long room

Rather than iterate the structure asynchronously, we defined the game in a single long in-person meeting: what the user does on each day of the scenario, the mode of interaction, and the user flow through the whole thing. Branching narratives get expensive to restructure once built, so the structure was worth stopping for.

Decision 06

Two rounds of high-fidelity testing instead of one lo-fi and one hi-fi

Partly principle — the same severity argument as Decision 03 — and partly a time crunch we had by then earned honestly. We ran two phases of HiFi testing instead.

Traded away: cheap early failure. Every structural mistake we made had to be found at a fidelity that was expensive to change, and the "half-baked functions" in the limitations below are the bill for it.