Why Multi-SKU Portfolio Accuracy Is a Governance Issue — Not Just a Planning Metric
By Paul R Salmon
Supply Chain Council
Executive Summary
Forecast accuracy is often reported as a single percentage — a neat, reassuring number that implies control.
But beneath that headline figure frequently sits a more dangerous problem: systematic forecast bias.
Bias distorts inventory levels, drives unnecessary working capital, masks structural underperformance, erodes executive confidence, and weakens resilience. When multiplied across hundreds or thousands of SKUs, the cumulative impact becomes material — financially, operationally, and strategically.
This paper argues that:
Forecast accuracy must be measured at portfolio level, not just SKU level. Bias matters more than average error. Governance mechanisms should detect sustained bias early. Multi-SKU diagnostics provide significantly better assurance than traditional reporting. Forecast performance is a proxy indicator for organisational maturity.
To support this, the Supply Chain Council has developed the ChainCheck Forecast Bias & Accuracy Diagnostic (Tool 03) — a structured Excel-based portfolio dashboard designed to quantify WMAPE, bias, tracking signal, and top error contributors across the full SKU base.
The issue is no longer “How accurate are we?”
It is:
“Where is error concentrated, and is our planning system structurally biased?”
1. The Comfort of Averages — and Their Danger
Many organisations report forecast accuracy as a single KPI:
“Our forecast accuracy is 82%.”
At face value, that appears healthy.
However, the average can conceal more than it reveals.
Consider a portfolio of 500 SKUs:
20 high-volume SKUs are forecast reasonably well. 200 mid-tier SKUs are structurally over-forecast. 80 critical items are persistently under-forecast. The remainder are low-volume and volatile.
The weighted average may still sit at 80–85%.
Yet operationally:
Inventory builds in the wrong places. Service failures occur where it matters most. Expediting becomes routine. Trust in the forecast deteriorates.
Averages smooth volatility and hide bias.
Portfolio diagnostics expose it.
2. Error vs Bias — A Critical Distinction
Forecast error is inevitable. Demand fluctuates. External shocks occur. Promotions distort signals. Noise exists.
Bias is different.
Bias is structural.
It reflects persistent over-forecasting or under-forecasting driven by behaviours, incentives, constraints, or governance gaps.
Error = Noise
Random variation.
Bias = Behaviour
Systematic distortion.
Persistent bias often stems from:
Incentive structures encouraging optimistic sales projections Supply constraints incorrectly treated as demand suppression Manual overrides without challenge Political smoothing of numbers Poor separation between demand and supply planning Inadequate segmentation (treating all SKUs identically)
While error increases variability, bias creates systemic inefficiency.
The cost of bias is cumulative and compounding.
3. The Financial Impact of Bias
Forecast bias directly influences:
3.1 Working Capital
Over-forecasting inflates safety stock, increases storage costs, and drives obsolescence risk.
A modest 5% portfolio improvement in weighted accuracy can release significant capital.
3.2 Service Performance
Under-forecasting causes stockouts, emergency transfers, and lost trust with customers.
In defence and critical infrastructure contexts, this can translate directly to readiness risk.
3.3 Capacity Planning
Persistent bias distorts:
Labour scheduling Production smoothing Supplier commitments Transport planning
Capacity is either underutilised or overloaded.
3.4 Governance Confidence
When senior leaders perceive forecasts as inflated or unreliable, they discount them.
At that point, the S&OP process becomes reactive rather than evidence-based.
4. Moving from Single-SKU Reporting to Portfolio Assurance
Traditional forecasting reviews often focus on SKU-level reports:
Last month’s forecast vs actual MAPE for a specific product Exception reports
While useful operationally, these do not provide portfolio assurance.
A portfolio-level dashboard introduces structured metrics such as:
4.1 WMAPE (Weighted Mean Absolute Percentage Error)
Reflects true financial and operational impact by weighting error by volume.
More representative than simple MAPE.
4.2 Bias %
Calculated as:
Σ(Forecast – Actual) / Σ(Actual)
Signed bias identifies persistent optimism or pessimism.
