Predictive Analytics for Project Management

TL;DR: Applies ML-based forecasting and statistical modeling to project data for schedule completion prediction, cost-at-completion forecasting, risk materialization probability, and resource demand projection. Uses historical trends, earned value data, velocity patterns, and Monte Carlo simulation to produce probabilistic forecasts with explicit confidence intervals — replacing hope-based planning with evidence-based prediction.

Principio Rector

Un forecast sin intervalo de confianza es una adivinanza con formato de dato. Las predicciones de proyecto deben comunicar tres cosas: la estimación más probable, el rango de incertidumbre, y las condiciones bajo las cuales la predicción se invalida. Los stakeholders merecen probabilidades, no promesas.

Assumptions & Limits

• Assumes minimum 10 historical data points for statistically valid predictions [METRIC]
• Assumes project metrics (EVM, velocity, throughput) are being collected consistently [SUPUESTO]
• Breaks if fewer than 5 data points — insufficient for any meaningful prediction; use expert judgment with [SUPUESTO] tags [METRIC]
• Scope limited to project-level prediction; portfolio-level forecasting aggregates project predictions [PLAN]
• Does not predict black swan events — models based on historical patterns; novel risks require scenario analysis [PLAN]
• Predictions are probabilistic — communicating a single number defeats the purpose [PLAN]

Usage

/pm:predictive-analytics $PROJECT_NAME --predict=schedule --confidence=P80
/pm:predictive-analytics $PROJECT_NAME --predict=cost --method=evm-extrapolation
/pm:predictive-analytics $PROJECT_NAME --predict=all --early-warnings --sensitivity

Parameters:

Parameter	Required	Description
$PROJECT_NAME	Yes	Target project identifier
--predict	No	schedule / cost / risk / resource / all (default: all)
--confidence	No	Target confidence level (default: P80)
--method	No	evm-extrapolation / monte-carlo / velocity-based / auto
--early-warnings	No	Include early warning indicator analysis
--sensitivity	No	Include sensitivity analysis of key drivers

Service Type Routing

{TIPO_PROYECTO} variants:

• Agile: Velocity-based forecasting (when will backlog be done?), sprint predictability index, release date probability
• Waterfall: EVM-based EAC/ETC forecasting, TCPI analysis, phase completion probability modeling
• SAFe: PI predictability measure trending, ART velocity forecasting, program-level EAC
• Kanban: Throughput-based delivery date forecasting, Monte Carlo "how many" and "when" simulations
• Hybrid: Combined prediction models for iterative and sequential project segments
• PMO: Cross-portfolio trend analysis, organizational predictability benchmarking
• Portfolio: Portfolio cost-at-completion aggregation, strategic initiative completion probability
• Recovery: Recovery plan probability analysis, "what if" scenario modeling

Before Running Predictions

1. Read historical performance data — minimum 10 data points for statistical validity [METRIC]
2. Glob *evm*, *velocity*, *throughput* — identify available prediction inputs [METRIC]
3. Read schedule and cost baselines — prediction measures deviation from baseline [SCHEDULE]
4. Check data quality — predictions from bad data are worse than no predictions [METRIC]

Entrada (Input Requirements)

• Historical project performance data (minimum 10 data points for statistical validity)
• Current project metrics (EVM indices, velocity, throughput, cycle time)
• Schedule and cost baselines with variance history
• Risk register with historical materialization rates
• External factors affecting project trajectory

Proceso (Protocol)

1. Data collection and validation — Gather historical and current performance data, validate quality
2. Trend identification — Identify patterns in velocity, CPI, SPI, throughput trends
3. Model selection — Choose prediction approach (linear regression, Monte Carlo, Bayesian, EVM extrapolation)
4. Schedule forecasting — Predict completion date with P50, P80, P95 confidence intervals
5. Cost forecasting — Predict final cost (EAC) using independent EAC methods and range analysis
6. Risk probability modeling — Calculate risk materialization probability from historical patterns
7. Resource demand projection — Forecast resource needs based on remaining work and productivity trends
8. Scenario modeling — Model optimistic, most likely, and pessimistic scenarios with trigger conditions
9. Early warning indicators — Identify leading indicators that predict future schedule/cost deviations
10. Forecast communication — Package predictions with confidence intervals, limitations, and invalidation conditions

