FORGE
Product Lifecycle Framework

FORGE

Most frameworks protect the organization from bad products.

This one protects good products from the organization.

Joe Nalley
FORGE IN THREE MINUTES

Two lanes. Pipeline ships products through six phases with two council gates -- concept and launch. Furnace explores ideas that don't have a business case yet through six modes with fixed clocks. Both run simultaneously. Neither contaminates the other.

Fifteen structural mechanisms replace culture as the governance layer. Named decision authority with 5-day clocks. Time-boxed disagreement. Blind pre-council ballots. Grief protocol for teams after correct kills. Annual zombie detection for products that stopped earning their place.

Portfolio capacity is finite and enforced. Three Full products in Build or Prove at any time. When the pipeline is full, new products queue. One person owns the queue.

The dual-lane design comes from the Pendulum -- the swing is structural, not behavioral. You can't coach your way out of it. FORGE doesn't pick a side. It builds for both positions and lets the architecture hold when the politics shift.

The framework includes a self-diagnostic that catches its own drift. Every system calcifies. This one is built to notice.

Why This Exists

Most products don't die from bad ideas. They die because the organization starves them -- committees that won't decide, leaders who won't kill what isn't working.

Every mechanism in this framework exists because something specific broke without it. The phases, the clocks, the kill criteria -- none of it is theoretical.

What Breaks

Committees that distribute accountability until it disappears. Timelines without clocks. Products that nobody will kill and nobody will fund. Frameworks that pick rigor or agility and break when the organization swings the other way.

The Architecture

Two lanes. Pipeline ships products through six phases with two council gates. Furnace explores ideas without a business case. Both run simultaneously. Neither contaminates the other.

The Discipline

Time-boxed phases. Named decision authority. Kill criteria written before work begins. Fifteen structural mechanisms replace culture as the governance layer.

6 lifecycle phases
2 council gates
15 people mechanisms
7 product archetypes
6 Furnace modes
What are you building?
Full Product
6 phases · 26-40 weeks · 2 council gates
New market entry, net-new capability, or platform build. Full lifecycle, full discipline.
Function Change
5 phases · 14-20 weeks · 2 council gates
New capability on an existing product. Compressed thesis, compressed launch.
Update
3 phases · 4-7 weeks · No council
Enhancement, fix, or iteration. Product owner decides. Clock enforces.
Integration
4 phases · 16-24 weeks · 1 council gate
M&A product absorption. The acquisition was Gate 1. Build, prove, launch, sustain.

Select a type above, or jump straight in -- the console is pre-loaded with a Payer De Novo configuration.

Configuration

Build the Path

Pick your constraints. The engine builds the path: phases, deliverables, kill criteria, governance.

Presets
Product Type
Org Structure
Risk Tolerance
Timeline
Parallel Furnace
Blueprint
Parallel Track: Furnace
Interactive Tools

Four tools built into FORGE. Configure a blueprint above, or use these directly.

Blueprint Generator
Configure product type, org structure, risk tolerance, and timeline. The engine builds the path: phases, deliverables, kill criteria, governance.
Health Diagnostic
Five questions scored automatically. Find out whether FORGE is operating as designed or drifting toward calcification.
Kill Criteria Generator
Test whether your kill criterion is specific and measurable enough to be binding. Validates against FORGE quality checks.
Archetype Decision Tree
Answer five questions. The first yes determines the archetype: De Novo, Partnership, Platform, Regulatory, Sustain, Exploration, or Acquisition.
One Product Through the System

A payer's maternity product went from signal to launch in 29 weeks with FORGE. The same organization attempted the same product the year before. It took 14 months and never launched.

Read the full walkthrough →
The System

What Holds It Up

FORGE has three load-bearing systems. Remove one and the others fail. The lifecycle defines what happens and when. The people mechanisms define who decides and how disagreements resolve. The portfolio layer defines how many products can run at once and what happens when the pipeline is full. Most PLMs nail the lifecycle and leave the other two to culture. Culture does not survive a reorg.

The Furnace
Exploration with protection. The Pipeline handles what you know. The Furnace handles what you don't know yet. The three systems below are what hold it up.
Lifecycle
Phases, gates, handoffs, kill criteria, time boxes
People
Decision authority, disagreement protocol, escalation
Portfolio
How many products can run at once. Capacity limits, prioritization rules, resource contention. The constraint layer.
Full specification → Manual Section 07: The Three-Legged Stool
Leg 1: Lifecycle

How the Lifecycle Works

The console builds a configured path. This section is the complete reference: every phase, gate, handoff, and role.

Lifecycle Flow
Signal
1-2 weeks
Detect the signal. Is this worth investigating?
Checkpoint
Thesis
4-6 weeks
Build the structural case. Name the buyer, the price, the kill conditions.
Council Gate 1
Build
8-12 weeks
Build the minimum version that generates real market data.
Checkpoint
Prove
8-12 weeks
Put it in front of buyers. Measure what happens.
Checkpoint
Launch
4-6 weeks
Full operational readiness. The council's second and final look.
Council Gate 2
Sustain
Continuous
Justify your resources every quarter. Nothing is permanent.
Furnace
8-12 weeks, fixed clock
Parallel lane for ideas without a business case. Validate or kill. No extensions.
graduates to THESIS ↑
Gate Decision Mechanics

How Gates Work

Every gate follows the same protocol.

5 days before the gate: The product owner submits gate deliverables to the decision-maker. Not a deck. The actual artifacts -- the kill criterion, the market data, the prototype, whatever the phase required.

At the gate, the decision-maker reviews. The options are:

Advance
Deliverables meet the bar. Next phase activates. The clock starts.
Kill
Kill criterion was met, or deliverables don't hold up. Product stops. Team redeploys. Learnings documented.
Redirect to Furnace
The idea has merit but isn't ready for the pipeline. Move it to the Furnace for an 8-12 week exploration cycle.

There is no "conditional advance." There is no "come back with more data." Those are kills dressed up as hope. Yes or dead.

Council gates (Thesis and Launch only) follow the same protocol but the decision-maker is the council, not the product owner's direct leader. The council sees the product twice. Everything in between belongs to the product owner.

