Verification

What this phase is

Verification is establishing the validity of your effort to discover and implement values — a form of quality control, in Flanagan and Nissenbaum's terms. Discovery found the values, implementation built them into the game; verification asks whether any of it actually worked. Did the game embody the values you claimed, and did it deliver the learning, health or behavioural outcomes it was funded to deliver?

Why serious games face a harder test

Entertainment games are validated by the market: if it sells and people play it, it works. Serious games get no such shortcut. Stakeholders — clients, clinicians, funders, education departments — expect evidence that a game for learning, health or behaviour change does what it promises, and that evidence is expensive to gather on budgets and timelines that rarely allow for it.

The developer of physiotherapy games put the burden plainly: "in serious games, your approach has to be validated in a way that games do not have to, if a game is fun, it's fun. You know, that's it, move on. You know? Whereas serious games, you have that added responsibility..." (P2). Every designer interviewed valued verification and saw it as part of making better games. Almost none had the resources to do it properly.

Patterns from practice

The process problem

Formal verification changes what you can build. The clinical director of the physiotherapy games noted that a clinical trial would make their product "essentially immutable" (P2) — at exactly the moment weekly revisions were needed to stay competitive. Choosing rigorous verification is a commitment to a fixed, proven product over flexibility and continuous improvement. You cannot iterate your way past a completed trial without invalidating it. The early childhood assessment games show the same force from the clinical side: to preserve research and clinical integrity, each iteration had to stay very similar to the last — "the science behind it... impacted them more than the architecture and game engines that we're using" (P9).

Trials are also narrower than they look. Verification is regional and conditional: a trial run on a 9.7-inch tablet proves nothing about any other screen size, yet games are expected to run across a range of devices — especially for kids, who play on whatever screen is handy. There is a real gap between what has been verified and how the game will actually be encountered.

And clinical trials decide who gets to play at all. Millions of dollars in testing costs act as a barrier to entry that suits established players and their investors, while sidelining small studios and experimental designers. That prices regulatory security and continuity above rapid iteration and innovation.

The evidence problem

The ideal evidence base is well understood: multiple playtest sessions, observation, pre- and post-surveys, embedded in-game questions and tracking. Nearly every designer interviewed said lack of resources stopped them getting there. There is a standing tension between the methodological ideals of a research-led approach and the pragmatic limits of the funding frameworks serious games actually live in.

Long-term impact sharpens the problem. The conservation games aim at behaviour change over five to ten years — and measuring that impact is near impossible. The same holds wherever the game targets slow attitudinal change, feeds into a wider suite of clinical measures, or sits inside a larger learning package: isolating what the game contributed can be incredibly difficult, if not impossible.

The double standard

In the agricultural and mining sector, studios are routinely asked to prove their games are more effective than existing training methods — while the clients often have no idea how well their existing training works. Traditional formats get the benefit of the doubt; games get the burden of proof. Serious games are in a constant contest with entrenched methods and are held to a far higher evidentiary standard than the incumbents they seek to replace. Name this asymmetry early with your client: if the bar is "better than current training", ask what evidence exists that current training clears any bar at all.

Match your evidence to your claims

Align the weight of your evaluation with the strength and scope of what you claim. A rough ladder:

• Awareness or attitude claims — playtesting, observation, and pre/post surveys are usually proportionate.
• Skill or knowledge claims — add embedded in-game tracking and assessment items, so the evidence comes from play itself.
• Behaviour-change claims — plan follow-up studies over realistic periods, use proxy measures where direct measurement is impossible, and state the limitations honestly to clients and users.
• Clinical-efficacy claims — controlled trials, with eyes open to what they cost: an immutable product, device- and region-specific results, and a serious barrier to entry.

Two habits make any rung of the ladder workable. Decide your metrics early and collaboratively — designers, clinicians and researchers together, agreeing on what is feasible to measure before development starts, not after launch. And communicate limitations honestly: a modest, truthful evidence claim serves your values better than a strong claim you cannot support.

Questions to ask your team

1. What exactly are we claiming this game does — awareness, skills, behaviour change, clinical outcomes? Does our evaluation plan match that claim, or exceed our budget's honesty?
2. Have designers, domain experts and researchers agreed on metrics before development, and are they feasible to actually collect?
3. If we run a formal trial, can we live with the product becoming effectively immutable? Who wins and who loses from that trade?
4. Will our evidence hold for the devices, regions and audiences the game will really be played on — or only the ones we tested?
5. Our client wants proof we beat their existing training. What evidence do they have that the existing training works?
6. Can we embed measurement in the game itself, or would visible tracking corrupt the experience or the data?
7. Where long-term impact is the real goal, what proxy measures and follow-up periods are realistic — and have we told the client the limits?
8. If the honest answer is "we can't measure this", are we prepared to say so rather than overstate?