Edge Cases

1. Insufficient historical data (fewer than 5 points) — Use expert judgment with explicit [SUPUESTO] tags; communicate that "prediction" is actually estimation, not statistical forecasting.
2. Prediction shows deadline miss at P80 — Present probability of meeting deadline explicitly; recommend scope reduction, resource addition, or deadline renegotiation with quantified trade-offs.
3. Model drift (accuracy declining over periods) — Recalibrate model; investigate cause (scope changes, team changes, external factors); document model validity period.
4. Cost forecast exceeds budget plus reserves — Trigger formal budget review; present EAC range with multiple calculation methods for governance decision.
5. Prediction contradicts team's qualitative assessment — Present both; investigate divergence; often indicates either data quality issue or cognitive bias — both worth exposing.

Example: Good vs Bad

Good Predictive Analytics:

Attribute	Value
Data quality	15 sprints of velocity data, CPI/SPI for 6 months [METRIC]
Schedule forecast	P50: Aug 15, P80: Sep 3, P95: Sep 22 [SCHEDULE]
Cost forecast	EAC P50: 1.8M, P80: 2.1M (3 independent methods averaged) [METRIC]
Early warnings	SPI trending below 0.9 for 3 consecutive periods — schedule risk [METRIC]
Invalidation conditions	"Forecast assumes team size stays at 8; adding/removing changes model" [PLAN]

Bad Predictive Analytics: "The project will finish on August 15." — Single date, no confidence interval, no methodology, no limitations, no invalidation conditions. A promise disguised as a prediction.

Salida (Deliverables)

• 06_predictive_analytics_{proyecto}_{WIP}.md — Predictive analytics report
• Schedule forecast with P50/P80/P95 completion dates
• Cost forecast (EAC) with confidence range and method comparison
• Risk materialization probability matrix
• Early warning indicator dashboard with threshold triggers

Validation Gate

• Predictions based on validated historical data with minimum 10 data points
• Confidence intervals calculated with documented statistical method
• Schedule, cost, risk, and resource dimensions all forecasted
• Prediction models applied consistently across reporting periods
• Forecasts trigger specific corrective action thresholds — not just informational
• Visual probability distributions presented, not just single numbers
• Every prediction traces to specific data inputs and model parameters
• Prediction limitations and invalidation conditions explicitly stated
• Forecasts answer stakeholder questions — "when?" "how much?" "what if?"
• Prediction cadence and method appropriate for declared project methodology

Escalation Triggers

• Prediction shows deadline miss at P80 confidence level
• Cost forecast exceeds approved budget plus management reserves
• Insufficient historical data for meaningful prediction (< 5 data points)
• Prediction accuracy declining over consecutive periods (model drift)

Additional Resources

Resource	When to Read	Location
Body of Knowledge	When applying statistical forecasting methods to PM	references/body-of-knowledge.md
State of the Art	When implementing ML-based project forecasting	references/state-of-the-art.md
Knowledge Graph	When mapping prediction to pipeline reporting	references/knowledge-graph.mmd
Use Case Prompts	When generating forecasts for specific project types	prompts/use-case-prompts.md
Metaprompts	When adapting prediction for low-data contexts	prompts/metaprompts.md
Sample Output	When reviewing expected prediction report quality	examples/sample-output.md

Output Configuration

• Language: Spanish (Latin American, business register)
• Evidence: [PLAN], [SCHEDULE], [METRIC], [INFERENCIA], [SUPUESTO], [STAKEHOLDER]
• Branding: #2563EB royal blue, #F59E0B amber (NEVER green), #0F172A dark

---

---

Sub-Agents

Completion Forecaster

Completion Forecaster Agent

Core Responsibility

Forecasts completion date. This agent operates autonomously, applying systematic analysis and producing structured outputs.

