Topic 1 · System Fundamentals

Hardware = physical (chip, screen). Software = invisible instructions (app, OS).

Planning → Analysis → Design → Implementation → Testing → Deployment → Maintenance. Memory: P-A-D-I-T-D-M.

Feasibility study, rough budget, risks. Cheapest stage — biggest leverage.

Capture requirements from real stakeholders. Skipping this = building the wrong thing.

Architecture diagrams + UI mockups + data models. Fixing a flaw here is 10–100× cheaper than fixing it after launch.

Actual coding. Usually only ~30% of total project time.

Unit → Integration → System → User Acceptance Testing (UAT). UAT = real users approve.

Move new system live + migrate data + train users + plan rollback.

Corrective (bugs), Adaptive (env changes), Perfective (new features), Preventive (tidy code). Usually >50% of lifetime cost.

Linear, doc-heavy; each stage finishes before next. Best for stable + regulated (medical, aerospace).

Short sprints; working code each cycle. Best for uncertain + fast-feedback (apps, startups).

Most real projects mix both — Agile day-to-day, Waterfall for major releases.

Adding people to a late project makes it later. Communication overhead explodes.

Implementation ≠ Deployment. Implementation = coding; Deployment = rolling out.

Real users + real data + real load behave in ways tests cannot fully simulate.

Switch everyone overnight. Fastest, cheapest, riskiest. Only for small, bounded projects.

Run BOTH systems; compare outputs. Safest, most expensive. Use for mission-critical (payroll, tax, banking).

Roll out to a small group first → expand. Best when org has natural sub-units (one school, one ward, one branch).

Roll out one module OR region at a time. Best when system is modular or geographic.

Moving old data to new system. Hardest invisible work — old data is messier than expected.

Workshops + docs + videos + in-app help + super-users. Underbudgeting training = failed adoption.

Pre-planned UNDO if go-live fails. Triggers + mechanism + data handling + time budget.

Involve users early · train BEFORE launch · communicate WHY · provide fallback. People fear looking incompetent.

Big Bang migration disaster — 1.9M users lost access, CEO resigned, £330M compensation. Lesson: never Big Bang critical systems.

End users · clients · developers · project managers · ops/IT · external (regulators, public). Client ≠ end user.

What the system DOES. 'A user can reset their password.' Specific actions.

How WELL it does it. Performance, scalability, security, usability, '-ility' words.

Specific, Measurable, Testable. 'Fast' is bad; '<1 s on 4G' is good.

Interview (depth) · Survey (breadth) · Observation (hidden patterns) · Document analysis (legacy rules).