Team-Level Checkpoints

What Happens Between Gates

A checkpoint is not a gate. Same deliverables, different reviewer. The product owner signs a 1-2 page checkpoint memo. No meeting required. No scheduling delay. No committee.

The product owner's authority is bounded by three constraints: the kill criteria (written before the phase, ratified by council), the time box (non-negotiable), and the deliverable requirements (if they're incomplete, the checkpoint can't pass).

If a phase fails at checkpoint, the product owner can restart it once at half the original time box. A third attempt requires council escalation. There is no infinite retry.

Roles

Who Does What

  • Product Owner
    Owns the product through the pipeline. Responsible for all deliverables, all phase transitions, all kill decisions between council gates. One person per product. Named, not titled.
  • Decision Maker
    The person who approves or kills at each non-council gate. Usually the Product Owner's direct leader. Named before the phase starts.
  • Council
    The group that reviews at Gate 1 (Thesis) and Gate 2 (Launch). Composition varies by org. The council doesn't manage the product. It authorizes resources (Gate 1) and authorizes market entry (Gate 2).
  • Furnace Lead
    The person running a Furnace cycle. Operates outside the pipeline with a mandate from leadership. One person, maybe two. Reports to a tight oversight group, not to a BU.
  • Oversight Group
    2-3 senior leaders who set Furnace priorities and approve cycle plans. Not a committee. A mandate source.
Phase Handoff Logic

What Connects the Phases

Every phase produces outputs that become inputs to the next.

  • Signal → Thesis The hypothesis document and competitive scan become the foundation of the business case.
  • Thesis → Build The validated business case and market conversations become the build spec.
  • Build → Prove The working prototype and unit economics become the pilot design.
  • Prove → Launch The pilot data and refined economics become the go-to-market plan.
  • Launch → Sustain The launch metrics become the quarterly health check baseline.
  • Furnace → Thesis The exploration output substitutes for Signal and provides evidence for the Thesis gate.
The Two-Gate Philosophy

Most PLMs have a gate at every phase transition. This one has two. The first gate (Thesis) answers: should this product exist? The second gate (Launch) answers: is this product ready for the market? Everything between those two gates is execution. The product owner bears full accountability.

Why only two? Because gate fatigue is where products go to die. When a council reviews the same product five times, by gate three they're rubber-stamping. The review becomes a meeting about a meeting. Two gates forces two real decisions. The rest is trust -- in the product owner, in the kill criteria, and in the clock.

Kill Criteria

Kill criteria are written before the phase starts, when everyone is rational. The gate is where you honor that rationality instead of succumbing to sunk cost.

The most common failure in product development isn't building the wrong thing. It's continuing to build the wrong thing after the data said stop. Kill criteria exist because organizations are bad at stopping. Named criteria, written in advance, with a named decision-maker who can't delegate the call. That's the discipline.

Full specification → Manual Sections 04-05: Gates & Checkpoints
Leg 2: People Mechanisms

Structure Over Culture

Culture does not survive a reorg. These fifteen mechanisms do. People leave. These don't.

01
Named decision authority with a 5-day clock
One person is named as the decision-maker. Not a title. A person. They have 5 business days. If the clock expires without a decision, the default answer is yes and the product advances.
02
Time-boxed disagreement
Anyone can object. They have 10 business days to formalize the objection with evidence. If the window closes without a formal objection, the decision stands. Silence is consent.
03
Three-level escalation
Decision-maker decides. Objector objects with evidence. Executive resolves in 5 days. That's it. No steering committees. No monthly governance reviews. Three levels, three people, hard deadlines at each.
04
Cross-functional visibility without consensus
Every function sees the decision. Nobody gets a veto. Notification is not the same as approval. The product owner notifies, documents the notification, and moves. If you were notified and said nothing, you agreed. For products that directly impact a BU's P&L or client relationships, notification upgrades to consultation. The BU leader has the same 10-day objection window, but their objection triggers a mandatory conversation between the product owner and the BU leader before the decision-maker rules. The BU leader still doesn't have veto power. But they have a guaranteed seat in the room before the decision is made. Consultation means your perspective is heard. Veto means you can stop progress. One builds alignment. The other builds paralysis.
05
Furnace as the ultimate people solution
When the real problem is political (a VP who won't let a product die, a team that can't agree on direction), move the product to the Furnace. Structurally separate. 8-12 weeks. Let the evidence decide. The politics can't reach it there.
06
Resource contention protocol
When two pipeline products compete for the same resource, the product with the earlier gate deadline gets priority. If deadlines are equal, the product with stronger kill-criterion performance in its most recent phase gets priority. If neither rule resolves it, the CPO (or designated portfolio owner) decides within 48 hours. No standing resource allocation committees. The rule resolves 90% of contention automatically. The CPO resolves the rest.
07
Blind Gate
Anonymous pre-commitment ballot before council discussion begins. Every council member submits proceed/kill before hearing anyone else's position. Prevents the highest-ranking person from anchoring the outcome before the evidence is reviewed.
08
Kill Criteria Red Team
External reviewer challenges the falsifiability of kill criteria before Gate 1. If the criteria cannot be disproven within the phase time box, they get rewritten. Kill criteria that can't be disproven by available data within the time box get rewritten before Gate 1 convenes.
09
Grief Protocol
When the people who built honestly get punished for a correct kill decision, nobody builds honestly again. Structured debrief plus priority reassignment for displaced builders.
10
Transition Quarantine
30-day frozen scope for Furnace graduates entering the Pipeline. No feature requests, no scope changes, no political layering. The product enters the Pipeline as-is.
11
Zombie Detection
Annual review of every Sustain-phase product. Would we build this today? If no, 90-day sunset path. Products don't have tenure.
12
Cover Clause
Named decision-makers are protected from outcome-based punishment. A good decision with bad outcome is still a good decision. Without this, named authority becomes named liability.
13
Watch Flag
Intuitive objections tracked without blocking progress. Flag follows the product. If concern materializes, it escalates. If not, it expires. Pattern recognition without veto power.
14
Scope Fingerprint
Annual comparison of what the product does today against what Gate 2 approved. If they don't match, reclassify. Drift is normal. Invisible drift kills portfolio coherence.
15
Pendulum Self-Diagnostic
Annual assessment of whether FORGE itself has calcified. Are gates performative? Are time boxes routinely extended? Is the Furnace being raided? The framework audits itself because nobody else will.
DIAGNOSTIC

FORGE Self-Assessment

Five questions. Scored automatically. Find out whether FORGE is operating as designed or drifting toward the calcification it was built to prevent.