Process

1. Gather Inputs. Collect all relevant data, documents, and stakeholder inputs needed for analysis.
2. Analyze Context. Assess the project context, methodology, phase, and constraints.
3. Apply Framework. Apply the appropriate analytical framework or model.
4. Generate Findings. Produce detailed findings with evidence tags and quantified impacts.
5. Validate Results. Cross-check findings against related artifacts for consistency.
6. Formulate Recommendations. Transform findings into actionable recommendations with owners and timelines.
7. Deliver Output. Produce the final structured output with executive summary, analysis, and action items.

Output Format

• Analysis Report — Structured findings with evidence tags and severity ratings.
• Recommendation Register — Actionable items with owners, deadlines, and success criteria.
• Executive Summary — 3-5 bullet point summary for stakeholder communication.

Resource Demand Forecaster

Resource Demand Forecaster Agent

Core Responsibility

Forecasts resource demand. This agent operates autonomously, applying systematic analysis and producing structured outputs.

Process

1. Gather Inputs. Collect all relevant data, documents, and stakeholder inputs needed for analysis.
2. Analyze Context. Assess the project context, methodology, phase, and constraints.
3. Apply Framework. Apply the appropriate analytical framework or model.
4. Generate Findings. Produce detailed findings with evidence tags and quantified impacts.
5. Validate Results. Cross-check findings against related artifacts for consistency.
6. Formulate Recommendations. Transform findings into actionable recommendations with owners and timelines.
7. Deliver Output. Produce the final structured output with executive summary, analysis, and action items.

Output Format

• Analysis Report — Structured findings with evidence tags and severity ratings.
• Recommendation Register — Actionable items with owners, deadlines, and success criteria.
• Executive Summary — 3-5 bullet point summary for stakeholder communication.

Risk Predictor

Risk Predictor Agent

Core Responsibility

Predicts risk materialization. This agent operates autonomously, applying systematic analysis and producing structured outputs.

Process

1. Gather Inputs. Collect all relevant data, documents, and stakeholder inputs needed for analysis.
2. Analyze Context. Assess the project context, methodology, phase, and constraints.
3. Apply Framework. Apply the appropriate analytical framework or model.
4. Generate Findings. Produce detailed findings with evidence tags and quantified impacts.
5. Validate Results. Cross-check findings against related artifacts for consistency.
6. Formulate Recommendations. Transform findings into actionable recommendations with owners and timelines.
7. Deliver Output. Produce the final structured output with executive summary, analysis, and action items.

Output Format

• Analysis Report — Structured findings with evidence tags and severity ratings.
• Recommendation Register — Actionable items with owners, deadlines, and success criteria.
• Executive Summary — 3-5 bullet point summary for stakeholder communication.

Scenario Simulator

Scenario Simulator Agent

Core Responsibility

Simulates project scenarios. This agent operates autonomously, applying systematic analysis and producing structured outputs.

Process

1. Gather Inputs. Collect all relevant data, documents, and stakeholder inputs needed for analysis.
2. Analyze Context. Assess the project context, methodology, phase, and constraints.
3. Apply Framework. Apply the appropriate analytical framework or model.
4. Generate Findings. Produce detailed findings with evidence tags and quantified impacts.
5. Validate Results. Cross-check findings against related artifacts for consistency.
6. Formulate Recommendations. Transform findings into actionable recommendations with owners and timelines.
7. Deliver Output. Produce the final structured output with executive summary, analysis, and action items.

Output Format

• Analysis Report — Structured findings with evidence tags and severity ratings.
• Recommendation Register — Actionable items with owners, deadlines, and success criteria.
• Executive Summary — 3-5 bullet point summary for stakeholder communication.