Demand Forecasting

Predicting Tomorrow’s Sales Today

🔄 Quick Recall: In the previous lesson, we learned inventory management—reorder points, safety stock, and ABC analysis. All of those techniques depend on one crucial input: knowing how much you’ll sell. That’s where demand forecasting comes in.

Forecasting isn’t fortune-telling. It’s using patterns in your data, market knowledge, and external signals to make educated predictions about future demand. The better your forecast, the less inventory you need and the fewer stockouts you experience.

By the end of this lesson, you’ll be able to:

• Apply three fundamental forecasting methods
• Account for seasonality and trends in your predictions
• Use AI to generate and compare demand forecasts

Why Forecasting Matters

Every supply chain decision depends on demand estimates:

Without forecasting, you’re reacting instead of planning. Reactive supply chains cost more, serve customers worse, and create constant firefighting.

✅ Quick Check: Think about a product you manage or buy regularly. Has demand been stable, growing, or seasonal? How do you currently predict future needs?

Method 1: Simple Moving Average

The simplest forecasting method averages your most recent sales periods.

Formula: Next period forecast = Average of last N periods

Example (3-month moving average):

When April’s actual sales come in (say, 130), the average slides forward: | May | ? | (120+110+130)/3 = 120 |

Strengths: Simple, easy to calculate, smooths random fluctuations. Weaknesses: Lags behind trends; treats all periods equally; ignores seasonality.

How AI Helps

“Here are my monthly sales for the past 12 months: [list]. Calculate 3-month and 6-month moving averages and forecast the next 3 months using each. Which moving average would you recommend for my data and why?”

Method 2: Trend-Adjusted Forecasting

If your sales are growing (or declining), a simple average will always be wrong—it’ll underestimate growth or overestimate declining demand.

Trend-adjusted forecasting adds a growth rate to your baseline:

Example:

• Average monthly sales: 200 units
• Monthly growth rate: 5%
• Forecast for next month: 200 × 1.05 = 210 units
• Forecast for month after: 210 × 1.05 = 221 units

Calculating Growth Rate

Monthly growth rate = (Recent period sales / Earlier period sales) ^ (1 / number of periods between) - 1

How AI Helps

“My quarterly sales for the last 2 years are: [list]. Calculate the underlying growth trend and forecast the next 4 quarters. Show both the trend-adjusted forecast and a simple average, so I can see the difference.”

Method 3: Seasonal Forecasting

Most businesses have seasonal patterns. A sunscreen company sells more in summer. A toy company spikes in November-December. Ignoring these patterns guarantees forecasting errors.

Seasonal indices measure how each period compares to the annual average:

Example:

Using seasonal indices: If your baseline forecast for July is 200 units, the seasonal forecast is: 200 × 1.50 = 300 units.

How AI Helps

“Here are my monthly sales for the last 3 years: [list by month and year]. Calculate seasonal indices for each month, identify the peak and valley months, and forecast the next 12 months combining the trend and seasonality. Present as a table.”

Combining Methods: The Practical Approach

The best forecasts combine methods:

1. Start with historical average as your baseline
2. Adjust for trend (growing? declining? stable?)
3. Apply seasonal indices for each period
4. Add qualitative judgment (marketing campaigns, competitor changes, economic conditions)

This layered approach is exactly what AI excels at—processing multiple factors simultaneously.

How AI Helps

“I sell outdoor furniture. Sales data for 24 months: [list]. I’m planning a 20% off promotion in April and a competitor is closing their online store in March. Create a 6-month forecast that accounts for: (1) historical baseline, (2) growth trend, (3) seasonal patterns, and (4) these market events. Show each adjustment layer.”

Forecast Accuracy: Measuring and Improving

No forecast is perfect. The goal is to be close enough to make good decisions.

Mean Absolute Percentage Error (MAPE) measures accuracy:

MAPE = Average of |Actual - Forecast| / Actual × 100%

Try It Yourself

Run a forecasting exercise with AI:

“Here is my sales data: [paste monthly data]. Please:

1. Calculate a 3-month moving average forecast
2. Identify any growth trend
3. Calculate seasonal indices
4. Create a combined forecast for the next 6 months
5. Rate the likely accuracy and recommend how much safety stock to add”

Key Takeaways

• Moving averages smooth fluctuations but lag behind trends
• Trend-adjusted forecasts account for growth or decline
• Seasonal indices capture predictable peaks and valleys
• Combine all three methods plus qualitative judgment for the best forecasts
• AI can process historical data and generate multi-layered forecasts in seconds

Up Next

In Lesson 5: Vendor Management, we’ll focus on the relationships that feed your supply chain. You’ll learn to evaluate, score, and manage suppliers so you get the quality, pricing, and reliability your business depends on.