PM Methodologies

Agile vs Waterfall: Comparison Table, Origins, Hybrid

By Vact Published

Use waterfall (a predictive, plan-driven life cycle) when scope, regulation or contract terms can be fixed before work starts and changing course late is expensive. Use agile when the right answer will only emerge from users or from the work itself, so the plan has to be revisable every few weeks. Most real projects sit between the two. That is why PMI’s PMBOK Guide, in both the Seventh Edition (2021) and the Eighth Edition (2025) that is current as of November 2026, treats predictive, adaptive and hybrid as development approaches to select and then tailor to the project.

Waterfall (predictive)Agile (adaptive)
Core ideaPlan the whole scope first, then build in phasesDeliver in short cycles and adjust the plan from feedback
OriginUsually traced to Winston Royce’s 1970 paper, which never uses the word “waterfall”Manifesto for Agile Software Development, February 2001
Typical cycleOne pass through phases, or a few large releasesIterations of one month or less (Scrum)
ChangeControlled through a formal change processExpected; the backlog is reordered as priorities change
Best fitStable requirements, fixed-price or regulated workUncertain requirements, products used and revised by customers

Where waterfall came from: Royce, 1970

Winston W. Royce presented “Managing the Development of Large Software Systems” at the IEEE WESCON conference in August 1970; the quotations below come from a scanned reprint of the paper. The paper is the usual citation for the sequential model, although the word “waterfall” appears nowhere in it. Its Figure 2 runs from system requirements through software requirements, analysis, program design, coding and testing to operations. Figure 3 adds iteration “with the preceding and succeeding steps but rarely with the more remote steps in the sequence.”

Royce’s verdict on this scheme is widely quoted: “I believe in this concept, but the implementation described above is risky and invites failure.” His reason was testing. It comes at the end, and it is “the first event for which timing, storage, input/output transfers, etc., are experienced as distinguished from analyzed.” If those fail to satisfy the system’s external constraints, “invariably a major redesign is required,” and “one can expect up to a 100-percent overrun in schedule and/or costs.” The caption of his Figure 4 states the problem in one line: “the design iterations are never confined to the successive steps.” Even so, he judged “the illustrated approach to be fundamentally sound” and set out “five additional features that must be added to this basic approach to eliminate most of the development risks.” Laurent Bossavit’s close reading of the paper infers from where the sentences sit that “risky” refers to Figure 2, the cascade with no feedback loops, and “fundamentally sound” to Figure 3.

The five features, under Royce’s own step headings:

  1. Program design comes first. A preliminary program design phase sits between software requirements and analysis, so that the designer “assures that the software will not fail because of storage, timing, and data flux reasons.”
  2. Document the design. Asked how much documentation, Royce answered “quite a lot” and called “ruthless enforcement of documentation requirements” the first rule of managing software development. His Figure 6 requires at least six documents, from the software requirements to the operating instructions.
  3. Do it twice. For a program “being developed for the first time,” he advised that “the version finally delivered to the customer for operational deployment is actually the second version insofar as critical design/operations areas are concerned.” A 30-month effort might give 10 months to the pilot model; a 12-month effort perhaps three.
  4. Plan, control and monitor testing. Testing is “the phase of greatest risk in terms of dollars and schedule.” Royce wanted test specialists, a second person to scan every piece of analysis and code by eye, and every logic path tested “at least once.”
  5. Involve the customer. The customer should be involved “in a formal way so that he has committed himself at earlier points before final delivery.” Figure 9 places three reviews after requirements definition: a preliminary software review, a critical software review during program design and a final software acceptance review.

Read in full, the paper keeps the sequence and treats a single unguarded pass as the risk; the five features are risk controls added to it. Royce conceded that each “costs some additional sum of money,” but wrote that in his experience “the simpler method has never worked on large software development efforts.” Craig Larman and Victor Basili’s history of iterative development in IEEE Computer (2003) says many people “incorrectly” view the paper as “the paragon of single-pass waterfall,” and that its feedback-based step “has been lost in most descriptions of this model.” They also quote Royce’s son, Walker Royce, describing his father as “always a proponent of iterative, incremental, evolutionary development.”

Where agile came from: the 2001 Manifesto

The Manifesto for Agile Software Development came out of a meeting of 17 people at the Snowbird ski resort in Utah on 11-13 February 2001, according to the manifesto site’s history page. The signatories include Kent Beck, Martin Fowler, Robert C. Martin, and Ken Schwaber and Jeff Sutherland, who had developed Scrum. Its four value statements read:

  • Individuals and interactions over processes and tools
  • Working software over comprehensive documentation
  • Customer collaboration over contract negotiation
  • Responding to change over following a plan

The closing sentence matters for every comparison: “while there is value in the items on the right, we value the items on the left more.” Plans, documents, contracts and tools remain in scope; they rank second.

