Skip to main content

Funnel Engineering

Your funnel is not your website. It is the system that converts qualified interest into committed revenue.

A sales funnel is a measurement instrument. Each stage answers a question. If the answer is wrong, the deal doesn't advance. If the answer is never asked, the deal stalls and nobody knows why.

Overview

AttributeValue
PurposeDesign, measure, and optimize the stages deals move through from qualification to close
TriggerQuarterly funnel review, or when conversion rate drops below target at any stage
FrequencyContinuous tracking, quarterly architecture review
DurationOngoing — 30 min/day pipeline review, 2-4 hours quarterly redesign
OwnerSales Leadership (human strategy) + RevOps (AI analytics)
OutputStage-by-stage conversion data, bottleneck diagnosis, optimization actions

Human Role: Stage design, conversion strategy, deal progression decisions AI Role: Pipeline analytics, forecasting, velocity tracking, bottleneck detection Spectrum: AI-Assisted


Prerequisites

Tools Required

ToolPurposeAccess
CRM pipelineVisual deal tracking by stageSales CRM
Sales forecastingPredict deal close probabilitySales Forecasting algo
Analytics dashboardConversion metrics by stageCRM reporting

Knowledge Requirements

  • Understanding of buyer's decision process (not your selling process)
  • Historical deal data (close rates, cycle times, deal sizes)
  • Product/market fit assumptions validated through lead qualification

Inputs

InputSourceRequired?
Qualified leadsLead QualificationYes
Historical deal dataCRMIf available
Buyer journey understandingCustomer interviews, ICP researchYes
Revenue targetsBusiness planYes

Process

Phase 1: Design the Funnel

Duration: 2-4 hours (annual, refresh quarterly) Responsibility: Human-led

The funnel architecture should mirror the buyer's decision process, not your internal org chart.

Step 1.1: Define Stages

Each stage represents a BUYER decision, not a seller activity.

StageBuyer QuestionEvidence RequiredExit Criteria
Lead"Should I learn more?"Responded to outreachAccepted a conversation
Discovery"Do they understand my problem?"Attended discovery callProblem confirmed, mutual interest stated
Qualified"Is this worth my team's time?"Shared internal requirementsDecision maker engaged, budget discussed
Proposal"Does the solution match my needs?"Reviewed proposal or demoSpecific feedback given
Negotiation"Can we make the terms work?"Active discussion on price/termsTerms agreed in principle
Won"Let's do this"Signed agreement or POPayment or commitment received
Lost"Not now / not this"Explicit decline or 30-day silenceReason captured, learning documented

Step 1.2: Set Stage Metrics

StageTarget ConversionTarget DurationDeals Expected
Lead → Discovery40-60%1-2 weeks?
Discovery → Qualified50-70%1-2 weeks?
Qualified → Proposal60-80%1-3 weeks?
Proposal → Negotiation40-60%1-2 weeks?
Negotiation → Won50-70%1-2 weeks?
Overall: Lead → Won5-15%4-12 weeks?

Step 1.3: Work Backward from Revenue

Revenue target:        $X/quarter
Average deal size: $Y
Close rate: Z%
Deals needed: X / Y = N deals
Pipeline needed: N / Z = P pipeline deals
Leads needed: P / stage_conversion_rates = L leads

This math tells you exactly how many leads to generate. If the numbers don't work, the problem is either deal size, close rate, or volume — and you know which to fix.

Phase 1 Output: Funnel architecture with stages, metrics, targets


Phase 2: Track Pipeline Health

Duration: 30 minutes daily, 1 hour weekly Responsibility: AI-led tracking, human interpretation

Step 2.1: Daily Pipeline Review

CheckWhat to Look ForAction
Stalled dealsAny deal in same stage >2x target durationAsk: what's blocking? Who needs to act?
Overdue follow-upsTasks past due dateComplete today or reschedule with reason
New entriesLeads that entered pipeline todayVerify qualification, assign next action
Stage changesDeals that advanced or regressedCelebrate advances, diagnose regressions

Step 2.2: Weekly Pipeline Metrics

MetricFormulaWhat It Tells You
Pipeline valueSum of all deal values × probabilityExpected revenue
Pipeline coveragePipeline value / quotaNeed 3-4x coverage minimum
Velocity(# deals × avg value × win rate) / avg cycleRevenue per time period
Stage distributionCount of deals per stageHealthy = pyramid shape
Average deal ageDays since deal createdAging = stalling

Step 2.3: Funnel Shape Diagnosis

ShapeWhat It Looks LikeDiagnosis
Pyramid (healthy)Many at top, few at bottomNormal — leads filter through stages
CylinderEqual at every stageNot qualifying enough — too many deals advancing
Inverted pyramidFew at top, many at bottomPipeline will dry up — lead gen needed
HourglassMany top, few middle, many bottomBottleneck in middle stages — qualification or proposal problem
FlatVery few at all stagesPipeline emergency — stop everything else, generate leads

Phase 2 Output: Pipeline health report with diagnosis


Phase 3: Optimize Conversions

Duration: 2-4 hours quarterly Responsibility: Human analysis, AI data

Every stage has a conversion rate. Improving any one rate compounds across the whole funnel.

