Shuichiro Ogawa
日本語

Notes · updated 2026-08-07

Overview

The question of this note is whether services exist for learning system design scenario-based. The analytical frame is the seven components of Goal-Based Scenarios (GBS, the learning-environment design framework by Schank and colleagues): learning goals, mission, cover story, role, scenario operations, resources, and feedback. The theory and its scholarly literature are covered by the GBS literature map; this note is its industry counterpart. Twenty-eight data points were collected across two streams, vendor-primary and developer-voice (ledger: source/review/system-design-scenario-learning/industry.md). The mapping to the seven components is this note’s analysis, not the services’ self-description.

The conclusion first: learning services equipped with scenario structure (narrative, role, mission, feedback) are plentiful in industry, but few make system design itself the scenario operations (the learner’s central activity). Most services’ operations are operational work (building, incident response); the ones whose operations are design decisions are led by Architectural Katas and AWS SimuLearn, and they are a handful. And no service was found that implements the GBS resources component (delivering expert stories at the moment they become needed).

Exercises Whose Operations Are Design

Exercises that make design decisions the learner’s activity were confirmed as follows.

Architectural Katas (originated by Ted Neward, maintained by Neal Ford) is the most complete form. A team of three to five plays the architect role, receives a fictional client’s deliberately vague requirements (for example, a sandwich chain that wants an online ordering system), questions a moderator playing the client to refine the requirements, designs an architecture, presents it, and receives critique and votes from the other teams. Neward states the motive plainly: architects get to architect only a few times in a career, so they need a place to practice. This matches the learning-by-doing premise of GBS. It is free and self-hostable, and the O’Reilly competitive version adds judge feedback (Neal Ford, Mark Richards, and others).

AWS SimuLearn has the learner interview a generative-AI virtual customer, elicit requirements, and propose an architecture. An architecture diagram updates in real time as the dialogue proceeds, and evaluation is multi-axis: technical accuracy plus communication and customer focus. It makes the front half of design work (requirements to proposal) the operations, which distinguishes it from build labs. Some learning plans are free and available on demand.

AWS Cloud Quest is a 3D role-playing game solving virtual-city residents’ problems, with each quest structured Learn, Plan, Practice, DIY; the Plan phase involves architecture design, though the center of gravity is the subsequent build. Microsoft Learn’s Challenge project format has learners pick a customer and produce architecture analysis and design documents along a planning workbook, though assessment is a pass/fail quiz.

Exercises Whose Operations Are Operations Work

The richest scenario structures are found not in design but in operations exercises.

AWS GameDay is a team competition set at the fictional Unicorn.Rentals (the world’s largest mythical-creature rental company): with only fragmentary documents left by a collapsed management team, participants join as new hires and stabilize the system. A real-time leaderboard and deductions for missing best practices provide feedback. It is event-only, not an always-available material.

KodeKloud Engineer is a free role-play platform: hired as a SysAdmin at the fictional xFusion Corp Industries, the learner receives tasks through a ticket system and resolves them. Narrative, role, mission, and automated judgment are all present, but the tasks are operations on a predefined infrastructure; no design decisions are involved.

Google SRE’s Wheel of Misfortune is a role-play reenacting real past incidents: a game master presents the scenario, the on-call players talk through diagnosis and mitigation, and the GM responds in real time. Using real postmortems as material reads as industry practice independently arriving at the case-based-reasoning foundation of GBS (learning from experts’ cases). Commercially, Uptime Labs offers an incident simulator in which AI-played stakeholders (CEO, CTO, customer support) apply pressure while the learner acts as incident manager, with skill-gap analysis across 40+ metrics. PagerDuty publishes its internal incident-response training (a six-role system, reenactment recordings of a real incident) for free. In security, TryHackMe’s SOC Simulator and Hack The Box provide fictional-organization narratives and roles, useful as reference cases for scenario design.

The Industrial Entrenchment of Cover Stories

The fictional-company device is entrenched in all three major vendors’ materials. AWS uses Unicorn.Rentals; Google Cloud uses the Cymbal series (Cymbal Bank is a national retail bank, and its Challenge Labs cast the learner as “a cloud engineer at Cymbal Bank”); Microsoft uses Contoso, and Applied Skills labs deliver tasks through an interface mimicking emails from a boss or customer. Cover story and role, in other words, are standard equipment in industrial learning design. These narratives exist for motivation and context, however, and are not connected to the design principle GBS theoretically demands: that expectation failure drives learning.