The twelve principles turn the values into working rules. The ones that most separate agile from a plan-driven life cycle are:

  • Welcome changing requirements, “even late in development” (principle 2).
  • Deliver working software frequently, “from a couple of weeks to a couple of months, with a preference to the shorter timescale” (principle 3).
  • Business people and developers “must work together daily throughout the project” (principle 4).
  • “Working software is the primary measure of progress” (principle 7).
  • Sponsors, developers and users “should be able to maintain a constant pace indefinitely” (principle 8).
  • “At regular intervals, the team reflects on how to become more effective” (principle 12).

Scrum in brief: team, events and artifacts

The Scrum Guide (Ken Schwaber and Jeff Sutherland, November 2020) defines Scrum as “a lightweight framework”. A Scrum Team has one Scrum Master, one Product Owner and Developers, typically 10 or fewer people in all. Work happens in Sprints of one month or less. The Daily Scrum is a 15-minute event; for a one-month Sprint, Sprint Planning is capped at eight hours, the Sprint Review at four hours and the Sprint Retrospective at three, and shorter Sprints usually have shorter events. The three artifacts are the Product Backlog (with its Product Goal), the Sprint Backlog (with the Sprint Goal) and the Increment (with the Definition of Done). An Increment may be delivered to stakeholders before the Sprint ends, so the Sprint Review is not a release gate.

Scrum is one approach among several. The Agile Practice Guide that PMI and Agile Alliance created together in 2017 also describes Extreme Programming (XP), Kanban, Feature-Driven Development, the Scaled Agile Framework and Large Scale Scrum, among others.

Agile vs waterfall comparison table

The table is an original synthesis for planning decisions. Entries that name Royce, the Manifesto, the Scrum Guide or GAO follow those sources; the rest describe typical practice.

DimensionWaterfall (plan-driven)Agile (Scrum as the reference)
PlanningDetailed plan for the full scope up front: a work breakdown structure, schedule and budget baselineProduct Goal and an ordered Product Backlog; detailed planning only for the next Sprint
Scope and change controlScope baselined; changes go through a formal change request with cost and schedule impactThe Product Owner orders the backlog and can reorder it as priorities change; during a Sprint, no changes are made that would endanger the Sprint Goal
Delivery cadencePhases end in handoffs; the user sees working product at or near the end, or at major releasesA usable Increment every Sprint (one month or less), which can be released before the Sprint ends
DocumentationHeavy and formal; Royce required at least six documents, from software requirements to operating instructionsEnough to build and maintain; working software outranks comprehensive documentation
Customer involvementFormal reviews at set points, like Royce’s preliminary, critical and final acceptance reviewsDaily collaboration between business people and developers (principle 4), plus the Sprint Review
RolesProject manager, functional leads, specialist teams by phaseProduct Owner, Scrum Master and Developers in one cross-functional team
RiskManaged through analysis up front and a maintained risk register; late discovery is the main exposureReduced by short feedback loops; risks surface as work is demonstrated; a risk register still helps on larger efforts
ContractsFits fixed scope and price, because the scope is defined before signingFits time-and-materials, capped or incremental funding; in US federal IT, FITARA has CIOs at covered agencies certify that certain investments adequately implement incremental development, per GAO
MetricsProgress against baseline: schedule and cost variance, milestones, critical path floatWorking software as the primary progress measure; Sprint Goal outcomes and the Definition of Done
Team stabilityStaff move between phasesPersistent team; constant pace (principle 8)
Failure modeBuilding the wrong thing precisely, found lateDrift without architecture, scope never converging, or a backlog with no real Product Owner

When each approach fits

These fits are practical rules of thumb that follow from how each approach handles change and evidence.

Waterfall fits

  • Building construction and fit-out. Physical work is hard to redo, drawings and permits must be approved before building, and trades follow a dependency order. A critical path schedule is the natural control tool.
  • Regulated and safety-critical delivery such as medical device or aviation software, where evidence of design, verification and approval has to exist before release. The kind of documentation Royce demanded doubles as the audit trail.
  • Fixed-price contracts with a defined specification, such as a data-centre migration with a known inventory.
  • Compliance and infrastructure deadlines where the end date comes from outside (an election, a legal effective date) and the required scope is fixed as well, so the work is mostly sequencing known tasks.

Agile fits

  • New software products and features, where users cannot specify what they want until they see something. Short cycles convert guesses into evidence.
  • Services with continuous improvement, such as an internal tool used by a support team that sends requests every week.
  • R&D and prototypes, where the plan for month four depends on what month two teaches.
  • Projects where the customer can staff a Product Owner who decides daily. Without that person, the Manifesto’s customer collaboration has nobody to attach to.

Neither wins outright

Most projects mix certainty with uncertainty. A hospital IT programme has a fixed go-live date and a regulated data migration (plan-driven) alongside clinician-facing screens whose design needs iteration (agile). That mixed profile is where hybrid approaches belong.

Hybrid agile-waterfall approaches

PMI treats the choice as a spectrum. Its FAQ on the PMBOK Guide Seventh Edition says the guide’s performance domains “reflect the full spectrum of approaches from predictive to adaptive,” that the 12 project management principles in its standard are “not specifically agile principles” but “universal regardless of approach,” and that the standard and guide “do not prescribe a particular approach.” The Eighth Edition (2025), current as of November 2026, has six principles and seven performance domains, gives predictive, adaptive and hybrid development approaches a section each, and keeps a chapter on tailoring.