Question 1
How long does a typical gate review take to schedule and complete?
1 week2 weeks4 weeks6+ weeks
Question 2
What percentage of time-boxed phases received extensions in the last year?
0%10-20%30-40%50%+
Question 3
What percentage of your Furnace allocation was redirected to Pipeline work in the last year?
0%10-20%30-40%50%+
Question 4
How many products were killed at a gate in the last 12 months?
3+210
Question 5
When a decision clock expires, what usually happens?
Decision defaults to proceedClock gets reset
Full specification → Manual Section 08: 15 People Mechanisms
Leg 3: Portfolio Governance

The Constraint Layer

How many products can run at once. Max three Full products in Build or Prove simultaneously. When the pipeline is full, new products queue at Thesis until a slot opens. One person owns the queue. The queue is ordered by market urgency, not internal politics.

CPO Decision Framework

Five decisions belong to the CPO. Not a committee. Not a working group. One person with the authority and the accountability to make these calls on a weekly cadence.

01
Queue prioritization
Which products enter Build next. The queue is rank-ordered: external deadline beats competitive pressure, which beats internal demand. The CPO reviews the queue weekly, updates the order, and publishes the list. No portfolio review committee. A list, owned by a name, updated on a cadence.
02
Resource allocation
How to split people across Pipeline and Furnace. The default is 85/15. When the Pipeline is under capacity, excess resources do not flow to the Furnace. They stay idle or get applied to technical debt. The Furnace's 15% is a floor, not a ceiling. Raiding the Furnace to accelerate a Pipeline product is the single fastest way to kill long-term portfolio health.
03
Sunset decisions
When to kill a Sustain-phase product. Every Sustain product gets a quarterly reckoning against three questions: Is it still earning its operating cost? Is it blocking a queue slot? Has it drifted from what Gate 2 approved? If the answer to any is yes, the CPO initiates a sunset review. Sunset is not failure. Sunset is hygiene.
04
Conflict resolution
Pipeline vs. Furnace resource conflicts. When both lanes need the same person or team, the product with the earlier gate deadline gets priority. If deadlines are equal, the product with stronger kill-criterion performance wins. If neither rule resolves it, the CPO decides within 48 hours. No standing resource allocation committees. The rule resolves 90% of contention automatically.
05
Capacity overflow
What happens when the queue is full and a high-priority signal arrives. Three options, in order: accelerate a current product's gate review to free a slot, pause the lowest-priority queued product, or hold the new signal in queue. The CPO does not create a fourth slot. The constraint is the discipline. Breaking it once teaches the organization it is decorative.
Portfolio Rules

Concrete constraints. Not guidelines. Not recommendations. Rules that trigger specific consequences when violated.

R1
Capacity limit
Max three Full products in Build or Prove simultaneously. Function and Update types do not count against this limit. Integration types count as half a slot. This number is configurable per organization, but once set, it is binding. Exceeding it requires a written exception from the CPO with a sunset date on the exception.
R2
Furnace floor
Minimum 15% of total product budget is protected for Furnace activities. This is a floor, not a target. The money does not revert to Pipeline if unspent. It carries forward within the Furnace. Redirecting Furnace budget to Pipeline requires a written exception that expires at the end of the quarter.
R3
No gate skipping
No product skips a gate. No exceptions. Not for the CEO's pet project, not for the "strategic imperative," not for the product that's been running in stealth for six months. If it exists, it goes through the gates. Products discovered mid-stream get classified retroactively and enter at whatever phase matches their current state. They still face the next gate on schedule.
R4
Sunset cadence
Every Sustain product faces a quarterly reckoning. The product owner presents three numbers: operating cost, revenue or value generated, and drift from the original Gate 2 approval. If operating cost exceeds value for two consecutive quarters, the CPO initiates a sunset timeline. No product is grandfathered.
R5
Queue transparency
Every product's position, phase, and queue rank is visible to the organization. Published weekly. No private queues, no shadow priorities, no products that exist outside the list. If it is not on the list, it is not authorized. Visibility is the cheapest governance mechanism available. Use it.
Portfolio Health Indicators

What the CPO monitors. Not vanity metrics. Signals that the system is working or drifting.

Pipeline velocity
Time from Signal to Launch, segmented by product type. Tracks whether the system is accelerating or calcifying. Velocity increasing without quality degradation is healthy. Velocity decreasing across all types signals process bloat.
Alarm: average velocity exceeds time-box sum by 40%+
Furnace conversion rate
Percentage of Furnace cycles that produce Pipeline candidates. Healthy range: 15-30%. Below 15% means the Furnace is exploring without direction. Above 30% means the Furnace is being used as a Pipeline staging area.
Alarm: conversion below 10% or above 40% for two consecutive quarters
Kill rate
Percentage of products killed before Launch. Healthy organizations kill 40-60% of products in the pipeline. Below 30% means gates are rubber stamps. Above 70% means signal detection is broken and too many bad ideas are entering the system.
Alarm: kill rate below 25% for two consecutive quarters
Resource utilization
Pipeline-to-Furnace resource split against the 85/15 target. Tracks whether Furnace budget is being raided. Also tracks idle capacity: if Pipeline slots are unfilled and resources are idle, the organization is underinvesting in signal detection.
Alarm: Furnace allocation drops below 10% of total budget
Gate pass rate
Percentage of products that pass each gate. If everything passes, gates are decorative. If nothing passes, gates are punitive. Healthy pass rates at Gate 1: 50-70%. Healthy pass rates at Gate 2: 60-80%. The difference reflects that Gate 1 should be the harder filter.
Alarm: Gate 1 pass rate exceeds 85% for three consecutive reviews
Queue age
How long products sit in queue before entering Build. Products waiting longer than six months are either low priority (kill them) or blocked by capacity constraints (address the constraint). A growing queue with stable capacity is a signal to tighten Signal-phase filtering.
Alarm: any product in queue longer than 6 months without a gate decision
Queue Management

How products move through the queue. Explicit rules that remove ambiguity about what happens next.

