3 Hard Truths About Forecasting You Need to Know
Every forecast is wrong, accuracy fades over time, and aggregated beats disaggregated. Here are the three hard truths every inventory planner should accept — and why forecasting still matters.
Forecasting is foundational to every supply chain decision you make — from how much inventory to order, to how many staff to schedule, to which warehouse to ship from. Yet most teams operate under assumptions about forecasting that set them up for frustration. Before you build your next demand plan, internalize these three hard truths.
Rule #1: Forecasts Are Always Wrong
No model can predict the future perfectly. Demand shifts, surprises happen, and uncertainty is part of the game.
This is not a failure of your tools or your team — it is a fundamental property of prediction. Consumer behavior is influenced by weather, competitor actions, viral trends, economic shifts, and countless micro-decisions that no algorithm can fully anticipate. Even the most sophisticated machine learning models produce estimates, not certainties.
The goal of forecasting is not to be right. It is to be less wrong than the alternative — which is guessing, or not planning at all.
The practical implication: stop evaluating forecasts on whether they are "correct" and start evaluating them on whether they are useful. A forecast that is consistently within 20% of actual demand is far more valuable than no forecast at all, even though it is technically "wrong" every single time. Pair your forecasts with prediction intervals and safety stock to account for the uncertainty that will always be present.
Rule #2: Accuracy Drops Over Time
The further out you try to forecast, the more variables pile up. Short-term forecasts are more reliable. Long-term forecasts carry higher risk.
This happens because each additional month into the future introduces compounding uncertainty. A one-month forecast only needs to account for near-term trends and recent order patterns. A twelve-month forecast must also anticipate seasonal shifts, competitor launches, economic changes, and supply chain disruptions that haven't happened yet.
Forecast error roughly doubles for every doubling of the time horizon. Your 1-month forecast might achieve 15% error, but your 6-month forecast could easily exceed 30%. Plan accordingly — use tighter safety stock for near-term months and wider buffers for distant months.
The practical implication: weight your decisions toward your short-term forecasts and re-forecast frequently. Rather than locking in a 12-month purchasing plan, revisit and refresh your demand plan every month. Rolling forecasts that are continuously updated outperform static annual plans every time.
Rule #3: Aggregated Beats Disaggregated
Forecasting demand at an aggregated level — such as by product category or region — is typically more accurate than forecasting at the SKU level.
Why? Aggregation reduces the impact of random fluctuations, making patterns more stable and easier to predict. When you forecast a single SKU, every unusual order (a bulk purchase, a return spike, a promotion) creates noise that distorts the signal. When you forecast a category of 50 SKUs, those random fluctuations cancel each other out, and the underlying trend becomes clearer.
Use aggregated forecasts for strategic decisions (budget planning, warehouse capacity, staffing) and SKU-level forecasts for operational decisions (purchase orders, reorder points). This layered approach gives you the best of both worlds.
This does not mean SKU-level forecasts are useless — you still need them for purchase orders and inventory allocation. But it means you should expect higher error rates at the SKU level and plan your safety stock accordingly. The most effective approach is to forecast at multiple levels and reconcile them, a technique called hierarchical forecasting.
So Why Forecast at All?
Because supply chains need direction. Even flawed forecasts guide inventory, staffing, and logistics decisions that would otherwise be made on gut instinct alone.
Forecasting isn't about being right — it's about being ready.
A business without a forecast is reactive by definition. It over-orders when anxious, under-orders when optimistic, and scrambles when reality diverges from hope. A business with a forecast — even an imperfect one — has a baseline to plan against, a framework for measuring deviation, and an early warning system when demand shifts.
Accept that your forecasts will be wrong, that they will degrade over longer horizons, and that individual SKUs will always be noisier than categories. Then build your planning process around these realities: use accuracy metrics like wMAPE to track improvement, backtest your models to validate them, and layer in safety stock to absorb the uncertainty that will never fully disappear.