Skip to content

Cheat Cards

One-page printables to keep alongside your workstation or pin to a team wall. Each card is designed for A4 landscape. A LaTeX-rendered PDF is linked at each card heading; alternatively use your browser’s Print or Save as PDF (Ctrl/Cmd+P) on this page — the print stylesheet renders one card per landscape page.

A combined PDF of all cards is available: Download all cards (PDF).


#ViewWhat it describesPrimary audience
3.1Logical ViewComponents, services, patterns — what the solution isArchitects, Developers
3.2Integration & Data FlowData flows, APIs, integrations — how components talkIntegrators, Architects
3.3Physical ViewHosting, compute, networking, environments — where it runsInfrastructure, DevOps
3.4Data ViewData stores, classification, retention — what data it holdsData Architects, Compliance
3.5Security ViewIAM, encryption, monitoring, threats — how it’s protectedSecurity, CISO, Compliance
3.6ScenariosUse cases, ADRs — how it behaves and whyAll stakeholders

Remember: No single view tells the full story. Together, they answer every reasonable question from any reasonable reviewer.


Tier 2: Quality Attributes (Section 4) — cross-cutting

Section titled “Tier 2: Quality Attributes (Section 4) — cross-cutting”
#AttributeAssess across every viewKey questions
4.1Operational ExcellenceCan we operate it in production?Observability, runbooks, alerts, incident response
4.2Reliability & ResilienceWhat happens when something fails?RTO, RPO, DR strategy, backup, fault tolerance
4.3Performance EfficiencyDoes it meet its targets at load?P95/P99 latency, throughput, concurrency, growth
4.4Cost OptimisationIs it priced to the value delivered?Capex/opex, unit economics, commitment discounts, exit cost
4.5SustainabilityIs it environmentally responsible?Carbon baseline, efficient resources, SCI metric

Document tradeoffs explicitly — when one quality attribute constrains another, name the trade-off and the rationale.

Security is a View (3.5) — it’s cross-cutting enough to warrant its own structure, so don’t look for it here.


Section 6.1–6.5 Decision Making & Governance

Section titled “Section 6.1–6.5 Decision Making & Governance”

Constraints (6.1) — fixed limitations the design must work within

IDConstraintCategoryImpactLast Assessed
C-001regulatory / technical / commercial / organisational / time

Assumptions (6.2) — things treated as true but not verified

IDAssumptionImpact if falseCertaintyStatusOwner
A-001high / medium / lowopen / closed

Risks (6.3) — potential events with negative impact

IDRiskSeverityLikelihoodOwnerMitigationResidual
R-001H/M/LH/M/LH/M/L

Dependencies (6.4) — external factors this design relies on

IDDependencyDirectionStatusOwner
D-001inbound / outboundcommitted / not-committed / resolved

Issues (6.5) — problems that have already materialised

IDIssueImpactOwnerResolution PlanStatus
I-001H/M/Lopen / in-progress / resolved

Golden rule: Every row needs an owner and a date. Anonymous = unaccountable.


ScoreLevelWhat it means
0Not AddressedNo evidence for this area
1AcknowledgedConcern recognised, no design or evidence
2PartialSome requirements met, significant gaps
3Mostly AddressedMost requirements met, minor gaps (passing grade)
4Fully AddressedAll requirements met with evidence
5ExemplaryReference-quality — best practice

The overall score is the lowest individual section score, not an average. A solution with 5 in Performance but 1 in Security is a “1”, not a “3” — the Security gap is the concern.

14 sections: 1, 3.1–3.6, 4.1–4.5, 5, 6 (Executive Summary, the six Views, the five Quality Attributes, Lifecycle, Decision Making).

Organisation thresholds (suggested defaults)

Section titled “Organisation thresholds (suggested defaults)”
  • All sections ≥ 3 → production approval
  • All sections ≥ 4 → Tier 1/2 critical systems
  • Section = 0 or 1 → remediation plan required

DepthWhen to useTypical effortGovernance gate
MinimumPoC, dev/test systems, Tier 5 internal tools1–3 hoursDevelopment review
RecommendedProduction-bound systems, Tier 3–41–2 daysProduction approval
ComprehensiveRegulated, Tier 1–2 critical1–2 weeksEnterprise review
  • Tier 1 Critical → Comprehensive
  • Tier 2 High → Comprehensive
  • Tier 3 Medium → Recommended
  • Tier 4 Low → Recommended
  • Tier 5 Minimal → Minimum

Over-documenting a simple tool burns goodwill. Under-documenting a critical system hides risk. Calibrate to risk, not to tradition.


Where carbon footprint is actually decided

Section titled “Where carbon footprint is actually decided”

Most of a SAD’s environmental impact is locked in by a small number of decisions. Get these right and Section 4.5 becomes evidence rather than aspiration.

DecisionWhere in the SADQuick win
Cloud region3.3 Physical ViewPick a region with high renewable energy share. Carbon intensity varies 5–10× across regions of the same provider.
Non-prod runtime3.3 + 5.5 OperationsAuto-shutdown dev/test out of hours. Typical saving: 60–70% of non-prod compute cost and carbon.
Compute family3.3 Physical ViewARM (Graviton, Ampere, Cobalt) and latest-generation x86 deliver 20–40% better performance-per-watt than older instance types.
DR posture3.3 + 4.2 ReliabilityA warm standby running 24×7 doubles compute footprint. Match DR mode (cold / pilot light / warm / hot) to the actual RTO.
Data retention3.4 Data ViewSet a retention policy and automate expiry. Indefinite “keep just in case” is the most common waste.
Cold storage tiering3.4 Data ViewMove logs/snapshots/backups to archive tiers (S3 Glacier, Azure Archive). Cost and carbon both drop sharply.
Caching & async3.1 Logical ViewAvoid recomputation. Cache responses, debounce polls, replace busy-wait with events.
Right-sizing cadence5.5 OperationsQuarterly right-sizing using cloud advisor tools catches drift. Without a cadence, over-provisioning grows linearly.
PrincipleMeaningTypical action
Carbon efficiencyEmit the least carbon for a given outputRegion choice, ARM compute, retention
Energy efficiencyUse the least energy for a given taskCaching, async, code-level efficiency
Carbon awarenessDo more when the grid is clean, less when it isn’tTime-shift batch jobs, defer non-urgent work

A Tier 5 internal tool running on shared SaaS doesn’t need a Software Carbon Intensity baseline. Match the depth of treatment to the system’s actual footprint. A 50-instance e-commerce platform deserves the full Section 4.5 treatment; a fortnightly batch job on shared compute does not.



  • Use Ctrl/Cmd+P in your browser
  • Select “Save as PDF” or your preferred printer
  • Select “A4” paper size
  • The print stylesheet automatically hides navigation, sidebar, and theme controls
  • Each card above is designed to break to a new page, so one card per sheet

These cards are designed to sit alongside the full standard — not to replace it. Use them for reference during drafting and review; read the full standard pages for detail and context.