Q1
Queue entry
Products enter the queue when they pass their Signal checkpoint. Not before. An idea without a completed Signal phase is not a queued product. It is a conversation.
Q2
Queue position
Determined by three factors in order: strategic alignment with current portfolio gaps, resource availability for the required product type, and market timing pressure. The CPO assigns position. Position can change weekly. The list is public.
Q3
Queue pause
Products can be paused without being killed. Paused products retain their queue position but do not consume capacity. Valid reasons to pause: resource dependency not yet available, market timing not yet right, regulatory prerequisite not yet met. Paused products get a 90-day review. At 90 days: resume or kill. No second pause.
Q4
Queue expiration
No product lives in queue longer than six months without a gate decision. At six months, the CPO reviews the product with the original Signal documentation. If the signal is still valid and the queue position is justified, the product gets a one-time 90-day extension. If the signal has degraded, the product is killed. No product queues indefinitely.
Q5
Slot release
When a product exits Build or Prove (by completing, failing, or being killed), the slot opens immediately. The next product in queue enters Thesis within one week. No gap between slot release and next product entry. The queue exists so this transition is instant.
Full specification → Manual Section 09: Portfolio Governance

Why Frameworks Fail

Every framework eventually becomes the problem it was built to solve. Stage-Gate started as a discipline tool and became a compliance ritual. The swing is inevitable.

FORGE is designed with that swing in mind.

The Pipeline is the inhale: structured phases, binding kill criteria, hard clocks. The Furnace is the exhale: protected exploration, political extraction, long-range scanning.

The Pendulum thesis says organizations oscillate between rigor and agility, and the swing is structural, not behavioral. You can't coach your way out of it. You build for both positions simultaneously and let the architecture hold when the politics shift. The Pipeline holds under agility pressure. The Furnace holds under rigor pressure. Neither collapses when leadership changes.

Read the full Pendulum framework →

Reading the Swing

The Pendulum tells you where the organization sits. The answer determines which lane leads.

Rigor Phase
Pipeline is dominant. Gates are tight, clocks are enforced, kill criteria are binding. The Furnace runs lighter -- fewer experiments, smaller budget -- but it never stops. If the Furnace goes dark during a rigor phase, the next agility swing will have nothing to draw from.
Agility Phase
Furnace gets more oxygen. More experiments, faster cycles, more tolerance for ambiguity. The Pipeline holds the line -- clocks don't loosen, kill criteria don't soften -- but priority shifts to the products already in motion. If the Pipeline relaxes discipline during an agility phase, the portfolio fills with half-built things.
Mid-Swing
Both lanes run full. This is the highest-risk moment. Leadership is changing, priorities are shifting, and each lane will be pulled toward the other. The swing is where things break. If you feel the pull to merge the lanes or borrow from one to feed the other, you are in the swing. Protect both.
PRODUCT ARCHETYPES

Seven Ways to Configure FORGE

Not every product enters the pipeline the same way. The archetype determines which phases apply, which deliverables change, and how long the clock runs. Pick wrong and you waste time on governance that does not apply. Pick right and the framework fits the product instead of the other way around.

De Novo
Net-new product, no existing infrastructure. Full Pipeline.
Use when building something that does not exist yet in your portfolio or market. No codebase to inherit, no client base to migrate, no partner to integrate. You own every decision from Signal to Sustain.
Full 6-phase lifecycle. All deliverables apply. Two council gates. Longest timeline (26-40 weeks standard). No shortcuts.
Building a new specialty pharmacy governance program from scratch.
Partnership
Co-developed with an external partner. Shared ownership, shared kill criteria.
Use when the product depends on a partner's capability, distribution, or clinical infrastructure. The partner brings something you cannot build in time. Adds partner alignment deliverables and joint governance at every gate.
Prove phase absorbed into Build. Adds partner agreement, joint operational model, and shared GTM plan as required deliverables. Kill criteria include partner-side triggers.
Payer-provider joint venture for episode-based care management.
Platform
Infrastructure that other products build on. Longer Build, continuous adoption.
Use when the product is not sold directly but enables other products to function. Platform adoption is continuous, not event-driven, so the traditional Launch phase folds into Sustain.
Extended Build phase (12-20 weeks). Adds platform architecture, multi-tenant validation, and API surface documentation. Launch folds into Sustain. Longest total timeline (30-48 weeks).
A governance data platform that Cadence, Curated, and Caliber all use.
Regulatory
Compliance-driven. The external clock dominates.
Use when a regulatory change or mandate is the signal. The compliance deadline sets the timeline, not the product team. Signal compresses because the signal is already known. Everything bends to the regulatory calendar.
Signal compresses to 1 week (or less). Adds regulatory review gates and compliance deadline mapping. Prove phase may extend for state-level approval cycles.
A prior authorization automation tool that requires state-level regulatory approval.
Sustain
Existing product optimization. Enters at Sustain phase.
Use for iterations on products already in market. The product has buyers, the economics are known, the question is whether the enhancement justifies the build cost. Shorter cycle. No council involvement.
Three phases only: Signal, Build, Sustain. No Thesis, no Prove, no council gates. Build compresses to 3-4 weeks. Total timeline: 5-7 weeks standard.
Quarterly refresh of a mature specialty pharmacy program.
Exploration
Furnace-native. No Pipeline entry unless extraction criteria are met.
Use when the idea has no business case and no buyer validation. Pure hypothesis testing. If the experiment validates, the product graduates to the Pipeline at Thesis. If not, it dies in the Furnace where the cost is lowest.
Furnace only. No Pipeline phases. Fixed 8-12 week clock. Required deliverables: hypothesis document, cycle plan, kill criterion, final output brief. No engineering resources. No council.
Investigating whether AI-driven claims review could replace manual audit.
Acquisition
Integrating an acquired product. The acquisition was Gate 1.
Use when absorbing a purchased product into your portfolio. Due diligence already happened. The thesis was validated by the acquisition decision. The question is whether you can integrate without destroying the value you paid for. Enters at Build or Prove depending on maturity.
Skips Signal and Thesis. Starts at Build with integration-focused deliverables: system mapping, talent retention plan, client migration plan, cultural integration checkpoint. Compressed Prove phase (4-6 weeks). Kill criteria shift to integration risk.
Absorbing a purchased billing integrity platform into an existing portfolio.
WHICH ARCHETYPE?

