A Deep Dive into Forecastability and Accuracy

📦Understanding Demand Forecastability

In today’s dynamic supply chain environment, accurately forecasting demand is not just desirable — it’s essential.

Companies that anticipate customer needs effectively can reduce waste, improve service levels, and gain a decisive competitive advantage.

However, not all products are equally predictable. Some have stable, repeatable demand, while others are volatile and sporadic, making them far harder to forecast.

Forecastability measures how reliably future demand for a product can be predicted — in other words, how much confidence we can have in our forecasts.

🔍Key Drivers of Forecastability

Two core metrics determine how forecastable a product is:

1. Average Demand Interval (ADI)

Measures how regularly demand occurs over time.

Low ADI → frequent, consistent demand

High ADI → infrequent, irregular demand

Example:

Low ADI → daily essentials like milk or bread

High ADI → seasonal goods or industrial spare parts

Formula:

Where:

Ti = time between two consecutive demand periods

N = number of demand periods

2. Coefficient of Variation (CV²)

Measures how variable the demand quantity is from one period to another.

Low CV² → stable order sizes

High CV² → large fluctuations in demand

Example:

Low CV² → basic food staples

High CV² → promotional or trend-driven items

Formula:

Threshold Values

To classify demand behavior:

Metric	Threshold	Meaning
ADI	1.32	Regularity cutoff
CV²	0.49	Variability cutoff

Using these, products can be grouped into four distinct demand pattern categories.

🧭 Demand Pattern Classification

Pattern	ADI	CV²	Description	Example Products	Forecasting Ease
🟢 Smooth	< 1.32	< 0.49	Consistent in both timing and quantity	Milk, bread, staple foods	Easy
🟡 Intermittent	≥ 1.32	< 0.49	Irregular timing but steady quantities	Spare parts, seasonal items	Moderate
🟠 Erratic	< 1.32	≥ 0.49	Regular timing but fluctuating volumes	Fashion, electronics	Difficult
🔴 Lumpy	≥ 1.32	≥ 0.49	Irregular timing and highly variable volumes	Capital equipment, project-based services	Very Difficult

Visual Summary

ADI ↓ / CV² →	Low (<0.49)	High (≥0.49)
Low (<1.32)	🟢 Smooth	🟠 Erratic
High (≥1.32)	🟡 Intermittent	🔴 Lumpy

📉 Implications for Forecast Accuracy

Accurate forecasting is vital for:

Production Planning → Match manufacturing to true demand

Inventory Optimization → Minimize excess while avoiding stockouts

Strategic Decisions → Support smarter purchasing and resource allocation

However, forecast accuracy is directly linked to forecastability:

Smooth demand → easier, highly accurate forecasts

Intermittent, erratic, or lumpy demand → requires advanced models and greater safety stock

🧰 Strategies to Improve Forecast Accuracy

1. Match Methods to Demand Type

Use forecasting techniques appropriate for each pattern:

Pattern	Recommended Methods
Smooth	Simple Moving Average, Exponential Smoothing
Intermittent	Croston’s Method, SBA (Syntetos–Boylan Approximation)
Erratic	Machine Learning Models, ARIMA, Neural Networks
Lumpy	Judgmental Forecasting, Scenario Analysis

2. Leverage Historical Data

Analyze past demand to uncover:

Seasonality

Trends

Promotion-driven anomalies

Consistent data cleansing and segmentation improve model performance.

3. Incorporate External Drivers

Integrate macro and micro factors such as:

Economic indicators

Competitor actions

Marketing campaigns

Weather or event data

This external context enhances demand signal accuracy.

4. Continuously Review and Refine

Forecasting is not static — it’s a living process.

Regularly:

Compare forecasts to actuals

Evaluate accuracy metrics (MAPE, WAPE, Bias)

Adjust models and parameters

Continuous feedback loops are key to building resilient forecasting systems.

✍️Key Takeaways

FIn a world where supply chain precision is paramount, Forecast Value Add (FVA) is more

Insight	Impact
Forecastability depends on ADI & CV²	Helps categorize products by predictability
Smooth vs. Lumpy matters	Smooth = predictable, Lumpy = volatile
Tailor forecasting methods	Match algorithms to demand type
Monitor FVA & accuracy metrics	Drive continuous improvement
Iterate & refine regularly	Build agility and resilience

✅ Understanding and managing demand patterns is foundational for achieving supply chain excellence.

✅ By classifying products based on Average Demand Interval (ADI) and Coefficient of Variation (CV²), organizations can:

Select the right forecasting methods

Improve accuracy

Optimize inventory and production

Strengthen responsiveness

✅In short, knowing your demand pattern is the first step toward forecasting mastery — transforming uncertainty into opportunity.

Demand Patterns