PMI’s tailoring explainer for the Seventh Edition sets out four steps, and the Eighth Edition’s tailoring chapter keeps the same four headings:

  1. Select an initial development approach (predictive, adaptive or hybrid).
  2. Tailor for the organisation, considering governance, key reviews, quality assurance, policy compliance and PMO or VDO approval.
  3. Tailor for the project, adding, removing or modifying aspects based on size, criticality and other factors.
  4. Implement ongoing improvement through retrospectives, lessons learned and efficiency improvements.

The 2017 Agile Practice Guide, as previewed by Agile Alliance, includes suitability filters, tailoring guidelines and a model “for assessing the suitability of Agile, hybrid, and predictive approaches.” Its second edition, also created by Agile Alliance and PMI, covers “continuum and hybrid models” and “fit-for-purpose approach selection.” Because the mix of agile and predictive elements differs from project to project, a team should write down what “hybrid” means for its own project.

Patterns for combining stage gates with sprints

The four arrangements below are practical patterns for combining the two, drawn up for this comparison.

PatternHow it worksSuits
Plan-driven frame, agile buildA fixed initiation, budget approval and release gate surround sprints that build the productCorporate software with a funding cycle and a firm launch date
Agile front, predictive deliveryShort discovery sprints settle requirements; once scope is stable, a baseline plan governs construction or rolloutSystems with design questions to settle first, then a large deployment
Parallel tracksSoftware runs in sprints while hardware, facilities or legal work follows a schedule; milestones link the tracksProduct launches with physical components
Gated incrementsEach gate reviews an increment’s evidence (tests passed, demo accepted) instead of a documentRegulated environments that want agile delivery

Three practical rules keep a hybrid from becoming the worst of both. Fix the interface between tracks, which is usually a milestone list from milestone planning. Make gate criteria testable outcomes. Keep one authority for scope per track so change control does not run twice.

Government buyers face the same mix. GAO’s Agile Assessment Guide, released on 28 September 2020, cites the Federal Information Technology Acquisition Reform Act (FITARA) of 2014, which requires CIOs at covered agencies to certify that certain IT investments are adequately implementing incremental development. GAO names Agile as one framework for incremental development.

Common misconceptions, corrected from the sources

“Waterfall means one pass with no feedback.” Royce’s paper shows iteration between successive phases in Figure 3, warns that in practice design iterations “are never confined to the successive steps,” and adds a “do it twice” pilot and formal customer reviews. Larman and Basili call the single-pass reading of the paper incorrect.

“Agile means no documentation.” The Manifesto states that there is “value in the items on the right,” which includes comprehensive documentation. The Scrum Guide requires a Definition of Done and a Product Backlog. The ranking changes; the documents stay.

“Agile means no planning.” Scrum has a Product Goal, a Sprint Goal and an eight-hour maximum Sprint Planning event for a one-month Sprint. Planning happens in smaller batches and more often.

“Agile is a fixed process you install once.” Under principle 12 the team reflects at regular intervals, “then tunes and adjusts its behavior accordingly,” and the Scrum Guide time-boxes the Sprint Retrospective at up to three hours for a one-month Sprint.

“Scrum equals agile.” Scrum is older than the Manifesto: the Scrum Guide says Schwaber and Sutherland developed it in the early 1990s and first presented it at the OOPSLA conference in 1995, and both later signed the Manifesto. The Manifesto names values and principles, not a process, and the Agile Practice Guide describes XP, Kanban and other approaches alongside Scrum.

“PMI’s PMBOK Guide is waterfall.” PMI’s FAQ on the Seventh Edition calls its principles “universal regardless of approach,” and the Eighth Edition gives predictive, adaptive and hybrid approaches a section each.

“A Sprint Review is the release gate.” The Scrum Guide says an Increment may be delivered to stakeholders before the end of the Sprint and that the Sprint Review “should never be considered a gate to releasing value.”

Decision checklist

Score each statement for your project. Count the ticks in each column.

StatementTick for waterfallTick for agile
Requirements can be written and signed off before work startsYesNo
The cost of changing course after the design is done is very high (physical build, certified hardware)YesNo
An outside party requires fixed price, scope or documented approvalsYesNo
Users will only know what they want after seeing working outputNoYes
A decision-maker from the customer side can answer questions dailyNoYes
The team can release something usable every month or soonerNoYes
The end date is fixed by an external event and scope can flexEitherYes
Regulators want design evidence before releaseYesAdapt as hybrid

How to read the result: a clear majority on one side points to that approach. A split result points to a hybrid; draw the boundary where certainty changes, give each track its own cadence, and follow PMI’s tailoring steps, starting with the selection of an initial approach and finishing with retrospectives. Whichever approach wins, a work breakdown structure still defines the deliverables, and a risk register still tracks the things that can go wrong.

Sources