Forecast Cost

The Real Cost of Forecast Error: How to Quote and Cost Demand Planning

Where forecast error actually costs money and how to price it: stockouts, excess holding, expedites, and the fully loaded cost of a planning function.

Forecast error has two price tags, and you must cost both directions. Under-forecasting causes stockouts: the cost per lost unit is contribution margin plus any expedite or substitution penalty, often 20% to 45% of selling price for a lost sale and far more if it triggers a line-down charge. Over-forecasting causes excess: the cost is the annual holding rate, typically 18% to 28% of unit cost covering capital, warehousing, insurance, and obsolescence, applied to units that sit. The Forecast Error Cost calculator asks for both the shortage penalty and the holding rate so a single error number resolves into dollars.

Build the cost-per-unit of error from a real example. Say WMAPE is 12% on a family shipping 500,000 units a year at a $40 unit cost. Roughly half that error skews long and half short. On the excess half, 30,000 units of average overhang at a 24% holding rate on $40 cost is 30,000 times $9.60, or $288,000 a year in carrying cost. On the short half, 30,000 lost units at a $14 contribution margin is $420,000 in missed margin, before expedite freight. That is $708,000 in annual error cost on one family, or about $1.42 per unit shipped.

Expedite premiums are the hidden line that wrecks quotes. When a forecast misses high, planners recover with air freight, overtime, and premium spot buys. Air versus ocean commonly runs 4 to 8 times the freight cost per unit; overtime adds a 50% labor premium and often a 15% to 25% yield hit from rushed changeovers. If 8% of volume gets expedited at a $2.10 per unit premium, that is $0.17 per unit across the whole line. Estimators who quote off standard cost and ignore this systematically underprice volatile SKUs by 5% to 12%.

Cost the planning function itself when you build a business case. A demand planner fully loaded runs $95,000 to $140,000 a year and can own 300 to 800 SKUs depending on volatility and tooling. Add planning software at $15 to $60 per SKU per year for a mid-market suite. So a 2,000 SKU portfolio might carry 3 to 4 planners plus $60,000 in software, roughly $450,000 to $650,000 fully loaded, or $0.90 to $1.30 per SKU-week. Weigh that against the error cost above: if better planning cuts error dollars by even 20%, the payback is usually under nine months.

Do not quote demand planning as a flat overhead line. Allocate it by volatility, because a CV under 0.2 SKU costs a fraction of what a CV over 0.7 SKU costs to plan and buffer. A defensible internal charge splits cost into a base rate per active SKU plus a variability surcharge: for example $30 per SKU baseline and an added $0.50 per point of coefficient of variation above 0.3. That way lumpy, promotion-driven items carry their true share, and the finance team stops cross-subsidizing the hard SKUs with the easy ones.

Overhead and scrap-equivalent losses belong in the quote too. In demand planning the analog to scrap is written-off inventory: obsolescence and markdowns on stock that aged out because the plan was wrong. Healthy operations write off 1% to 3% of inventory value a year; poor forecasting pushes it to 5% or higher. On $8,000,000 of average inventory, moving from 4% to 2% obsolescence saves $160,000. Fold that delta into the cost case rather than treating write-offs as a separate accounting event that nobody owns.

Where estimates go wrong is scope and timing. Estimators cost the forecast at the SKU level but forget the S&OP cycle time cost: every extra week between plan lock and execution widens the error window and raises buffer. Shortening the cycle from 30 days to 18 days can shrink the demand-during-lead-time exposure enough to trim safety stock 10% to 20%. Use the S&OP Cycle Time and Inventory Buffer from Forecast Error calculators together so your quote reflects the process latency, not just the math error, and price the risk you can actually see.

Published 2026-07-01.