How Many Funnel Stages Do You Need? One Gate per Real Commitment15 min read

The funnel redesign meeting is an hour old and someone has just proposed a sixth stage. Nobody in the room can say what it would decide.

The argument then splits four ways:

• Keep the five-stage model, because that’s the standard.
• Add a stage every time something slips through that shouldn’t have.
• Cut back to three, since the last process was too slow and everyone still remembers it.
• Leave it as configured in whatever tool you bought.

None of the four starts from the decisions the funnel actually makes, only from a memory, a diagram, or a default.

Count the decisions that authorize, withhold, redirect, or stop a tranche of resources, and you’ve counted your gates. The idea-funnel list later uses four gates and five stages.

Too many stages add review waits and invite rubber-stamping. Too few empowered decisions let weak projects cross major commitments.

What a Stage Is Actually For

A stage groups work that reduces a defined set of uncertainties before the next decision. A gate authorizes, withholds, redirects, or stops work or resources.

The U.S. Department of Energy’s 2007 program guidelines list four gate functions:

• A set of criteria judges the project’s progress.
• The project goes ahead, gets delayed, or gets stopped.
• Funding is approved for the next stage.
• A path forward for the next stage is presented and approved.

Three of those four produce a decision: whether work continues, whether funding is released, and which path is approved. The criteria define the threshold.

A review without authority to make a binding gate decision is a checkpoint wearing a gate’s clothes. Checkpoints belong inside a stage rather than between stages.

Where the Five-Stage Default Comes From

A 2020 review of Stage-Gate says the original model starts with the ideation stage, Discovery, and culminates in the Post-Launch review.

Stage and gate counts use different conventions. This article counts intake as Stage 1 and counts only the decisions between consecutive stages, so five stages produce four gates.

The same review describes context-based approaches that skip gates and stages to cope with different degrees of complexity.

So five is a common template, not a universal requirement. Applying the full template to a minor process change can make a funnel heavy.

What Too Many Stages Actually Costs

A new gate creates one more place where work waits for review. The queue-health model matters when proposals arrive faster than reviewers can decide.

Kellogg School of Management’s 2015 piece on controlling queues illustrates how quickly a single-server wait can grow:

• The illustrated wait more than doubles as utilization rises to 80%.
• Going from 80% to 90% doubles it again.

Gate panels aren’t solitary servers, so treat those numbers as a capacity warning rather than a forecast. Waiting still rises as reviewer utilization approaches full.

An internal checkpoint is cheaper only when it avoids the same approval path and reviewer delay. Expect review load to rise when added gates share reviewers.

Harvard Business Review’s 2016 collaborative overload piece reported time in collaborative activities ballooning by 50% or more over the prior two decades.

Reporting Layers Are a Warning, Not a Gate Count

McKinsey’s 2019 report on a 2018 decision-making survey compared speed and quality across reporting layer counts:

These are organizational reporting layers, not funnel stages. The survey shows an association, so it can’t predict the effect of adding one gate.

Use it as a warning when a redesign also adds hierarchical approvals. A separate question found 61 percent judging most decision time ineffective.

For managers at an average Fortune 500 company, McKinsey’s thought experiment puts ineffective decision time at roughly $250 million in annual labor cost.

That cost is not all gate time. Look for routing delay in the approval workflow as well as in the gate itself.

When Gates Stop Deciding

McKinsey’s 2017 decision-gate article says meetings at most organizations too often become product-maturity reviews with limited decision focus.

It lists three related patterns:

• Gates read as process formalities that lack teeth.
• Funding is rarely delayed or halted at a traditional stage gate.
• Serious project issues don’t become evident until late in development.

A gate that never changes work, resources, or direction is a recurring status meeting with a calendar invite attached.

Count how many times a gate in your funnel changed an outcome last year. That count tells you more than the number of gates does.

Complexity Has Its Own Price

The same pattern turns up outside the innovation process itself. Bain’s 2016 report on growth barriers found 85% of executives citing internal barriers as the primary obstacles to growth.

Two of the four forces Bain ties specifically to stall-out:

• Unchecked complexity that kills growth, named by 42%.
• Bureaucratization, named by 41%.

When stages fill without decisions, backlogs and stale work accumulate. The linked article attributes that pattern to unclear decision rights and uncapped dwell time.

