The agile-vs-waterfall debate is mostly noise. Real projects choose based on three questions, and many end up — correctly — with a hybrid.
The three questions
- How stable are the requirements? If scope can genuinely be discovered iteratively (software, product design), iteration adds value. If the deliverable is fixed by contract and physics (a bridge, a refinery), pretending to discover it in sprints is theatre.
- What does change cost? In software, changing course costs a sprint. In construction, changing a foundation design after pour costs months and crores. High cost of change → invest in up-front definition (waterfall's whole point).
- What does the regulator or contract require? Stage gates, design approvals, and EPC contract structures impose sequence regardless of methodology preference.
What each does well
Waterfall / predictive delivers when scope is definable up front: clear baseline, meaningful EVM, contractual clarity, and a schedule you can hold parties to. Its failure mode: discovering in month 18 that the requirement was wrong in month 2.
Agile delivers when value can ship incrementally: early feedback, re-prioritisation, and no big-bang integration risk. Its failure modes: scope-creep dressed as "responding to change", and weak long-range cost/schedule visibility — which is exactly what sponsors and CFOs need.
Hybrids that actually work
- Stage-gated shell, agile inside. Fixed gates for funding and design approval; iterative delivery between the gates. Common in banking and pharma IT.
- Predictive structure, agile workstreams. A construction megaproject scheduled in P6, with the digital/BIM and commissioning-software workstreams run in sprints, integrated at milestones.
- Rolling-wave planning. Detailed planning for the next 90 days, planning packages beyond — predictive in shape, adaptive in detail. Native to both P6 practice and PMBOK.
Making the hybrid controllable
The hard part is progress measurement across methods. Practical answer: map sprints to work packages in the WBS, earn value on completed, accepted increments only (0/100 per sprint outcome), and roll it into the same S-curve as the predictive work. Sponsors get one integrated picture; teams keep their working rhythm.
The honest test
If your "agile transformation" still has a fixed scope, fixed date and fixed price — you've chosen waterfall with extra meetings. And if your "detailed 3-year baseline" is re-planned every quarter — you're doing rolling-wave; formalise it and stop calling the variance a failure.
Explore delivery-approach content in the Knowledge Hub, or compare tooling in the Software Hub.