As an exception, the Cloud Resume Challenge (Forrest Brazeal) has no narrative, but its creator states he deliberately under-specified the instructions so learners would fall down late-night research rabbit holes — a mission-driven design that intentionally builds in expectation failure, making it the most GBS-interesting case. Its feedback is also human rather than automated: personal code review by Brazeal for finishers and peer review in the community. The lineage of this deliberately-designed-failure approach (learning-science theory and industry implementations) is organized separately in Designing Failure Into Learning.

Checking Against the Seven GBS Components

Mapping the main services against the components (○ = confirmed, △ = partial, − = absent; learning goals omitted since all services carry them implicitly):

Servicemissioncover storyroleoperations = design?resourcesfeedback
Architectural Katas○○○○− (questions to moderator only)○ peer critique + votes
AWS SimuLearn○○○○ (requirements → proposal)△ (guided labs)○ multi-axis real-time
AWS Cloud Quest○○○△ (Plan phase only)△ (Learn-phase videos)○ auto-verify + badge
AWS GameDay○○○△ (operations-centric)−○ leaderboard + deductions
Cymbal Challenge Lab○○○− (build)−○ auto-graded
MS Applied Skills○○○− (configuration)−○ auto-graded
KodeKloud Engineer○○○− (operations)−○ auto-judged
Wheel of Misfortune○△ (real cases)○− (response)△ (postmortems = real cases)○ GM’s immediate responses
Cloud Resume Challenge○−−△ (configuration choices)−○ human review
Uptime Labs○○○− (response)−○ multi-axis report

Two structural absences emerge from this mapping. First, the combination of design and narrative is rare. The services with the richest narratives and roles (GameDay, KodeKloud Engineer, Cymbal) skew toward operations and building, while the services whose operations are design (Katas, SimuLearn) have comparatively spare narrative devices. SimuLearn is nearly alone in having both, and even there the narrative is confined to a single scene of customer dialogue.

Second, the absence of resources. GBS requires delivering experts’ stories at the moment a learner fails and needs them. No service implementing this was confirmed. The closest is Wheel of Misfortune’s use of real incident postmortems, but those are scripts the GM narrates in advance, not just-in-time delivery in the context of failure. Feedback across services is dominated by automated grading (pass/fail, deductions, leaderboards); the coaching-in-failure-context that GBS calls for is only partially served by SimuLearn’s multi-axis evaluation and the human GM format.

Implications for Student Education

For teaching students system design scenario-based, using the seven GBS components as a design checklist and composing pieces is more realistic than adopting any single service. As the design-as-operations core, Architectural Katas is free and self-hostable, and an instructor playing the moderator (client) can tune the cover story, the role assignment, and the requirements dialogue on the spot. AI-driven requirements elicitation can be supplemented with SimuLearn’s free plans, and the missing resources component has to be hand-built by the instructor: bringing published incident reports and design documents out at the moment of failure. To ground this assembly in evidence, the state of GBS’s empirical research (skewed to quasi-experiments, no confirmed meta-analysis) should be checked in the GBS literature map before designing. Reference material and review tools for correct design are covered in Tools for Learning System Design, and simulators that show a configuration’s consequences in the simulation note.

Gaps

  • No service was found corresponding to a full GBS implementation: system design as scenario operations, with narrative and role, plus just-in-time expert stories.
  • No study validating the learning effectiveness of Architectural Katas was found (only the format description and the creator’s intent).
  • Educational-institution terms for the vendor exercises (GameDay, Cloud Quest, SimuLearn) could not be confirmed from official pages.
  • No first-person testimony was obtained on whether any of these services were designed with reference to GBS or other learning theories; on current evidence, the narrative/role/mission equipment is an industrial convergence independent of the theory.
  • No independent survey shows the educational effectiveness of scenario-based exercises; effectiveness research belongs to the scholarly stream of the GBS literature map.

References

All accessed 2026-08-07. For ledger details (methodology, position judgments, all verification flags), see source/review/system-design-scenario-learning/industry.md. For the scholarly grounding of the seven GBS components (Schank, Berman & Macpherson 1999 and others), the references of the GBS literature map are canonical.

Vendor official (T1v)

Third party (T3)

← All Notes · Home