Decision Tree

Answer the questions in order. The first yes determines the archetype.

Does the product already exist in your portfolio?
Was it acquired from another company?
Is it speculative with no business case?
Does it require regulatory approval to operate?
Does it involve an external partner?
Is it infrastructure that other products build on?

The Second Lung

Where ideas live before anyone commits build resources. Protected budget, separate owner, no obligation to ship.

Exploration
Pure hypothesis testing. No business case required. Fixed clock, defined learning objectives, kill criterion written before the experiment starts. One person, part-time. Access to data and subject matter experts. No dedicated engineering.
Incubation
Signal detected but too fragile for Pipeline discipline. Tighter milestones, still separate from Pipeline governance. Small team, 2-4 people. Budget for proof-of-concept, not product development.
Extraction
Pull a struggling Pipeline product into the Furnace. Remove the political pressure. Let the data speak without an audience. Reintroduce or kill based on evidence. The original product lead does not run the extraction. Fresh eyes are the point.
Adjacency
Test whether an existing product can serve a new market or population. Uses existing product data as starting signal. Faster cycle, narrower hypothesis.
Horizon Scanning
Long-range market and regulatory monitoring. No build activity. Output is a brief, not a backlog. The peripheral vision the Pipeline lacks.
Recovery
Structured retrieval from killed products. Within 30 days of a kill, two senior people can petition the Furnace Lead. One 90-day cycle. Did the kill criteria miss something? The assessor must not have been involved in the original work or the kill decision.

The Pendulum Connection. The Pipeline creates the pressure that forces decisions. The Furnace creates the release that prevents the Pipeline from suffocating the next generation of products. Both run simultaneously.

Full specification → Manual Section 06: The Furnace
Furnace Mode Specifications

What Each Mode Produces

The Pipeline has six phases, each with named deliverables, kill criteria, and time boxes. The Furnace needs the same discipline. Without per-mode specifications, the Furnace becomes a parking lot for ideas nobody wants to kill.

Exploration
Pure signal detection. No thesis required. Looking for problems worth naming.
Time Box
6-8 weeks
Required Deliverables
  • Problem statement brief -- Names the gap, the population affected, and why existing solutions miss it. One page.
  • Signal source log -- Every data point that supports the hypothesis: market reports, customer conversations, regulatory filings, claims data, competitor moves. Dated and attributed.
  • Buyer landscape scan -- Who would care about this problem being solved? A list of specific buyer types with specific pain points, assembled from evidence, not intuition.
  • Cycle-end recommendation -- Promote to Incubation, pivot the hypothesis, or kill. One page with evidence.
Optional Deliverables
  • Adjacent product impact map -- Does this problem overlap with anything already in Pipeline or Furnace?
  • Regulatory landscape summary -- If the problem sits in a regulated space, the relevant compliance calendar and constraints.
Kill Criterion
No named problem with an identifiable buyer after the time box expires. If the team cannot articulate who hurts and why, the signal was noise. Also kills immediately if the problem is already being solved by a Pipeline product the team was not aware of.
Exit Condition
A named problem with at least three independent signal sources and a buyer type willing to discuss it. That is enough to enter Incubation.
Incubation
Early-stage concept development. Hypothesis exists but is untested.
Time Box
8-12 weeks
Required Deliverables
  • Hypothesis document -- States the gap, the buyer, and the test that determines merit. Falsifiable. One page.
  • Experiment design -- Week-by-week learning plan. What evidence will be generated, how it will be measured, and what constitutes a positive or negative result.
  • Buyer validation notes -- Minimum two conversations where the concept was described and the buyer's reaction documented. Not customer discovery theater.
  • Kill criterion -- Written before the cycle starts. Specific conditions that would stop the experiment early.
  • Final output brief -- Two to three pages. What the experiment tested, what the results were, and whether the concept should move to Extraction or die.
Optional Deliverables
  • Rough prototype -- If the concept benefits from something tangible, a sketch-level prototype.
  • Unit economics sketch -- Early-stage cost and revenue assumptions. A napkin calculation that identifies whether the economics could ever work.
Kill Criterion
Hypothesis cannot be validated within the time box. Buyer conversations produce indifference or active disinterest. Also kills if the experiment reveals a structural barrier that makes the concept unworkable.
Exit Condition
A validated hypothesis with buyer evidence and a recommendation to extract into Pipeline.
Extraction
Pulling a proven Furnace concept into Pipeline for productization.
Time Box
4-6 weeks
Required Deliverables
  • Extraction brief -- Summarizes the Furnace evidence, names the Pipeline entry point (Thesis), and identifies what the Furnace work already answered. This replaces the Signal phase.
  • Scope lock document -- Defines exactly what transfers to Pipeline and what does not. Prevents scope inflation during the handoff.
  • Resource and ownership transfer plan -- Names the Pipeline product owner, team composition, and handoff timeline. The Furnace Lead does not become the Pipeline product owner. Clean break.
Optional Deliverables
  • Technical debt assessment -- If a prototype was built during Incubation, what needs to be rebuilt vs. what can carry forward.
Kill Criterion
No Pipeline capacity within 30 days. Also kills if the CPO determines the evidence does not meet Pipeline Thesis standards after reviewing the extraction brief.
Exit Condition
A product enters Pipeline at Thesis with a named product owner, a locked scope, and the Furnace evidence as its foundation. The 30-day Transition Quarantine activates.
Adjacency
Extending existing products into neighboring use cases.
Time Box
6-8 weeks
Required Deliverables
  • Adjacency hypothesis -- Names the existing product, the new market or population, and why the product might transfer. Cites existing product performance data.
  • Gap analysis -- What is different about the adjacent market? Where does the existing product fit and where does it break?
  • Buyer validation (minimum 2 conversations) -- Does the existing product's value proposition translate, or does it need to be reframed?
  • Recommendation brief -- Extend the existing product, fork a new product for the adjacent market, or kill. One page with evidence.
Optional Deliverables
  • Cannibalization risk assessment -- Will the adjacent product compete with the existing product for the same buyers or resources?
Kill Criterion
The adjacent market requires more than 30% modification to the existing product. At that point it is a new product and belongs in the Pipeline as a de novo entry.
Exit Condition
Extend the existing product (Function Change) or fork a new product (enters Pipeline at Thesis via Extraction).
Horizon Scanning
Long-range market, regulatory, and technology monitoring.
Time Box
Continuous (quarterly brief)
Required Deliverables
  • Quarterly horizon brief -- Covers three horizons: regulatory changes (12-24 months), market shifts (6-18 months), and technology developments (12-36 months). Names specific changes, assigns probability, and flags anything that should trigger an Exploration cycle.
  • Signal escalation list -- Any horizon signal that crosses from "interesting" to "requires action." Named signals with recommended next steps.
  • Prior quarter accuracy check -- What was predicted? What was missed? What was predicted and did not happen? The scanner must account for its own accuracy.
Optional Deliverables
  • Deep-dive memo -- When one signal warrants more investigation than a brief allows. Three to five pages.
Kill Criterion
Two consecutive quarters with no usable signals and less than 30% prediction accuracy. Pause for one quarter and reassess scope.
Exit Condition
Horizon Scanning does not exit. It produces inputs. When a signal crosses the action threshold, it triggers an Exploration cycle.
Recovery
Rescuing a stalled or failing Pipeline product by returning it to Furnace conditions.
Time Box
90 days (single cycle, no extensions)
Required Deliverables
  • Kill decision review -- What triggered the kill? Was the criterion correctly constructed? Was the data accurate? Written by someone not involved in the original work or the kill decision.
  • Revised hypothesis -- What was wrong with the original thesis and what evidence now suggests it may still be viable.
  • Evidence test -- A targeted experiment that addresses the specific gap the original work could not close. A focused test of the exact question that killed the product.
  • Final recovery recommendation -- Re-enter Pipeline at Thesis, return to Furnace archive, or confirm kill. Two pages with evidence.
Optional Deliverables
  • Market condition comparison -- Has something structural changed since the kill that makes the product viable now?
  • Process failure analysis -- If the kill was caused by process rather than product, what broke and what would prevent it.
Kill Criterion
The evidence test confirms the original kill was correct. Also kills if the 90-day clock expires without a clear recommendation. Recovery is not a reprieve. If the second look produces the same answer, the product stays dead.
Exit Condition
The product re-enters Pipeline at Thesis with new evidence and a new product owner, or the kill is confirmed and the archive is updated.
Furnace Lead Governance