Trott and colleagues argue in a Prometheus critique that high uncertainty requires the flexibility to change fundamental elements of a project, including the underlying concept and the target market.

Their conclusion is that stage-gate isn’t well suited to innovation processes addressing these contemporary challenges. Rigid staging assumes the concept will hold still while the stages run.

What Missing Stop Authority Costs

Stage count alone doesn’t determine whether a project can be stopped. The risk is crossing a major commitment without an empowered go, hold, or stop decision.

A three-stage process with gates only between stages has two interstage gates. Their timing determines whether they arrive before major spend.

One Reason Projects Don’t Stop on Their Own

Barry Staw’s 1976 experiment put 240 undergraduate business students through a simulated investment decision, manipulating personal responsibility and positive versus negative results.

The high-responsibility, negative-results condition produced the largest second allocation. Responsibility didn’t significantly change allocations after positive results.

Negative results also didn’t significantly change allocations under low responsibility. The experiment suggests separating start and continuation reviewers after negative results, but it didn’t test that design.

What It Costs to Find Out Late

NASA’s Johnson Space Center reports Boehm’s relative cost-to-fix figures for a software problem found at four successive phases:

To be fair, that’s software engineering data and not a study of innovation funnels. It’s here for the shape of the curve and nothing more.

The same paper quotes a line about finding and fixing a problem after delivery costing upwards of 100 times more.

Expect a funnel with no working stop to surface problems later. Whether that costs more in your business is worth measuring, not assuming.

A separate 1999 Technische Universität Hamburg-Harburg paper summarizes Cooper and Kleinschmidt’s evidence on underfunded pre-development work.

Pre-development activities, also called the fuzzy front end, received only 6% of dollars and 16% of man-days. Execution quality there differed strongly between winners and losers.

When Nothing Stops It

Two documented cases run the same arc, one in consumer electronics and one in state government:

Isabelle Royer’s account of SelectaVision is the better documented of the two. Fourteen more years of resources went in after that first prototype.

RCA kept building new models and investing in production capacity. The 1981 launch drew a tepid response and changed nothing.

Keil and Montealegre, who documented SACSS, describe executives who become so wedded to a project that they continue when they should pull out.

Separate recommendation from kill authority. In that split the venture lead recommends, and the venture board decides using pre-agreed kill criteria.

What a Real Kill Recovers

A 2019 McKinsey case describes a global food producer that designated a full-time project killer.

Over three years, its portfolio fell from more than 560 projects to just over 200. That left roughly 360 fewer projects competing for resources.

McKinsey’s 2025 article describes a long tail of low-ROI initiatives soaking up resources.

One consumer-electronics company rebalanced its portfolio, freeing 20 percent of the budget. The exercise identified opportunities for a 10 percent portfolio-level ROI increase.

A portfolio-pruning framework makes zombie-project costs visible and supports evidence-based decisions about what to fund or prune.

Count the Commitments Before You Count the Stages

There’s a number floating around for how many decision points a project should have. It works as a rough sanity check and not much more.

McKinsey’s 2017 piece recommends identifying five to ten critical management decisions over the life of a project, from feasibility assessment through launch.

No performance data is attached to that range in the article. Treat it as expert guidance rather than a tested optimum.

It calls those five to ten decisions decision gates. The same piece separately recommends two to five repeated component decisions and one to three progress reviews.

That product-development design doesn’t validate a four-gate idea funnel. Use it as a contrasting benchmark rather than a target to copy.

Run the Audit First

Walk your current funnel end to end. Mark each point where a tranche of work or resources is authorized, withheld, redirected, or stopped.

Also mark external commitments that become hard to reverse.

A signed supplier contract counts, and so does a hire against a new headcount line. Tooling orders and publicly announced dates belong on the same list.

Then run each candidate stage through three questions:

1. Which uncertainties does this stage reduce, and what evidence will show that?
2. Which decision does its gate make?
3. What changes if the answer is no?

The third question helps distinguish a gate from a checkpoint. If nothing changes on a no, you’re looking at a checkpoint.

Checkpoints are worth having, and the audit doesn’t ask you to remove any. It asks you to stop turning them into stage boundaries.

Moving a review inside a stage removes a formal boundary only when it doesn’t recreate the same approval path or reviewer queue.

