How Much Do Forecasts Really Cost Us? Why Accuracy Alone Is Not Enough

In supply chain management and retail, we constantly talk about "forecast accuracy". But measuring accuracy purely in volume units doesn't reveal the true financial impact. In an ideal world, we would track one far more important metric: What percentage of total revenue are we losing due to forecast errors?

Ideal Case: Metric as "Lost Revenue %"

An ideal metric should answer a clear question for every business period: "You lost X% of total revenue due to undershoot (stockouts on shelves) and overshoot (waste, markdowns, and holding costs)."

The Two Sides of Error: Undershoot vs. Overshoot

A forecast error is never symmetric in terms of financial cost. To measure true losses, we must separate errors into two key directions:

What Must a Good Metric Satisfy?

To use a metric for operational decision-making, it must meet three key requirements:

• Focus on revenue drivers: High-revenue or high-capital items must have a proportionally larger weight on the metric.
• Asymmetric scoring by category: High-margin items hurt more when missing, while short-shelf-life items cost more when overstocked.
• Cross-business comparability: Enable comparisons across different periods, stores, or sales channels on a unified percentage scale.

Step 1: What-If Demand Modeling

Before calculating lost revenue, we must estimate unconstrained true demand. Traditional sales logs only record what was sold, not what would have been sold had stock been available on the shelf.

This is where what-if demand modeling comes in. It answers the question: "How much would we have sold if the product had been continuously available under identical conditions?"

A what-if model allows us to:

• Accurately quantify undershoot (lost sales and margin).
• Realistically estimate overshoot (true surplus ending as write-offs or dead capital).
• Establish an accurate baseline for advanced error metrics.

wMAPE with Business Weights as a Practical Approximation

Standard weighted MAPE (wMAPE) sums absolute errors divided by total volume. It solves scale issues and outputs a clean percentage.

Adding business weights to wMAPE:

• Item Margin: Increases the weight of undershoots on high-margin products.
• Overstock Cost: Accounts for waste rates, holding costs, and capital costs.
• Categorization: Treats fresh goods differently than shelf-stable inventory.

A weighted wMAPE combined with what-if demand modeling closely approximates the actual percentage of lost revenue and profit, because we no longer measure how many units we missed, but what it actually cost us.