Who Runs the Furnace

The Furnace Lead owns the Furnace the way a Product Owner owns a Pipeline product. One person. Named, not titled. Runs all active cycles, manages the Furnace budget, decides what enters and exits. The role requires someone comfortable with ambiguity, comfortable killing their own experiments, and comfortable telling a VP that their pet idea failed a six-week test.

Not a junior role. Not a rotation assignment. A senior product leader who reports directly to the CPO and operates outside the Pipeline reporting structure. If the Furnace Lead reports to a Pipeline Product Owner or a BU leader, the Furnace becomes a feature factory for the loudest stakeholder.

  • Reporting Structure
    Furnace Lead reports to CPO. Not to a Pipeline Product Owner, not to a BU leader, not to an innovation committee. Direct line to the person who owns the portfolio.
  • Decision Rights: Furnace Lead Alone
    Start or kill any Exploration, Incubation, Adjacency, or Recovery cycle. Allocate Furnace budget across active cycles. Reject cycle requests that do not meet the entry bar.
  • Decision Rights: Requires CPO
    Extraction decisions. Any single cycle consuming more than 40% of quarterly Furnace budget. Changes to overall Furnace budget allocation. Overriding a Furnace Lead kill decision.
  • Escalation Path
    Furnace Lead vs. Pipeline Product Owner conflicts: CPO resolves within 5 business days. If the clock expires, the Furnace Lead's position holds. The Furnace is structurally protected because it is structurally weaker. Without explicit protection, the Pipeline always wins.
  • Budget Allocation
    15-20% of total product budget, ring-fenced. Set annually, reviewed quarterly, cannot be raided for Pipeline overruns. The moment the Furnace budget becomes a Pipeline reserve fund, the Furnace is dead.
Furnace Cycle Review

Governance Without Overhead

The Pipeline has council gates. The Furnace needs governance too, but lighter. Two cadences.

Monthly Furnace Review
30 minutes. Furnace Lead presents to CPO and Oversight Group. Status of every active cycle: on track, at risk, or recommending kill. Budget burn vs. plan. Any cycles requesting entry or recommending Extraction. If no decisions are needed, the review takes 15 minutes.
Format: Written update 2 days before. Meeting covers only decision items. If none, the meeting is canceled. Attendance: Furnace Lead, CPO, Oversight Group (2-3 people). No observers.
Extraction Decision
The formal gate for moving a Furnace concept into Pipeline. Triggered by the Furnace Lead when a cycle produces a positive recommendation. Not scheduled. Called when needed. The CPO reviews the extraction brief, scope lock, and transfer plan. Decision: approve extraction (Thesis entry with 30-day Transition Quarantine), return to Furnace for another cycle, or kill.
Decision authority: CPO. One person, one decision, one accountability.
The Thesis

Every other PLM framework assumes the organization is the asset and the process protects it. This one starts from a different premise: most products die of oxygen deprivation, not poor decision-making. The clock, the kill, the named owner, and the Furnace are the corrections. What follows is the operating system.

This framework was built inside a top-three national health plan's product organization after 50+ interviews with product leaders, medical directors, and operations executives across health plans, health systems, and health technology companies who described the same failure modes. The phases are not theoretical. The kill criteria are not aspirational. The people mechanisms are not borrowed from a textbook. Every element exists because something specific broke without it.

The two-lane architecture came from a specific observation: organizations swing between rigor and agility, and every swing breaks whatever the last swing built. A framework that picks one side becomes a weapon for the other. FORGE has two lanes because the Pendulum is real. The Pipeline holds when leadership wants speed. The Furnace holds when leadership wants control. Neither lane exists to win. Both exist to survive the swing.