Expect the audit to leave you with fewer stages than the tool you bought came configured with.

What Survives the Audit in an Idea Funnel

Run that audit on your idea funnel and you’ll probably land close to the list below.

Expect roughly four moments where a decision has to be made:

• An intake screen that authorizes discovery
• An evidence gate that authorizes validation or pilot work
• A build gate that commits engineering or development resources
• A scale gate that commits launch, hiring, manufacturing, or market spend

The last three commit hard resources, while the intake screen releases or refuses discovery capacity. Those four gates separate intake, discovery, validation, build, and scale.

Those counts are outputs of the audit, not a template you start from. Lighter work can skip gates, as the lanes below show.

Running this audit once is easy. Defending the result a year or two later is the harder half.

Keep a record of each gate decision, including its owner, evidence, rationale, and revisit trigger. Software can preserve that record with the initiative.

AI can surface where reviewers diverge on the same idea and flag inconsistent scoring against shared anchors. People at the gate still make the actual call.

Scale the Count to the Risk

One count for the whole portfolio is the wrong shape for the problem. A minor packaging change and a new platform don’t carry the same risk.

A 2017 Portland State University paper reviews Stage-Gate variants and an Agile hybrid for the fuzzy front end.

A single fixed funnel ignores differences in complexity and uncertainty. Use lanes for work suited to gated governance, and discovery before lane assignment for work that isn’t.

For governed work, set a number per lane, and let idea type and risk shape its rules. The whole portfolio doesn’t need one count.

Set the Lanes Before the Gates

Here’s a starting shape, built from the sourced material above plus judgment rather than any single published tier definition.

Highly ambiguous work does not fit a predefined gate sequence. Use iterative discovery before assigning it to one of the lanes below.

Route new-platform or new-market work to Full only after discovery defines its core assumptions.

When signals conflict, use the highest applicable lane.

The trigger column is the one you have to write yourself. Set thresholds against your balance sheet, risk appetite, regulatory exposure, and strategic commitments.

A 2009 tollgate methodology paper in the Project Management Institute’s library says the effort invested has to be proportional to the size and relevance of the project.

The same paper recommends a selection matrix for organizations running different project sizes. That matrix is what sets the gate count for each size.

Let Thresholds Do the Sorting

The UK’s Gateway Review process, since renamed the Gate Review Process, showed the logic. The original OGC model listed six reviews, although Strategic Assessment applied only to programs.

Review was typically required above a prescribed risk threshold, which combined four things:

• Upfront expenditure
• Lifetime cost
• Degree of special interest to the government
• Risk

The Department of Energy overview mentioned earlier went further, since projects could be initiated at whatever stage suited them and funding could end wherever it fit.

That’s a funnel with configurable entry and exit rather than a fixed conveyor. DOE counted five work stages and four interstage gates.

The conventional model adds a gate before its first formal stage. That’s a counting difference, not evidence that either arrangement is better.

Iterate Inside the Stage

Lanes decide how much evidence a gate demands, and iteration decides how that evidence gets produced.

Cooper now describes an Agile Stage-Gate hybrid, adding Agile elements while retaining structure and rigor. He calls the result the Triple A System: adaptive and flexible, agile, and accelerated.

The 2020 review cited earlier reports that 44.8% of best-performing firms run build, test, feedback and revise iterations, against 26.3% of average-performing firms.

The review calls this first evidence and says statistical proof remains rare. It also warns that short-sighted customer feedback can favor incremental innovation.

The association doesn’t show that iteration caused stronger performance. Use iteration as an evidence-generating practice inside a stage.

Set the Count, Then Defend It

The redesign is easy compared with what comes after it, because you should expect someone to ask for a stage back inside a quarter.

Three moves for the next redesign:

• Publish the commitment list before you draw a single box.
• Give each gate a named owner who can refuse or redirect the next tranche of work or resources.
• Re-run the audit whenever a threshold or a funding rule changes.

A published audit gives reviewers a rationale they can inspect. A named gate owner makes that rationale enforceable rather than advisory.

A gate without someone who can say no is a scheduled meeting.

Count the commitments and test whether four gates and five stages fit your idea funnel. Let anyone who wants a sixth name the decision it makes.

Download our ebook on idea evaluation and governance, or book a demo to see how Accept Mission structures idea funnels around real decision points.