Absolute bias indicates scale of distortion.
4.3 Tracking Signal
Tracking Signal = Cumulative Forecast Error / Mean Absolute Deviation
This identifies sustained bias rather than isolated fluctuations.
Large positive or negative tracking signals suggest systemic distortion.
4.4 Top Error Contributors
Portfolio visibility allows planners to identify:
The SKUs driving the majority of error Whether error is concentrated or dispersed Where governance effort should be focused
This is a fundamental shift from descriptive reporting to diagnostic insight.
5. Introducing the ChainCheck Forecast Bias & Accuracy Diagnostic
To support structured governance, the Supply Chain Council developed the ChainCheck Tool 03 – Forecast Bias & Accuracy Diagnostic.
This Excel-based tool provides:
5.1 Multi-SKU Portfolio View
Total SKU count Portfolio WMAPE Portfolio Bias % Portfolio Tracking Signal Top 10 absolute error contributors
5.2 Individual SKU Drilldown
SKU-level WMAPE Bias % Tracking Signal Actual vs forecast trend chart
5.3 RAG Threshold Framework
Configurable tolerance levels Automatic status classification (Green / Amber / Red) Executive-ready summary narrative
5.4 Governance Guidance
The tool includes a scoring guide outlining:
What each metric indicates Common causes of distortion Corrective levers
It is designed not as a forecasting engine — but as a forecast assurance instrument.
Its purpose is to improve decision confidence.
6. Why Portfolio Visibility Improves Maturity
Forecast performance reflects broader organisational capability.
A mature forecasting system demonstrates:
Low sustained bias Stable tracking signals Transparent override governance Clear segmentation (ABC-XYZ) Regular bias review cadence Evidence-based model refinement
An immature system typically shows:
Persistent optimism or pessimism Volatile tracking signals Override culture without accountability One-size-fits-all policies Reactive corrections
Forecast maturity correlates strongly with supply chain maturity.
7. The Governance Shift: From Blame to Learning
Forecast bias diagnostics should not be punitive.
They are diagnostic.
The aim is not to ask:
“Who got this wrong?”
But rather:
“What is the system telling us?”
Bias often reflects structural constraints:
Promotions poorly communicated Allocation rules distorting history Demand and supply signals mixed External policy shifts
A portfolio-level view enables constructive governance.
It supports structured challenge rather than reactive criticism.
8. Implications for Defence and Critical Supply Chains
In defence environments, forecasting accuracy is not simply a financial concern.
It directly affects:
Equipment availability Readiness posture Spares provisioning Sustainment modelling Whole-life cost
Bias in defence forecasting can distort war reserve calculations and maintenance planning.
Small structural errors compound across large fleets and long time horizons.
In such contexts, portfolio-level assurance is essential.
It becomes part of risk governance.
9. Practical Recommendations
Organisations seeking to strengthen forecast governance should:
9.1 Measure Weighted Accuracy
Use WMAPE rather than simple averages.
9.2 Track Signed Bias Monthly
Monitor structural optimism or pessimism.
9.3 Monitor Tracking Signal
Identify sustained bias early.
9.4 Segment the Portfolio
Different policies for:
Stable A items Volatile C items Intermittent demand items
9.5 Focus on the Top Error Drivers
The Pareto principle applies strongly in forecasting.
9.6 Introduce Formal Bias Review Cadence
Monthly governance review with challenge.
9.7 Separate Demand and Supply Plans
Avoid capacity constraints suppressing demand history.
10. Conclusion
Forecast accuracy is not just a planning metric.
It is a signal of organisational discipline, behavioural incentives, and governance maturity.
The question is not:
“What is our forecast accuracy?”
But rather:
“Where is error concentrated, and is our system structurally biased?”
Portfolio-level visibility provides the answer.
When forecast bias is managed deliberately:
Working capital improves Service stability increases Capacity smooths Executive trust strengthens Resilience improves
In complex supply chains, predictability is power.
And predictability begins with confronting bias — openly, systematically, and with the right tools.