10 Non-Negotiables
01
Kill criteria at every gate
Written before the phase starts, falsifiable, and binding. If you can't name what would kill this product, you haven't earned the right to fund it.
02
The clock is the discipline
Every phase is time-boxed. The clock forces decisions that consensus never will. One extension per phase, max 3 weeks, with written justification. A second request triggers a kill review.
03
One person decides
Named, not titled. Not a committee. Committees distribute accountability until it disappears.
04
Market data over internal consensus
Buyer conversations, not alignment meetings. The market is the arbiter. Internal agreement is a comfort metric.
05
Dual-lane architecture
Pipeline for products with a business case. Furnace for ideas that don't have one yet. Both run simultaneously. Neither contaminates the other.
06
People problems solved by structure
Named decision authority, time-boxed disagreement, three-level escalation. Process outlasts personality.
07
Fail-fast economics are explicit
The cost of running an 8-week experiment is knowable. The cost of a product that should have died in month three but ran for eighteen months is also knowable. Name both numbers.
08
Two gates, not five
The council sees the product at concept and at launch. Everything between belongs to the product owner. Five gates teaches an organization to rubber-stamp.
09
The deliverable load is constant
Whether the council reviews or the product owner reviews, the standard doesn't change. The reviewer changes. The rigor does not.
10
The framework reviews itself
Annual Pendulum assessment of whether FORGE has calcified. Are gates performative? Are time boxes routinely extended? Is the Furnace being raided? The framework audits itself because nobody else will.
Full specification → Manual Section 10: Non-Negotiables
Comparison

FORGE vs. Stage-Gate

Stage-Gate is the default. Most organizations run some version of it. Here is where the two frameworks diverge and why the differences matter.

Dimension Stage-Gate FORGE
Council gates Gate at every phase transition (typically 5) Two council gates. Concept and launch. Everything else belongs to the product owner.
Time discipline Phases end when deliverables are "ready" Every phase is time-boxed. The clock forces the decision.
Kill mechanism Gates can kill, but rarely do in practice Kill criteria written before the phase starts. Binding. No conditional advance.
Decision authority Committee-based governance One named person decides at each non-council gate. Not a title. A person.
Exploration Pre-development "Discovery" stage Parallel Furnace with its own governance, clock, and kill criteria.
Configurability One path fits all product types Seven archetypes, each with modified phases, deliverables, and timelines.
People problems Assumed to be solved by culture Fifteen structural mechanisms. Named authority, time-boxed disagreement, escalation protocol.
Portfolio layer Separate from the PLM Built in. Capacity limits enforced. Queue owned by a name.
Full specification → Manual Section 15: Comparative Analysis
WHAT FORGE KILLS

10 Organizational Pathologies

1. The Zombie Product
A product that should have been killed three quarters ago. Still consuming budget because nobody will make the call.
Killed by: prospective kill criteria + time boxes + Zombie Detection (Mechanism 11).
2. The Golden Child
A product that gets special treatment because a senior leader championed it.
Killed by: the Furnace quarantines it. Separate governance, fixed clock, same clock as everything else.
3. Death by Committee
Every phase transition requires committee approval. Adds 10-20 weeks.
Killed by: two gates, not five. The product owner runs everything between.
4. The Consensus Trap
Every function has an implicit veto. Nobody says no. Everybody says "not yet."
Killed by: notification, not consensus. Time-boxed disagreement. 15 days max.
5. Scope Creep Camouflage
An Update grows into a Function without reclassification.
Killed by: fixed classification + time box enforcement. Exceeds the box? Reclassify or ship what's ready.
6. The Evidence-Free Launch
Product reaches market without buyer validation.
Killed by: Gate 2 requires documented buyer willingness to pay. Not "we think they'll pay." Documented.
7. The Permanent Pilot
A pilot that never ends because the results are "encouraging but not conclusive."
Killed by: Prove phase time box. Two cycles max. Then you go to Gate 2 with whatever you have.
8. The Orphaned Product
Product launches. The builders move on. Nobody is named to run it.
Killed by: Sustain is an explicit phase with a named owner, quarterly reviews, and kill criteria.
9. The Indecision Default
Nobody makes the call. Weeks pass.
Killed by: 5-day decision clock. Default is proceed. Indecision falls on the indecisive, not the team.
10. The Over-Stuffed Pipeline
Too many products, not enough resources. Everything is slow.
Killed by: portfolio capacity limits. Max 3 Full products in Build/Prove. Hard constraint.
Full specification → Manual Section 13: 10 Pathologies FORGE Kills
HOW IT WORKS IN PRACTICE

One Product Through the System

A payer's specialty care team identifies a gap: high-risk maternity members are generating avoidable NICU admissions. The clinical signal is real. The cost data supports it. Nobody has built a structured intervention.

Signal (2 weeks)
The product owner documents the signal: 3,200 high-risk pregnancies annually, average NICU cost of $78,000 per admission, 22% of admissions flagged as potentially avoidable through earlier intervention. Buyer hypothesis: health plan medical directors managing maternity spend. Competitive scan: two vendors offer point solutions, neither integrates with existing care management. Go/kill recommendation: proceed to Thesis.
Checkpoint
Product owner signs. No committee. Two weeks elapsed.
Thesis (5 weeks)
Written thesis with falsification criteria: "This product should not exist if fewer than 3 buyer conversations confirm willingness to pay at $2.50+ PMPM, or if the clinical intervention cannot demonstrate measurable impact within a single pregnancy term." Five buyer conversations completed. Three confirmed pricing interest. Unit economics modeled: break-even at 1,800 members. Kill criteria for Build written.
Gate 1 -- Concept Approval (45 minutes)
Council reviews. The thesis holds. The economics are credible. The kill criteria are falsifiable. Decision: proceed to Build. Time box: 10 weeks.
Build (10 weeks)
Working prototype: a clinical pathway with risk stratification, provider-facing dashboard, and member outreach workflow. Not polished. Functional. Technical architecture documented. Operational cost model built on actual infrastructure requirements, not Thesis estimates. Kill criterion status: none triggered.
Checkpoint
Product owner signs. 10 weeks elapsed.
Prove (10 weeks)
Pilot with 400 members across two markets. Results: 18% reduction in avoidable NICU admissions. Member satisfaction 4.2/5. Provider adoption slower than projected (62% vs. 80% target). Revised unit economics: break-even at 2,100 members (higher than Thesis model due to provider onboarding costs). Buyer feedback: "We'd pay $2.75 PMPM but need the provider adoption story to improve." Kill criteria not triggered -- adoption is below target but economics still viable.
Gate 2 -- Launch Approval (55 minutes)
Council reviews pilot data against kill criteria. The 18% NICU reduction is strong. The provider adoption gap is real but addressable. Go-to-market plan has named owners and dates. Operational readiness confirmed by the team that will run it, not the team that built it. Decision: proceed to Launch.
Total elapsed: 29 weeks. One product, two council meetings, zero committee cycles between them.
WITHOUT FORGE, THIS PRODUCT TOOK 14 MONTHS AND NEVER LAUNCHED.