Step 3.1: Identify the Bottleneck

The bottleneck is the stage with the lowest conversion rate relative to target.

Stage TransitionTargetActualGapPriority
Lead → Discovery50%???
Discovery → Qualified60%???
Qualified → Proposal70%???
Proposal → Negotiation50%???
Negotiation → Won60%???

Fix the bottleneck first. A 10% improvement at the bottleneck has more impact than 10% everywhere.

Step 3.2: Diagnose Root Cause

Bottleneck StageCommon CausesDiagnostic Questions
Lead → DiscoveryPoor outreach, wrong ICPAre replies relevant? Do they ask questions?
Discovery → QualifiedWeak discovery, no pain confirmedDid they name the pain unprompted?
Qualified → ProposalValue not clear, stakeholders not alignedDid the decision maker attend?
Proposal → NegotiationPrice shock, feature gap, wrong frameWhat specific objection killed it?
Negotiation → WonCompeting priorities, internal politicsWhat changed between "interested" and "silent"?

Step 3.3: Run Improvement Experiment

Pick ONE bottleneck. Design ONE change. Measure for ONE cycle.

HYPOTHESIS: If we [change], then [stage] conversion will improve by [%]
EXPERIMENT: [Specific change to process, messaging, or qualification]
MEASURE: [Stage conversion rate, n=20+ deals]
DURATION: [One sales cycle, typically 4-8 weeks]
DECISION: If improved → standardize. If not → diagnose further.

Phase 3 Output: Identified bottleneck, root cause, improvement experiment


Phase 4: Forecasting

Duration: 30 minutes weekly Responsibility: AI-led prediction, human adjustment

Step 4.1: Stage-Weighted Forecast

StageDeal CountAvg ValueStage ProbabilityWeighted Value
Discovery??10%?
Qualified??25%?
Proposal??50%?
Negotiation??75%?
Total pipeline$?

The Sales Forecasting algorithm provides more sophisticated predictions using deal velocity, engagement signals, and historical patterns.

Step 4.2: Commit vs Upside

CategoryCriteriaExpected Close
CommitVerbal agreement, terms discussed, timeline confirmedThis month
Best caseProposal reviewed, positive feedback, decision pendingThis quarter
PipelineQualified but not yet in active evaluationNext quarter
UpsideEarly stage, could accelerateFuture

Phase 4 Output: Revenue forecast with confidence levels


Outputs

OutputFormatDestination
Pipeline health reportMetrics dashboardWeekly sales review
Stage conversion dataCRM analyticsQuarterly optimization
Revenue forecastWeighted pipelineLeadership reporting
Bottleneck diagnosisAnalysis + experiment planProcess improvement

Downstream Consumers

Downstream WorkflowWhat It NeedsLink
Customer LifecycleWon deals entering onboardingLifecycle
Revenue forecastingWeighted pipeline dataBusiness planning
Lead GenerationPipeline gap → volume neededLead Gen

Success Criteria

Quality Metrics

MetricTargetMeasurement
Forecast accuracyWithin 15% of actual revenueMonthly comparison
Stage probability calibrationStage % matches actual conversionQuarterly calibration
Deal aging alerts100% of stalled deals flaggedCRM automation
Win/loss capture rate100% of closed deals have reasonCRM data quality

Performance Metrics

MetricTargetTimeframe
Pipeline coverage3-4x quotaMonthly
Sales cycle length<12 weeks (avg)Monthly
Win rate>30% overallQuarterly
Deal velocity>$X revenue per weekMonthly

Failure Modes

FailureSymptomDiagnosisSolution
Happy earsEvery deal is "looking good"No objective stage criteriaDefine evidence required per stage
SandbaggingPipeline always low, deals appear at closeReps hiding deals until certainCreate safe-to-forecast culture
Pipeline rotLarge pipeline, low close rateStale deals inflating numbersMonthly pipeline hygiene — kill or advance
Stage skippingDeals jump from Lead to ProposalSkipping qualificationEnforce gate criteria before advancement
Single-threadAll deals through one contactOne person leaves, deal diesMap buying committee, multi-thread

Context