In this guide
- 01Plan stage learning
- 02Review the evidence
- 03Authorize a commitment
- 04Revisit conditions
Stage-Gate separates the work of developing an opportunity from the decisions to continue investing in it. A stage produces learning and deliverables; a gate examines those results and authorizes a next commitment, requests more work, pauses or stops. The distinction matters more than the appearance of a pipeline diagram.
Use the framework when projects compete for resources or move between commitments that need explicit review. Innovation management sets the wider context, while innovation budgeting explains why funding should follow a useful next question.
What the original approach says
Stage-Gate is associated with Robert Cooper’s work on new-product development. The model owner’s Discovery-to-Launch description includes discovery, scoping, business-case work, development, testing and validation, and launch. It describes gates through deliverables, criteria and outputs, including resource decisions.
This is the owner’s account of its model. Its promotional claims should be distinguished from independent evidence about performance. A company calling its process Stage-Gate may also implement something quite different from the described approach.
Cooper’s 2008 update explicitly discusses misconceptions about rigidity and linearity, together with governance and bureaucracy. A sensible assessment therefore asks how the gates operate, rather than assuming that all gated development is a fixed sequence with no iteration.
Separate stages from gates
Consider a hypothetical team developing a refill dispenser for an everyday household product. During a stage, it investigates dispensing behaviour, material compatibility, service requirements and customer understanding. Those activities may happen in parallel and may expose reasons to revisit the concept.

A gate asks what the evidence justifies next. A promising demonstration may support another bounded test while leaving manufacturing or user questions unresolved. The team should not need to pretend those questions are answered to obtain a sensible next commitment.
Keep three questions separate:
| Question | Where it belongs | Example |
|---|---|---|
| What work will reduce uncertainty? | Stage plan | Examine how the dispenser behaves under stated conditions |
| What evidence is needed for the next commitment? | Gate criteria | Identify unresolved issues relevant to the proposed trial |
| What will the organization authorize? | Gate output | Fund a limited trial, revise the proposal, pause or stop |
The table is a simplified teaching aid. It does not reproduce an official Stage-Gate template or certify the fictional project against a proprietary process.
Set criteria before the presentation
Agree what the gate is deciding and which evidence is proportionate. A first inquiry should not be evaluated as though its financial forecast were already a reliable sales plan. A launch commitment should not rely only on an enthusiastic concept interview.
NASA’s decision-analysis guidance offers a useful general discipline: clarify the decision, alternatives, criteria and rationale. Applied here, that means considering a smaller commitment or a revised question as alternatives to full approval and total rejection.
For the fictional dispenser, a gate before an operational trial might examine the task being tested, unresolved technical conditions, participant support, operating ownership and the resources needed. Appropriate specialist checks remain necessary for the actual product and materials. A management framework cannot establish product safety or regulatory suitability.
Use assumption mapping to identify the uncertainties that could undermine the next commitment. Define what must be understood now and what can reasonably remain open. Record the distinction so the next reviewer does not mistake deliberate uncertainty for a forgotten task.
Make the gate decision concrete
Invite people who can assess the relevant evidence and authorize the resources. A committee whose members can only recommend expenditure should make that limitation explicit. A project cannot proceed on an approval that nobody is able to honour.

Record one of several understandable outcomes. Continue means that a defined next stage has authority and resources. Rework means that the team must address stated questions before reconsideration. Hold means that a named condition prevents progress and someone will review it. Stop means that no further commitment is authorized, with the reasoning preserved.
Avoid a conditional approval with dozens of unowned actions. Separate conditions that must be satisfied before work begins from questions the next stage is authorized to investigate. If a condition concerns another department, obtain that department’s commitment instead of assigning it silently to the project team.
A short gate record can contain:
| Field | Required meaning |
|---|---|
| Decision | What is authorized, changed, paused or stopped |
| Evidence | The observations supporting that choice |
| Open questions | What remains uncertain and why it matters |
| Resources | People, time, facilities and spending authority |
| Owner | Who is accountable for the next commitment |
| Review trigger | A date or event that brings the decision back |
Use a decision memo to prepare this reasoning before the gate, then update it with the actual outcome. A record of what was decided is more useful than a slide showing that a meeting occurred.
Allow iteration inside a real commitment
Stage-level investment decisions and iterative development can coexist. The Agile Manifesto values responding to change; it does not by itself define a company’s portfolio governance. A team can revise a prototype within a bounded commitment while returning for authorization before a materially larger one.
Define the boundary. If the team discovers that the proposed dispenser requires a different manufacturing approach, that may change cost, timing or risk enough to require a new decision. Adjusting a research script or testing another design detail may remain within the authorized inquiry.
Strategyzer’s Test Card explanation separates the hypothesis, test, measure and decision threshold. That is useful within a stage, without implying that every product question can be settled through one metric. Experiment design explains how to match a test to what it can reasonably reveal.
Keep learning visible at the gate. If a trial contradicts an assumption, the team should be able to recommend a changed commitment. Treating every gate as a sales presentation can make the very evidence needed for a good decision harder to discuss.
Avoid bureaucracy and automatic continuation
One failure mode is a paperwork gate: every field is complete but the next resource decision is unclear. Another is a ceremonial gate: every project proceeds because stopping it is politically difficult. A third is excessive uniformity, where a small local service change receives the same demands as a major product launch.
Scale the review to the decision while preserving its substance. Remove duplicated reporting, clarify who decides and examine whether criteria discriminate between plausible options. Do not remove a difficult question simply because it slows approval; decide whether it is relevant to the commitment.
The Oslo Manual’s implementation distinction is helpful at the far end of the process. A launched offer or implemented process is different from a concept that passed internal reviews. Gate completion is a governance event, not evidence by itself that an innovation delivers value in use.
Evidence limits and the first gate to repair
Much of the explanatory material here comes from model owners and methodological guidance. It establishes the intended approach, not a universal profitability effect or an optimal number of gates. The dispenser narrative is hypothetical; its table and decision record are original adaptations.
Choose one existing gate and inspect its last decision. Identify what evidence was reviewed, what resources were committed and who owned the next step. If any of those is missing, repair that decision before redesigning the whole pipeline. The resulting practice can then be reviewed through continuous improvement.
Sources:
- Stage-Gate International (n.d.). Stage-Gate Discovery-to-Launch Process.
- Robert G. Cooper (2008). The Stage-Gate Idea-to-Launch Process: Update, What’s New, and NexGen Systems.
- Kent Beck and other Agile Manifesto signatories (2001). Manifesto for Agile Software Development.
- NASA (n.d.). 6.8 Decision Analysis.
- Strategyzer (n.d.). Validate your ideas with The Test Card.
- OECD and Eurostat (2018). Oslo Manual 2018.