The same organization attempted the same product the previous year. Month 1: scope defined. Month 2: undocumented strategy shift by a VP who wanted to expand scope to include postpartum depression. Months 3-5: cycling iterations between product, clinical, and the BU over what "maternity" means.

Month 6: pricing model rebuilt for the third time because the expanded scope changed the economics. Month 8: compliance review flagged TCPA risk on member outreach. Month 10: compliance review completed, vendor contract expired during the delay. Month 12: new RFP required. Month 14: deprioritized by incoming leadership. Zero members served.

The product was never bad. The organization was never incompetent. The structure made speed impossible and nobody had permission to say so.

That is what FORGE prevents.

Full specification → Manual Section 14: One Product Through the System
Interactive Tool

Kill Criteria Generator

Test whether your kill criterion is specific and measurable enough to be binding.

Deliverable Deep-Dives

What Each Phase Actually Produces

Detailed deliverable specifications for each phase are available through the configuration console. Select a product type in Quick Start, then expand any phase card to see the full deliverable list with descriptions, quality bars, and context for why each artifact exists.

Read the full FORGE Operating Manual →
Templates

Working Documents

Editable templates for every deliverable, gate review, and governance document.

Click any template to preview, edit, print, or save.

Full specification → Manual Section 11: Deliverables by Phase

If FORGE is how you want to run your product org, I've implemented it.

joe.nalley@showyourwork.health

Available for advisory, implementation, and portfolio governance.

Draft saved
The Problem
Most products don't die from bad ideas. They die because the organization kills them slowly. Three pathologies show up everywhere.
The Zombie Product
A product that should have been killed three quarters ago. Still consuming budget because nobody will make the call.
The Orphaned Product
Product launches. The builders move on. Nobody is named to run it. It drifts until someone notices the cost line.
The Golden Child
A product that gets special treatment because a senior leader championed it. Evidence doesn't matter. Politics do.
The Architecture
FORGE runs two lanes simultaneously. Products with a business case enter the Pipeline. Ideas without one go to the Furnace. Neither contaminates the other.
Pipeline
Six phases: Signal, Thesis, Build, Prove, Launch, Sustain. Two council gates. Time-boxed. Kill criteria written before work begins. The structured path from idea to market.
Furnace
Six exploration modes with fixed clocks of 8-12 weeks. Separate governance. No business case required. When the evidence is ready, it graduates to Pipeline. When the clock runs out, it stops.
The dual-lane design comes from the Pendulum -- the structural swing between rigor and agility that every organization makes. FORGE doesn't pick a side. It builds for both.
The Discipline
Fifteen structural mechanisms replace culture as the governance layer. Five of them do the most work.
Named Authority
One person owns the decision. Not a committee. They have 5 business days. If the clock expires, the default is yes.
Time-Boxed Disagreement
Anyone can object -- with evidence, within 10 business days. Silence is consent. No implicit vetoes.
Kill Criteria
Pre-defined conditions that trigger product termination. Written before work begins. Falsifiable by available data within the time box.
Blind Gate
Anonymous ballot before council discussion. Every member commits proceed or kill before hearing anyone else. Prevents the highest-ranking voice from anchoring the outcome.
Grief Protocol
Structured acknowledgment when a product is killed. Debrief plus priority reassignment. If builders get punished for correct kills, nobody builds honestly again.
The Constraint Layer
Structural limits that prevent the pipeline from consuming everything. Not guidelines. Hard constraints.
Max Three in Build or Prove
No more than three Full products in Build or Prove at any time. Beyond that, quality degrades and timelines slip. The constraint forces prioritization before work begins.
15% Furnace Ring-Fence
Fifteen percent of product budget is ring-fenced for the Furnace. Protected. Non-negotiable. If the Furnace starves, the pipeline eventually starves too.
One Queue Owner
One person owns the queue. Not a committee. Not a rotating role. One person who decides what enters Build and in what order. Visible to everyone. Accountable to the CPO.
Weekly Portfolio Publish
The portfolio is published weekly. Every product, its phase, its clock, its owner. No hidden projects. No stealth work. If it is not on the board, it does not exist.
See It Work
A payer's high-risk maternity product. Same organization, same product, two different outcomes.
Without FORGE

Month 1: scope defined. Month 2: a VP expanded scope to include postpartum depression without documentation.

Months 3-5: cycling iterations between product, clinical, and the BU over what "maternity" means.

Month 6: pricing rebuilt a third time. Month 10: compliance review completed, vendor contract expired during the delay.

14 months. Never launched. Zero members served.

With FORGE

Signal (2 wks): documented gap, buyer hypothesis, competitive scan. Checkpoint: product owner signs.

Thesis (5 wks): falsification criteria written, five buyer conversations, unit economics modeled. Gate 1: 45 minutes.

Build (10 wks): working prototype. Prove (10 wks): 400-member pilot, 18% NICU reduction. Gate 2: 55 minutes.

29 weeks. Launched. Two council meetings total.

Try It
You've seen the problem, the architecture, the discipline, the constraints, and a real example. Now configure it for your product.
Pick a starting point. The Blueprint configures FORGE for your product type, org structure, and risk tolerance. The Diagnostic tells you whether your current process is drifting.
1 / 6